<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by David Edwards on Medium]]></title>
        <description><![CDATA[Stories by David Edwards on Medium]]></description>
        <link>https://medium.com/@david.edwards_42657?source=rss-67973d56daea------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*j2RclyOHkPb9sqNuFdqYTA.png</url>
            <title>Stories by David Edwards on Medium</title>
            <link>https://medium.com/@david.edwards_42657?source=rss-67973d56daea------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 02:38:48 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@david.edwards_42657/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com" rel="hub"/>
        <item>
            <title><![CDATA[Portfolio News: Tuesday 30th September]]></title>
            <link>https://medium.com/dave-invest-wards/portfolio-news-tuesday-30th-september-db0960a0b314?source=rss-67973d56daea------2</link>
            <guid isPermaLink="false">https://medium.com/p/db0960a0b314</guid>
            <category><![CDATA[stocks]]></category>
            <category><![CDATA[value-investing]]></category>
            <category><![CDATA[investing]]></category>
            <dc:creator><![CDATA[David Edwards]]></dc:creator>
            <pubDate>Wed, 01 Oct 2025 11:45:00 GMT</pubDate>
            <atom:updated>2025-10-01T11:45:00.667Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>What’s new</strong></p><p>Whilst most of my portfolio focuses on value stocks, I have a small number set aside as “wildcards.” These are the kinds of businesses that Benjamin Graham wouldn’t think twice about, and which “The Intelligent Investor” recommends running away from at full speed. I, however, was curious to see whether taking a few of these might cause some interesting effects on the portfolio.</p><p>I have a couple of different categories, but while I was browsing across different categories, I came across a potential “Growth” stock which caught my attention. It was in the “Health” industry, which I don’t like to begin with because the finance model leads to a lot of small or speculative businesses on the market. This one, however, seemed to have extremely positive analyst sentiment, from its 100% buy recommendation to its insanely high price target. Its name was ATAI life sciences, so I took a look.</p><p>As I looked more closely, there were things which I was extremely concerned about. It had -0.7 earnings per share, and a -5,000% profit margin. What was the fuss about? Looking at what it was working on, it started to make sense — ATAI is pursuing research on psilocybin treatments for depression. We’d watched a documentary a few years ago about hallucinogenic drugs and magic mushrooms and their potential for the treatment of depression, so I was interested, but it seemed that the political environment may be turning in the favour of ATAI — Trump had won the election (but not started office yet), and it was clear that Robert Kennedy, his radical pick for Health Secretary, might be open to these alternative treatments. I took a single unit of ATAI, and if I lost it all, it wouldn’t have been the end of the world, but there was a chance that this speculative pick might develop.</p><p>Fast forward slightly: last week the company reported positive results in its most recent round of testing, and the stock price jumped from $4.60-ish to a peak of $5.84. From my buy price of $1.38, it is up 284%. I’m not a fan of speculative picks, and I’m not saying that anything other than pure luck led me to this stock. Even after I took it, I fully expected it to lose its entire value (to be honest, I still am), but I think it’s a fun experiment to diverge from your usual framework and at least take a look at these wildcard stocks, because occasionally they do behave like this.</p><p>It has since fallen back to 5.29, but I took another unit, just to see what happens…</p><p><strong>What I’ve been looking at</strong></p><p>This week I looked at European Real Estate stocks. Some of these I had seen before, but there were some new Swiss, Finnish, Danish and Austrian stocks, plus a ton of new French and German businesses to look at.</p><p>To begin with, I’m always interested in what the numbers tell about a given industry. In this case, the average Price to Book was 0.873, indicating that there is a fairly strong link between the physical assets of the company and its stock market value. In fact there were only 3 failed tests for Price to Book ratio from the sample I looked at.</p><p>The big problem with this industry was presumably the impact that Covid had. Less than half avoided negative earnings, across the past 5 years, and only 7 of the 48 companies showed EPS growth across both the 3 year and 10 year horizon. In addition, there were a lot of companies which Benjamin Graham would regard as being too small ie they had annual revenue under €500 million.</p><p>To sum up: There definitely weren’t any 6-star stocks, but there were 4 5-stars:</p><p>Merlin (Spain — slightly too small, its annual revenue was €465 million, so just under the tolerance. In my opinion, this is no big deal)</p><p>Unibail-Rodamco-Westfield (France — carrying slightly too much debt to assets, at 67.88%, but again this is not a blowout compared to the 50% reference value)</p><p>Aroundtown (Germany — had a 10 year EPS growth of -10.35% which knocked out its stability)</p><p>Aedas Homes (Spain — Lightyear didn’t display a 10 year EPS growth figure for this stock, and where data isn’t available, I regard this as a fail)</p><p>For the top 3 of these stocks, Lightyear had some analyst data, and Merlin seems very positive: 80% recommended buying the stock, and the price is projected to rise 12% to 14.36. Westfield had a similar upside (+11.39%) but less analyst enthusiasm (2/3s recommended buying it), while Aroundtown had rather cautious analyst views — the price is currently above its expected value by 2% so expected to shrink, and the overriding advice is “Hold.”</p><p>Of the other differentiators that I’m interested in, there was one standout for dividend yield: Aedas and it’s staggering 16.97% — they paid an annual dividend of €2.58 per share last year, and with a price of €21.18, you could get 5 of these for just over €100 netting a cool €12.90 minus tax. Will they pay this much next year? I don’t know, but it’s definitely worth a look. Aroundtown doesn’t pay a dividend which was a big minus against them, especially considering the other 2 paid 3%-ish (Merlin 3.96%, Westfield 3.12%).</p><p>Each of these had their merits then, and though I am less inclined towards Aroundtown, I added all 4 to my Euro diversification matrix. I already own one European real estate stock but it was a wildcard, so I’m quite keen to get a stable stock in the one remaining Real Estate slot.</p><p>Looking at the country diversification, to be honest, I think there are much more attractive German stocks for the 6 remaining German places, so it looks very unlikely that I will take Aroundtown, and the fact that there were 2 strong candidates for the 2 Spanish spots means I will probably lean towards one of Merlin or Aedas. We’ll see what happens, and I’ll let you know which one of these I end up taking.</p><p><strong>What I bought</strong></p><p>Nothing this week, and probably nothing next week, but I am updating my prioritisation logic and shortlist with this influx of new European stocks.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=db0960a0b314" width="1" height="1" alt=""><hr><p><a href="https://medium.com/dave-invest-wards/portfolio-news-tuesday-30th-september-db0960a0b314">Portfolio News: Tuesday 30th September</a> was originally published in <a href="https://medium.com/dave-invest-wards">Dave Invest-wards</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Portfolio News: Monday 22 September]]></title>
            <link>https://medium.com/dave-invest-wards/portfolio-news-monday-22-september-6ce97153de91?source=rss-67973d56daea------2</link>
            <guid isPermaLink="false">https://medium.com/p/6ce97153de91</guid>
            <category><![CDATA[value-investing]]></category>
            <category><![CDATA[stock-market]]></category>
            <category><![CDATA[investing]]></category>
            <dc:creator><![CDATA[David Edwards]]></dc:creator>
            <pubDate>Mon, 22 Sep 2025 11:59:40 GMT</pubDate>
            <atom:updated>2025-09-23T16:23:21.002Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>What’s new</strong></h3><p>This week I want to talk about <em>Intel</em>. For the longest time, this stock has been the absolute stinker in my portfolio. I bought my first block in October 2023 at $34.66 and by the end of 2023 it was over $50 — happy days. It immediately fell back, but I took a second block in July 2024 at $30.74…at which point it more or less immediately crashed and lost another $10 per share. Rather than panic, I doubled my position, and it became the only 4-unit stock in my portfolio.</p><p>I took Intel because in the semiconductor industry it seemed the most stable option. As Nvidia climbed higher and higher, I was discouraged by its price level — currently its P/E is at 50 (my usual threshold is 30) and its “Price to Book” value is an insane 42.85 (I usually hesitate at 2…). Intel seemed to tick all the boxes — it was an established company with sufficiently large sales, relatively low debt, paid a dividend, and it seemed appropriately priced (its Price to Book is currently 1.24, which is much more within my tolerance).</p><p>After the disastrous August 2024 earnings report, it probably wouldn’t have been a stock I would have picked, but I was already in so I decided I should take an additional stake while the price was low. It worked to an extent, bringing my average buy price down to $25.88. If I hadn’t done that, I would still be in the red, but I’d have missed my opportunity to buy this stock at a discount.</p><p>Fast forward 12 months, and this week the stock rallied… big time. We already had some good news, as the US government announced a partnership, Softbank acquired a $2 billion stake in August, and then this week a partnership agreement was announced with Nvidia and it absolutely launched on Thursday. Although it fell back a bit, it finished the week up 21.58% and month on month it’s up 28.5%. It finished the week at $29.58, up 14.3% on my new average buy price of $25.88.</p><p>What did I learn? I think this was a good example of why I use the Benjamin Graham framework and lean towards established giants and steady, defensive growth over high growth stocks. Intel certainly wasn’t too big to fail, and it has a long restructuring and journey back to profitability ahead, but its profile meant that I wasn’t the only one who found it attractive. I don’t know that if something similar happened to a smaller or less well-known industry player, these 3 bits of good news would necessarily have occurred and I would still be counting my losses rather than celebrating the fact that my big stinker has climbed back into the green.</p><h3>What I’ve been looking at</h3><p>Lightyear officially launched Austrian and Danish stocks, so I’m continuing to update my Euro stock watchlist. I noticed that a lot of additional Euro stocks have been added, so I’ve been doing a comprehensive analysis and working through the whole list, recalibrating my diversification matrix in the process.</p><p>This week I finished analysing 2 sectors. The first was the relatively small “Digital Hardware” sector and surprisingly there were not a lot of positive outcomes. Indeed, a solid half of the segment failed to return a Price per Earnings value, meaning they were not turning a profit, and of the remaining companies, only 7 fell under my target value of 30. However, whilst there were no 5-star stocks, I did find my first ever 6-star stock: take a bow <em>Jenoptik</em>. All 6 of the metrics I tested produced values within my tolerance and it even pays a modest dividend, and as it’s the only one above 5 in the segment, it goes into my watchlist.</p><p>“Communication” was the other segment I finished analysing. There were 40 stocks in here and though there weren’t any 6-stars, there were 7 5-stars in addition to Nokia, which I had already analysed earlier. The 7 5-star stocks were:</p><p><em>Ipsos (France)</em></p><p><em>Metropole Television (France)</em></p><p><em>1&amp;1 (Germany)</em></p><p><em>TF1 (France)</em></p><p><em>Telekom Austria (Austria)</em></p><p><em>Proximus PLC (Belgium)</em></p><p><em>Atresmedia (Spain)</em></p><p>All had sufficient revenue, none had experienced negative “eps” in the last 5 years, and all had PE ratios under 30, which was a good start. Atresmedia and Metropole Television offered over 9% dividend yields, and Proximus was my first Belgian stock, so these were particularly interesting.</p><p>Atresmedia was slightly overpriced against its book value, Proximus and Telekon Austria had debt ratios that were slightly outside of my tolerance, and the rest had shown negative 3-year EPS growth, meaning that while they weren’t making a loss, their earnings per share had shrunk across the last 3 years.</p><p>I added them all to my watchlist, let’s see how it goes. My next sector for analysis will be “Real Estate.”</p><h3>What I bought</h3><p>This week, my dollar fund was looking flush with dividend cash, so I took a look in the US market and bought a further unit of VICI. It’s a relatively stable up-and-down stock, bouncing between $27 and $34, but it’s a reliable dividend stock that just raised its dividend to $0.45 per share. According to my prioritisation logic, there aren’t any more stocks in my US watchlist with amazing growth potential or discounted prices, so VICI’s favourable dividend put it at the top of my list. Usually I would have waited and bought in the middle of the quarter, and the price of share immediately dropped once the ex-dividend passed, but I was keen to improve my position to take advantage of the October dividend payout so I paid the premium.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6ce97153de91" width="1" height="1" alt=""><hr><p><a href="https://medium.com/dave-invest-wards/portfolio-news-monday-22-september-6ce97153de91">Portfolio News: Monday 22 September</a> was originally published in <a href="https://medium.com/dave-invest-wards">Dave Invest-wards</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Continuous Localization at Pipedrive]]></title>
            <link>https://medium.com/pipedrive-engineering/continuous-localization-at-pipedrive-e3e276a1499d?source=rss-67973d56daea------2</link>
            <guid isPermaLink="false">https://medium.com/p/e3e276a1499d</guid>
            <category><![CDATA[localization]]></category>
            <category><![CDATA[development]]></category>
            <category><![CDATA[translation]]></category>
            <category><![CDATA[growth]]></category>
            <category><![CDATA[product]]></category>
            <dc:creator><![CDATA[David Edwards]]></dc:creator>
            <pubDate>Tue, 07 Jan 2025 08:19:19 GMT</pubDate>
            <atom:updated>2025-05-16T13:10:10.837Z</atom:updated>
            <content:encoded><![CDATA[<p><em>“The first thing you need to do, the one thing you absolutely must do, is tell your developers to deploy less.”</em></p><p>This was the advice given to me by a localization consultant, hired by the manager of a manager quite a few years ago to review our processes. Pipedrive has somewhere in the region of 11 product development and engineering tribes, each of which is working on missions, tasks and bug fixes simultaneously, meaning that 45 is a conservative estimate of the number of projects being worked on at any given time, and thus the proposed solution was unworkable; a solitary Localization Project Manager can’t just demand that the deployment cycle of an organisation of 250+ engineers and developers be amended to fit their preferences. The challenge, then, is how to lean into an environment in which you receive 10–15 micro-batches of copy a day, and as confusing and daunting as it was for the consultant, Localization at Pipedrive has been effectively working to the cadence of continuous deployment with a process called “continuous localization” for 11 of the 14 years that the company has existed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MH8IcwicT0BelQNfjfrhCQ.jpeg" /></figure><p>Continuous localization feels overwhelming. I can clearly remember when I started working here and seeing apparently an endless line of endless amounts of copy for localization all of which needed to be done immediately (and I certainly remember having a mug of coffee the size of my head like the Localization Manager in the image above).</p><p>Relentless is another way to describe it. There is no version release, no deadline to which you work; the tempo sometimes speeds up, sometimes slows down, but it never ends and it never takes a pause (fun fact: Christmas was often the busiest time for us, as teams and individuals tried to push out their last deploys before the year closed while the Marketing team also made a “Your Year in Review”-style campaign which also needed localization). The outcome was that you felt a lot like Sisyphus, except instead of the boulder rolling down the hill at the end of the day, the boulder got a bit heavier and the hilltop a bit further away.</p><p>So, in view of an overwhelming and relentless workload, how do you manage? How do you build a process of continuous localization that works?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KtWdZVpl2Yu4qQPqYcIUbQ.jpeg" /></figure><h4>Accessible Translation Hub and the Single Source of Truth</h4><p>The first thing is that <strong>you must have a ‘single source of truth’</strong> which is accessible to everyone. This needs to act as a library in which everyone can see what the <em>actual live translation</em> is and now that amending it here will be reflected live. Similarly, if it isn’t in this library, it isn’t getting translated. The system breaks down when such an openly accessibly source doesn’t exist: we had a look at a product where the translations were in a hidden library which needed to be manually updated and within a few months, we noticed that questions like “where is this translation,” “why hasn’t this been translated” and “why isn’t this amended translation showing live” were all too frequent. In most cases, this ‘single source of truth’ will be a Translation Management Platform like CrowdIn, Lokalise or Phrase.</p><p>Once you have established this library, the next crucial step is making sure that everyone knows their responsibility towards this tool. It is the developers’ responsibility to get their strings into the tool, and no further — once developers start ordering translations or, even worse, performing translations themselves, the system can start to creak. We also found that developers and translators often don’t speak the same language when it comes to discussing context. Similarly, the translators were expected to work autonomously according to the guidelines outlined and not to overstep.</p><p>In this case, everyone needs to understand that this is where translations are stored, developers are expected to make sure their strings end up here (in our case, that meant ensuring their content was coded correctly) and translators are expected to go in here and work on their own, flag duff strings as necessary and communicate with the localization project managers. To reiterate, successful functioning of the system is dependent on everyone having access to the tool, but doing so <strong>within the bounds of their own responsibility</strong> (and believe me, the most common cause of incidents was by far and away human error).</p><p>Once this is set up, the Localization Project Manager performs the role of “Overlord of Chaos” because truly, without these unsung heroes, what is a staggeringly efficient system can quickly become unsustainable. Strings go in, translations come out, and ensuring that everything works frictionlessly is what the LPM does.</p><h4>Rolling Backlog</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hbgm7Dftn6OiAOsiVbmBVA.png" /></figure><p>We used to operate (though not any more) a system called a <strong>Rolling Backlog</strong>. Fortunately, a few years ago I (literally) sketched out the complete system when I was super-bored on a ferry to Finland and was onboarding a new team member, but the basic principle is that the translators simply work down a list of “strings for translation” with minimal input from the LPM:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SxDtcWeP-XD2iNwUR3GUFw.png" /></figure><ol><li>Whenever a developer deploys or merges, an automation scans the code and compares it against what already exists in the translation library. Anything new is added to the bottom of the rolling backlog.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wkNghQaOCGfrfstu7Hy_Tw.png" /></figure><p>2. While the new string is added to the bottom, the translators for all languages start working down from the top of backlog. This means their workload can be a little esoteric, jumping from different strings in different projects with every new word, but ultimately they work towards our newly added string.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TukgxDAUGwEV_PxFsl7W4A.png" /></figure><p>3. As the translator works through the backlog and other developers deploy their content, our newly added string moves to the top of the backlog. Our deadline for the journey of this string from the bottom of the rolling backlog to the top, and ultimately complete, is one week, although in practice it usually takes less than this.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-o0YRDs8huRmyEvkBiylcQ.png" /></figure><p>4. The work of the Localization Project Manager is to oversee this list. They hide strings which have found their way into the backlog but which aren’t yet ready for translation (that’s the green/blue string in the diagram), prioritise urgent strings which may need to be translated with a much shorter timeframe (the yellow batch), and also make sure our red new string is completed in one week, ready to be displayed to the world in however many languages are necessary.</p><p>Therefore, the LPM is asking themselves the following questions:</p><ul><li>What new strings are in the backlog?</li><li>Have they been sent correctly?</li><li>Are they ready for translation?</li><li>What is the deadline?</li><li>Do they have appropriate context attached?</li><li>Have they been duplicated/can we use Translation Memory to pre-translate these?</li><li>How is our budget?</li><li>How much money have we used up?</li><li>How much money do we have left for the month?</li><li>Have the 1-week old strings been translated?</li></ul><p>Eventually, as we became less overworked and exerted more mastery of the system, we transitioned into a more traditional ticket/release system with a 1-week turnaround, but the principal remains the same —we accumulate strings in a rolling backlog backlog, and once or twice per week you collect together those that are ready for translation and send them to translators. This system is slightly more manual, but it allows us to have greater control over the quality of translatable strings and management of our budget.</p><h4>Frictionlessness through Automation</h4><p>The key concept behind a successful continuous localization system is <strong>becoming as frictionless as possible</strong>. Every time a manual action is required, it introduces friction into the process, and the more touchpoints you have, the more friction is produced; the more complicated things become, the longer it’s going to take to take and the greater the risk that something might go wrong.</p><p>There are always instances where developers forget to create a ticket or translators forget to notify someone when their translation is complete. In this case, it doesn’t matter because the synchronisation is completely automated, so everyone has less work: the dev doesn’t need to worry about an extra process, the translator’s work is immediately reflected in the live environment, and the localization project manager’s job is no longer running around chasing up different stakeholders.</p><p>Frictionlessness can be achieved through automating the appropriate points of the process pipeline. Our automation point happens at the merge/deploy step in a development pipeline, the logic being that this is an unavoidable step in the work of the developer, one which they cannot skip, and it is somewhere we can seamlessly add the automation. We simply scan the merged/deployed code for content, compare it with what exists in our ‘single source of truth’ and anything new is added and anything missing is deleted. Anything that is new is added to the ‘for translation’ backlog and is released as soon as it’s been checked by the LPM.</p><p>In addition to reducing the complexity of the process, there are other benefits. Developers can focus on building and coding great products rather than ticket management, and by making the interaction between the LPM and developer less transactional, it can become more creative. Now you can have discussions about international usability or multilingual user experience rather than chasing down tickets, and your LPMs are freed up to do more in the pre-translation space. Additionally, you can focus on more labour intensive projects like SEO, webpage optimisation and transcreation of marketing assets.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oo2VPQsIyt5jdDo7Rk2MLA.jpeg" /><figcaption>Real footage of an automated localization process without necessary oversight</figcaption></figure><h4>Risks of a Fully-Automated Continuous Localization Process</h4><p>Rather than engage in a lengthy discussion, I will simply list here the risks and mitigations to be considered when building a fully- or almost fully-automated continuous localization process:</p><ol><li>You have to be able to trust that your Translation Management Platform, or whatever you have selected as your ‘single source of truth,’ will operate reliably and without downtime. This is the centre of the operation, and as such it is critical that it works all the time.</li><li>You have to trust that your developers are well trained, and can work within the bounds of their defined responsibility. Remember, developers are best at coding, and when they start translating, communicating with translators, or even running irregular functions, this frequently causes breakdown (which usually means I am awake at 2am trying to fix things before the start of the next European business day).</li><li>One additional developer error which can occur, especially as the organisation grows bigger and more complex, was what we called a “<strong>deploy tug of war</strong>.” This is where two developers work independently on separate branches of the same code, and each time they deploy they overwrite the deploy of the other. “That’s strange,” says developer 1. “The merge I just made isn’t reflected in the monolith and isn’t showing live. I’ll deploy it again,” developer 2’s strings disappear, they redeploy their strings overwriting developer 1’s strings, and around we go again. Once we were alert to it, it was straightforward to spot, and not necessarily easy but possible to remedy. I think that potentially using the GitHub branching feature or something similar can avoid this, but we built our integration way before this was launched and migrating our process is deemed too risky, but hey, if it ain’t broke, right?</li></ol><p>One of the biggest challenges is getting the balance right between oversight and invisibility ie how much friction you <em>deliberately</em> introduce into the system. An over-automated system is possible but comes with its own risks.</p><p>4. <strong>Waste</strong>: If your translators are working completely autonomously, there is a risk of “duff” strings finding their way into, and eventually the top of, the active rolling backlog. In this case, you can easily end up paying to translate things like “Hello world” or “this is a test” where a dev is simply testing that his connection is working (both of which I do in fact now have in 24 different languages).</p><p>5. <strong>Cost overrun</strong>: If you aren’t monitoring the volumes of strings, it can be very easy for the costs to accumulate and overrun. In this case, you need to have a keen eye on how much cost you have accumulated, how much you have left in your budget and how much you have left to translate, otherwise you may end up going over budget.</p><p>6. <strong>Quality</strong>: If you aren’t actively managing your backlog and adding context or checking that variables or plurals have been correctly formatted, you could end up with either poorly localized texts, or paying double to retranslate when the strings are fixed. Having a rigorous quality check is key.</p><p>7. By far the biggest risk we found was <strong>internal invisibility</strong>: if your developers and translators work autonomously, it can create an illusion of invisibility for your localization team. I remember a few years ago that we were so successful in automating our processes that when my colleague and I gave a presentation to the company about our work, we were overwhelmed by comments of “I had no idea anyone was doing this” or “I assumed it was all magic.” That’s great in one sense, but this invisibility has drawbacks — if something breaks, no-one knows who to turn to to fix things. Additionally, having a well automated machine can cause a misconception at executive level that the localization team doesn’t need to exist or have become architects of their own redundancy, when in fact their role is more important than ever.</p><p>How do you mitigate such invisibility? The answer is certainly regular t-shaped engagement, stakeholder management, and ensuring visibility in other ways through presentation of outcomes. This could mean, for example, taking ownership of collecting and communicating global growth or usage metrics on a quarterly basis, to make sure everyone knows who you are, and why your role is important. Additionally, it is always worthwhile emphasising the team’s output in words translated, projects completed and associated signups.</p><p>Finally, as with any localization project, you need adequate QA controls and guidelines, but this time their importance is heightened. We invited an editor to proof-read everything to ensure that it is correct, and we also conducted a 6-month audit, which became an annual audit, which became an audit we didn’t need except when necessary, but that process of layering and controlling the quality of translation is an important guardrail against the potential chaos of a completely automated system</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*H2yeNshBUMgTj62zl4JSTQ.jpeg" /></figure><p><strong>This has not been touched by AI</strong></p><p>At this point, it would be remiss to publish any article on any topic without mentioning AI, by which I don’t mean the three-toed sloth of the northern Amazon, but artificial intelligence.</p><p>We don’t use AI in our system of continuous product localization. In my opinion, it adds another layer of complexity to an already complicated process, and there is little to win from it. Cost savings, speed to market, scale — these are the potential benefits, but unless your localization operation has started to hit the ceiling of affordability, AI or machine translation is unlikely to be able to offer any significant benefits once you implement the appropriate quality guardrails. Yes, AI can produce a translation in seconds, but by the time you wait for an editor to proof-read and approve the translation, you haven’t won any time over human translation (in which I have a much higher level of confidence). At best, all you are ever doing is reducing the size of the rolling backlog, but like Sisyphus, that boulder will never make it to the top of the hill, so what is the difference in the size of the backlog.</p><p>Maybe there are benefits. We recently used AI to go to market in Hungarian and Thai, and the speed to market was cut from 6 months to 1 week. In this case, the AI was useful not in reducing the rolling backlog, but as a pre-translate of the existing content so that the language teams could get into the backlog faster. This allowed us to launch an unedited beta immediately, but the launch was in fact no faster, given that it took 6 months to proof read the MT content. Given the amount of incorrect translations, I don’t think this is a process we will repeat in the future nor scale into our rolling backlog.</p><p><strong>“Deploy less”</strong></p><p>Sometimes it is possible to agree with your stakeholders to build a system which works for everyone, and sometimes you have to roll with the punches and create something that aligns with what you’re given. In the end, we didn’t need to tell the developers to deploy less; we built a system in which receiving 15 batches of new strings per day wasn’t a problem, and which removed as much friction as possible through automation. This isn’t to say that a continuous localization system should be able to run without any oversight, but as long as you are aware of the risks involved and take the relevant steps to mitigate them, you don’t need to spend sleepless nights worrying about when you will ever get the backlog down to zero: sometimes you just have to accept that the boulder will never reach the top of the hill, but as long as you’re pushing it forward fast enough, that’s all you need.</p><h3>Interested in working in Pipedrive?</h3><p>We’re currently hiring for several different positions in several different countries/cities.</p><h3><a href="https://www.pipedrive.com/en/jobs/open-positions?utm_source=medium.com&amp;utm_medium=ownedsocial&amp;utm_campaign=pipedrive-engineering">Take a look and see if something suits you</a></h3><p>Positions include:</p><ul><li>Junior Software Engineer</li><li>Software Engineer</li><li>Full-stack Software Engineer</li><li>ML Engineer</li><li>And several more…</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e3e276a1499d" width="1" height="1" alt=""><hr><p><a href="https://medium.com/pipedrive-engineering/continuous-localization-at-pipedrive-e3e276a1499d">Continuous Localization at Pipedrive</a> was originally published in <a href="https://medium.com/pipedrive-engineering">Pipedrive R&amp;D Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A Tbilisi Sunrise]]></title>
            <link>https://medium.com/@david.edwards_42657/a-tbilisi-sunrise-e33bcb25821a?source=rss-67973d56daea------2</link>
            <guid isPermaLink="false">https://medium.com/p/e33bcb25821a</guid>
            <category><![CDATA[travel]]></category>
            <category><![CDATA[georgia]]></category>
            <category><![CDATA[tbilisi]]></category>
            <dc:creator><![CDATA[David Edwards]]></dc:creator>
            <pubDate>Thu, 08 Jun 2023 10:56:33 GMT</pubDate>
            <atom:updated>2023-06-08T10:56:46.963Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XZT5YC04NNdndrJfZMn9cQ@2x.jpeg" /><figcaption>The eponymous Tbilisi sunrise</figcaption></figure><p>I love this photo. It was my first photo from my first day in Tbilisi, the first time I had ever visited Georgia. It was taken at 5:30 in the morning, and if the image is slightly hazy, a little blurry, that means it accurately reflects my state of mind at the time. My wife had warned me that flights to Georgia tend to land at weird times, and she was right. I arrived at about 3 in the morning after a transfer through Warsaw, and I didn’t grab a lot of sleep on the plane. I swapped places with a Ukrainian refuge, and I listened to her tell the Russian guy in the row behind about life in Kharkiv, while her son booted the seat tray and spilled water everywhere. I’m always a little nervous the first time I visit somewhere, but coming on my own, to a country and culture I had never visited before, at 3 in the morning, I was a little more nervous than usual.</p><p>The Bolt driver who picked me up from the airport didn’t help — he had a glass eye and a very imposing scar down the side of his face, necking energy drinks as he sped way over the speed limit on the completely deserted road from the airport to the city centre. The area where I stayed was also outside of my expectations — the old part of the city, the streets narrow, cobbled and pedestrianised, a smell of rotten eggs dominating the senses (the area is built on a volcanic sulphur-area with lots of natural baths and spas), and buildings full of character and history. At the end of the street was a thermal spa decorated in a very “central asian” style.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XD3Ol2guF5T1qdvYR46zPA@2x.jpeg" /><figcaption>First impression of Tbilisi</figcaption></figure><p>I definitely felt far from home, but at the same time as being intimidated by this unknown I was exhilarated to be stepping into such an adventure. By the time I reached my hotel to check in (and the receptionist had woken up from my calls), I had too much adrenaline to sleep, so I decided to talk a sunrise walk around the old town.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wKJsW7L7PmehrIJ9JCPMFw@2x.jpeg" /><figcaption>Pre-dawn from Tbilisi old town</figcaption></figure><p>I have something of a “get-up-and-go” attitude when it comes to travelling but anyone who knows me can tell you I’m not much of a morning person, so seeing the sunrise is a pleasure I rarely enjoy. On this occasion, I enjoyed it more than usual. I followed the empty walking trail (which I later discovered to be absolutely crammed full of tourists) to the waterfall. I said hello to the odd lazy cat. I wandered the deserted streets, feeling I had the entire city to myself. I climbed the hill to the castle, savoured the view across the sleeping city, and considered walking to the Mother of Georgia statue before deciding that sleep was beginning to overpower me and it would be better to return to the hotel room for some rest.</p><p>In hindsight, it seems like a dream, the memory filtered through a sleep-deprived haze of pre-dawn morning light and the buzz of travelling outside of my comfort zone. In Europe, and especially within Schengen, we have started to lose that excitement of heading into the unknown — a stamp in the passport, a queue at immigration, the chaos of an airport arrivals hall, a language and script completely alien, an uncertain lingua franca (should I use English or Russian?), a country seemingly outside of normal phone networks (or rather where roaming is wildly unaffordable), a completely unusual currency with a slightly odd exchange rate to remember (1 Georgian Lari = 0.36 Euro, 1 Euro = 2.8 basically 3 Georgian Lari) — all of this was whirring through my mind.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mRhiXyOaM-_0TDr_9z91Fw@2x.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1V1PfsdbBc6Tm2W74MB_Ow@2x.jpeg" /><figcaption>Beautiful Tbilisi</figcaption></figure><p>The following day is equally filtered, but with another familiar but not entirely welcome emotion — the loneliness of travelling alone. My wife was writing her Masters thesis, so she stayed in Estonia, and the flight my friends were taking didn’t arrive until the evening, so the emptiness of the streets somehow mirrored my own loneliness. Even after the city woke up I had Tbilisi to myself, and all of those memories I generated: walking to the Mother of Georgia statue, the visit to the Georgian National Museum, the khachapuri/soup/khinkhali/mukuzani lunch I enjoyed, getting caught in a downpour and dashing into the Holy Trinity Cathedral soaking wet , the conversation I had with the waiter at dinner about Estonia and Georgia and the Soviet Union and our common, unruly neighbour — all of these memories belong only to me. I recalled that this was behind the reason why I stopped travelling alone. travel memories are best shared and revisited. Even when you meet other travellers on the route, once everyone disperses those memories become owned by you and you alone afterwards, a story you can tell but to which others have no stake, can’t relate, weren’t involved in. Somehow it seemed so pointless — “I visited Prague and saw that” but without consequence, unconnected or related to anyone else.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*P8swg0FSA_6vo_l4AZ-w9w@2x.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2He79QVQ53QmFI0RSo-iRw@2x.jpeg" /><figcaption>Running from the rain, a typical Georgian lunch</figcaption></figure><p>The following day, we took a wine tour to the Kakheti region of Georgia with my friends. I can’t remember much of it; that evening we went to the house of a friend of our guide, had a meal with lots of wine and cha-cha, got completely drunk, more drunk than I have been in a long time, and as a result I have little to no memory of the day or evening. Apparently I drank wine from a roof tile; at one point I asked to borrow the family’s guitar while they were performing traditional Georgian music in order to play Don McClean’s “American Pie” — I have only very vague memories of these things (although photographic evidence does exist).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XyKUic56obO0PEeE344ubQ@2x.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qtxHvxXWGw7jvs_HxBynsg@2x.jpeg" /><figcaption>We drank a lot of Georgian wine — though the exact memory is lost, I remember enjoying the evening a lot!</figcaption></figure><p>Yet because I shared that lost evening with my friends (and my friend’s wife was sober, and thus able to take the aforementioned blackmail), the emotional attachment I have to that experience is somehow stronger than the morning I spent enjoying a Tbilisi sunrise on my own. Every time I’m teased about it, though the actual memory is lost, I have retained that feeling of enjoying one the best nights and travel experiences of my life, yet when I see the travel photos of that Tbilisi sunrise, I remember not only the exhilaration of that adventure into the unknown, but also the loneliness of not being able to share it with anyone.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e33bcb25821a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Flashcard App Review: Is Lingvist the answer?]]></title>
            <link>https://medium.com/@david.edwards_42657/flashcard-app-review-is-lingvist-the-answer-c20c6d6ad339?source=rss-67973d56daea------2</link>
            <guid isPermaLink="false">https://medium.com/p/c20c6d6ad339</guid>
            <category><![CDATA[language-learning]]></category>
            <category><![CDATA[language-acquisition]]></category>
            <category><![CDATA[flashcards]]></category>
            <category><![CDATA[language-app]]></category>
            <dc:creator><![CDATA[David Edwards]]></dc:creator>
            <pubDate>Sun, 09 Oct 2022 15:53:21 GMT</pubDate>
            <atom:updated>2022-10-09T15:53:21.545Z</atom:updated>
            <content:encoded><![CDATA[<p>Flashcard apps are no longer my secret weapon when it comes to language study, not because they have gone out of style but because everyone should be aware of them by now. The question is no longer whether you should use a flashcard app, but which app to use. I’ve discussed my dissatisfaction with Memrise and the Little Green Bird in previous posts, but is there a tool on the market better than either of these big alternatives? I tried using Duolingo to learn Portuguese and I gave up, but I didn’t give up on the language — I decided to use it as opportunity to explore a different flashcard app, and I discovered Lingvist.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dzclHa_9yXG0yaeJVNgswA.jpeg" /><figcaption>Lingvist’s Home page</figcaption></figure><p>Lingvist is an Estonian company with a bit of a slow burn. I’d never heard of them, but I discovered them recently on the “Estonian Mafia Wall of Fame” and decided to try them out. Ironically, shortly after I downloaded the app I spotted a car decked in Lingvist livery driving around central Tallinn. In general, the app seems an improvement on Duolingo, but there are a few small tweaks to improve. Nonetheless, I’d recommend it, especially as after 5 weeks of usage, I was confident enough to experiment with the Portuguese I’d learned during a recent 3-week stay in Porto and received warm feedback.</p><p><strong>How Lingvist works</strong></p><p>The app functions by giving you batches of 50 “cards.” These cards either feed you new words you haven’t encountered yet, develop words you are learning or revise terms you “already know.” The result is that you completely relinquish control over your language learning journey to Lingvist’s “advanced algorithm,” and you trust the algorithm to develop your vocabulary as it deems most appropriate. Each card has a small status icon, indicating how “learned” the word is, and comes with a sentence of context and verbs also have an optional grammar explanation (which for Portuguese, is very necessary!). It’s important to note that, unlike in Duolingo, there is no ‘penalty’ for getting a card incorrect, making for a far less stressful learning experience.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DiiA76prOBy2n4q_zJVOMQ.jpeg" /><figcaption>Be careful! A typical “card” in Lingvist (after an incorrect answer)</figcaption></figure><p>The app shows how your knowledge is changing by breaking down your “word-list” into “Known” words, “Words in progress,” “Words at risk” and “Words Mastered,” with a number of “Total Words” encountered. The “Total Words” is also represented as a bar showing your progress within the context of the entire course. The course I am taking is Portuguese (Brazil), and the course length is 5,000 words, so I have a beautiful bar chart showing my progress in different colours reflecting the different stages each word is at.</p><p>The definition of “5,000 words” is a little tricky. When I took my very first steps in Russian, I bought a “my first 1,000 words” book in Riga and it took a good 6 months or so to slog through it on Memrise. I was surprised to see my word count fly past 1,000, until I realised that Lingvist is not talking about word stems but <em>all</em> words, so plural forms and conjugations are counted separately. This means Ligvist’s 5,000 words isn’t quite the same as my definition of 5,000 words, although given the amount of time I’ve spent learning and re-learning different tenses and conjugations of the <em>ser/estar </em>pair, maybe their is some truth, but nevertheless it feels a bit disingenuous to claim I have such a wide vocabulary when quite a lot is based on masculine/feminine or singular/plural forms of the same word, especially when you have mastered the patterns of the language.</p><p>In general, your language learning is very much placed in the care of Lingvist’s algorithm to manage. You hit the learn button, and you are fed your 50 ‘cards’ for the session. If you try to do more than two sessions in a day, a message appears warning that too much language learning can overload the brain and actually slow down the process, which is very much in line with research emphasising consistent, small bites of language and vocabulary learning are more efficient than rapidly burning through a course. This is quite the change compared to Duolingo, which constantly seems like it is trying to ‘hack’ your brain. Lingvist, on the other hand, are up-front about the amount of hard-work needed and by acknowledging that it’s going to take time and encouraging me not to overdo it, they have certainly earned some credit in my books by doing so.</p><p>The interface is very much “down to business,” and down to business certainly seems to be the theme of the app — the deep blue is professional and the layout is very clean and efficient, with the relevant data and infographics to show your progress. This makes a change compared to the cartoon-y faux-accessibility of Duolingo and the (in my opinion) excessive and confusing clutter of late-Memrise. More importantly, also absent are the competitive and gamification elements that define Duolingo: there are no leagues to be promoted or relegated from, no friends or community against whom to compare yourself, no stars or diamonds or gems or powerups — this is vocabulary learning stripped back to its core, which I very much like. It’s not trying to win the race, but training to finish marathon in your own time or improve your PB, which I believe is a fairer representation and more fulfilling approach to language learning.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gLYK8aaBVUvN6v7ot8c9hg.jpeg" /><figcaption>User-friendly, owl-free interface</figcaption></figure><p>That said, given your language learning journey is being managed algorithmically, much of the interface and interactive elements on the page are purely informational — you can see the status of your wordbank and how it is composed, and also review a list of the words you have already encountered, but as there is little agency in your personal ability to direct your learning, there isn’t much to do except hit “Learn” and work through another batch of 50 cards (although remember — no more than 2 per day).</p><p>With the absence of add-ons and powerups and boosters, it is not surprising to find that the pricing is very straightforward: €79.99 for the year, or €9.99 for the month. The language selection isn’t as large as Duolingo (9 languages at the time of writing, including 2 Spanishes and a limited Estonian course), so there are no endangered or made-up languages like Hawaiian or Klingon, but the main world languages are covered and there are quite a few extra languages from which you can learn English, hinting that perhaps in the future Japanese, Chinese, Polish and others will be added. One extra, very welcome feature which isn’t present on Duolingo is the option to create your own deck, so if you are learning a language that isn’t available yet, this theoretically shouldn’t stop you from being able to upload your own vocabulary and using Lingvist to learn it (although I didn’t try this with any languages with non-latin characters, nor is it immediately clear how learning your own language affects the pricing).</p><p>One other function I didn’t try out was the ability to photograph pages in the target language and have these added to a new deck. This seems like a very intriguing feature, opening up the frustrating “next steps” in language learning and allowing you to continue learning through reading, and very similar to an app a friend of mine once developed.</p><p><strong>What I liked and didn’t like about Lingvist</strong></p><p>The experience overall feels like a throwback to a simpler time, kind of how we thought the internet/smartphones would be before monetisation and gamefication came along and loaded us up with lootboxes and excessive notifications. It sits as a welcome companion in my compendium of apps, and my progress is driven more by my own personal motivation to learn than any growth-hack or passive aggressive notifications about how I made the cartoon bird angry, which I really like. I’m a 35 year old, 6-language polyglot, I’m learning Portuguese through a mixture of curiosity and practical interest, and I’m not interested in a lot of the value-added nonsense in Duolingo that becomes insanely infuriating. I went for a road trip one weekend and didn’t get time to use Lingvist and everything was fine; I didn’t screw up any league position, and I certainly didn’t spend any gems re-learning words. I took a trip to Germany for a friend’s wedding and missed a week of learning, and my phone wasn’t hassling me every 20 minutes with “You made Lingvist sad” emails or “Got time to practice Portuguese?” notifications. In that sense, I preferred that Lingvist treated me like an adult, and it seemed like a silent partner, ready to help in my language learning but not constantly trying to barge into by attention or grab my time. I’m sure the Duolingo approach is great for some language learners, but one of the reasons I dumped this app was because I started to develop a complex, feeding a Duolingo-based anxiety of Green owl-flavoured guilt whenever my life inconveniently dictated that I had other things more important than achieving my daily goal or preserving my streak. With Lingvist, I felt as though I was treated like an adult, trusted to go at my own pace and be in control, which is ironic because control is one of the things you definitely don’t have.</p><p>Remember those science fiction movies where the robot is gently rocking the baby to sleep, and there is a kind of “aw” moment followed by a slight unease with how much freedom we’re giving away to the machines and a brief worry that this ends up with The Matrix? (No? Just me?) Well that’s kind of what Lingvist feels like. It is a glimpse into a future where algorithms decide everything for us. “But Dave, it’s just a vocabulary-learning flashcard app” I hear you say, and I get it. Again, depending on your preference, if you’re not such a confident language learner or you are perfectly comfortable handing control of your language learning to the management of the algorithm, this is definitely for you. You won’t need to think about what topic to pick, whether to deepen your knowledge of one subject or refresh your knowledge of an already learned deck — the algorithm will decide that for you, maintain your word-bank in the optimum way, and actually if you can learn to let go, that’s kind of liberating. It frees you to sit back and enjoy the ride (to the extent that rote-learning vocabulary can be described as an enjoyable ride).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-vLGRg1i96TU1ftDOJmAJA.jpeg" /><figcaption>Word slider, the most eye-opening part of the app imho</figcaption></figure><p>As I said, this passive journey might be ok for some, but I prefer being a little more active, so it feels like a slightly less rewarding experience than other apps. Though I haven’t tried any of the other languages, the Portuguese is one big 5,000 word course, which also means that the words come thick and fast and occasionally without any rhyme or reason. I can see that they were uploaded in a formulaic way (eg the months of the year came in a batch, though out of sequence), but I came a little unstuck when I was presented with a group of demonstratives with little distinguishing context (here I had to turn to a Portuguese-speaking friend to guide me through the old <em>this/that/that but you can’t see it</em> debate).</p><p>Additionally, if I’m handing over such control over my language-learning experience, I would like a little more information on the status of that journey. There are infographics regarding “words at risk,” for example, but no way to access what those words <em>actually are. </em>If I were able to see, not necessarily practice but just see, what words were in the “at risk” category, I would feel much more reassured — ok, yes I understand that these are the words I’ve forgotten or that I’m finding tricky, I’ll spend a couple of minutes just browsing through them outside of the main “learning” so that maybe I’m a bit more prepared when they inevitably reappear in the main experience. As it is, that number is more or less meaningless, it’s just demonstrating that the algorithm is in control and it’s another reminder of the amount that you are trusting to be effectively managed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DRTvUHHfgx5yHRNvqSOPZg.jpeg" /><figcaption>What are the 24 words at risk, though?</figcaption></figure><p>The 5,000 word course seems very daunting (as I mentioned, when I learned Russian, I focused on the most common 1,000 and then released myself into the wild), but that big number actually consists of derivatives counted individually. Verb conjugations, plural forms, gender variations of adjectives — these are all counted individually, so although the number will be higher than 1,000, that 5,000 isn’t quite the wide vocabulary of 5,000 unique words as I was expecting, but rather the total number. That’s not a problem <em>per se</em>, but is worth noting if only to encourage you not to be too intimidated.</p><p>One problem that Lingvist does suffer from in common with other flashcard apps (which I alluded to earlier) and which it hasn’t been able to solve, is grammar explanations, in my case with verb forms, conjugations and tenses. I know certain facts about Portuguese grammar from some background reading, such as that in Portuguese you can’t <em>be</em> hungry or in pain, but you <em>have</em> hunger or pain. Such explanations aren’t really provided, so if you hadn’t encountered this before, it would be very easy to confuse the conjugations of the verb ‘to have’ as different ways of saying ‘to be.’ Depending on your preference, that might not be a bad thing — technically speaking, learning <em>I have hunger </em>as “I am hungry” in the source language is perfectly valid, after all, if you’d dropped into the country where no-one speaks English, this is the way you would learn it. Personally, I would prefer a little extra explanation. When the grammar book introduced this topic with the simple explanation “in Portuguese, these feelings are expressed with <em>to have,</em>” it was perfectly logical. Without solving this problem of explanation, I would likewise be confused about why <em>there is</em> phrases were also constructed with the verb to have, and I’d be wondering how many ways of saying <em>to be </em>there actually were in Portuguese.</p><p>It’s the same story with the past tenses — I understand that the contextual information on the card tells me “past perfect” or “imperfect” when asking for a form of “to be” alongside a sentence for context, but maybe it would be cool to be introduced to a little more explanation. Having not learning Spanish, I’m wrestling a little with the two different forms for ‘to be,’ the permanent “ser” and impermanent “estar,” as well as the seemingly endless tenses and forms, and though it is possible to learn this through the flashcard app, I can’t help but feel that with a little explanation, there might be an easier way.</p><p><strong>Conclusion</strong></p><p>Don’t misunderstand me though — I am enjoying Lingvist. Thoroughly. Being treated like an adult and trusted to find the time is a breath of fresh air compared to Duolingo, even if it does mean surrendering myself to the control of the algorithm. If the programme is concerned with deciding what I should be doing when I’m studying and managing what words to feed me rather than harassing me in my spare time or provoking anxiety attacks around what powerups to equip in order to avoid losing my 150-day streak, this is an improvement. This seems to be what the internet and algorithms were supposed to be used for.</p><p>I definitely like Lingvist, and I really feel like I’m making big strides forward with my Portuguese. I’d been using the app for just a few weeks, but I already felt confident enough to try things out. That said, I’m not entirely sure whether we can ascribe all of the success to Lingvist, as some of the early heavy lifting had already been done thanks to aborted Duolingo and offline attempts at language study. I love the “Calendar” and “History” views (if you are nervous about giving the machine control, it is quite satisfying to look at the line of “correct %” staying stable while the “new words” fluctuates), while the “Vocabulary” view is really eye-opening (there is a paragraph of text in the target language, and a slider where you can reveal more words as you learn more — my vocabulary is only 1,149 but I can understand far more than the 64% of the text the feature suggests I should be able to, even with 36% of the words blacked out).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gExBSsOS9GVGZabaik7upw.jpeg" /><figcaption>The “Knowledge Lab”</figcaption></figure><p>I would suggest that there are some features that feel like a work in progress, and others that can be developed. The “Knowledge Lab” doesn’t add a lot of value, and as I previously mentioned, being able to actually <em>see</em> what words were in which category such as <em>at risk</em>, <em>in progress</em>, <em>known</em> etc would add a lot! Although I do kind of mind that I don’t feel in complete control, it’s not the end of the world — I’m an experienced language learner, so I’d probably like a bit more control, but I largely trust the algorithm to direct me and the immediate feedback I’ve been getting from my friends in Porto has been really encouraging. If only I could get them to stop making fun of my Brazilian accent…</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c20c6d6ad339" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How we launched…Traditional Chinese]]></title>
            <link>https://medium.com/pipedrive-engineering/how-we-launched-traditional-chinese-6673c5690616?source=rss-67973d56daea------2</link>
            <guid isPermaLink="false">https://medium.com/p/6673c5690616</guid>
            <category><![CDATA[engineering]]></category>
            <category><![CDATA[localization]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[software-localization]]></category>
            <category><![CDATA[china]]></category>
            <dc:creator><![CDATA[David Edwards]]></dc:creator>
            <pubDate>Tue, 22 Feb 2022 09:28:49 GMT</pubDate>
            <atom:updated>2022-07-22T06:09:46.381Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EpHaqJcGt2yhN59Bblpy7Q.png" /><figcaption>The end result — Pipedrive in Chinese (Traditional), though it was a long road to get there</figcaption></figure><p>Building out the internationalisation of a software company is a step-by-step process. The odds are that you’re not going to begin with a complete or fixed language offering that you build once, release and never touch again. Instead, it’s more likely that your language portfolio is flexible and growing, which means you will probably find yourself in the situation Pipedrive was in during October, with a new language to launch. The translation journey wasn’t something we could do overnight, but thanks to some handy decisions we made and <a href="https://medium.com/pipedrive-engineering/building-a-localization-process-that-translates-into-results-f683419860fa">solutions</a> built into our platform, it was also a fairly painless one (unless of course, you are updating your internal tools at the same time — more on that later!). Hopefully, the following retrospective, based on the key questions we asked ourselves at each stage, will provide some insight into how we at Pipedrive add a new language to our platform from start to finish, and why we chose Traditional Chinese as language number 18.</p><ol><li><strong>Does this language/market align with our strategy?</strong></li></ol><p>According to <a href="https://blog.busuu.com/most-spoken-languages-in-the-world/#:~:text=Well%2C%20roughly%206%2C500%20languages%20are,less%20widely%20spoken%20than%20others.">Busuu</a>, there are 6,500 languages in the world. In 2019 Pipedrive offered 17, which meant there were 6,483 potential languages to add to our software. How did we end up identifying Traditional Chinese as the next language to add?</p><p>Although the process began much earlier, the first step was in December 2019. Within the Localization Team, we made an analysis of our customer base, the competitive landscape in the CRM sphere, looked at market potentials and other indicators, and internally published a shortlist of languages and markets of interest (if you’re interested in learning more, I recently discussed this process on the <a href="https://globalambitions.net/language-expansion-framework/">Global Ambitions podcast</a>). It was pretty important to us that we didn’t just make a random decision in isolation, so we published the list and canvassed feedback from internal stakeholders. Are any of these languages interesting, or are there any other languages that would help you in your work?</p><p>Traditional Chinese wasn’t on the list initially, but it emerged as a language of interest for two reasons. Firstly, we have reseller partners based in both Taiwan and Hong Kong, who were keen for us to support their efforts to sell to non-English speakers in their markets. Whilst that was useful, on its own it did little to distinguish the language from any other; Pipedrive has many <a href="https://marketplace.pipedrive.com/partners">resellers and partners</a> in many languages and markets for which we don’t yet offer a localized product, so although it certainly helped, on its own it wasn’t enough.</p><p>But with recent efforts to increase our presence in Japan and Korea, adding Chinese was seen as an important step in strengthening our position in the East Asian marketplace. With an important strategic focus on neighbouring markets, it became clear that launching Traditional Chinese could be a useful test run for launching its more widely used cousin, Simplified Chinese, at some point in the future. Suddenly, Traditional Chinese seemed like a more attractive proposition.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4TLZ0uJrd6YnVGbvH776qA.jpeg" /><figcaption>Chinese calligraphy</figcaption></figure><p>The decision wasn’t yet made, but it prompted us to look at the language in more detail. Written Chinese went through its first standardisation in the 2nd Century BC, some 2,000 years ago, although it actually dates back even further to 1250BC. Though the spoken language split into the mutually unintelligible variants we know today such as Mandarin (920 million native speakers), Cantonese (84 million speakers) or Wu (80 million native speakers), speakers of these languages all used a common form of the written language until the 1950s. At that point, a process of <a href="https://en.wikipedia.org/wiki/Simplified_Chinese_characters#History">character simplification</a> reduced the number of strokes and characters in the written language, but because of the complex geopolitical situation in the Chinese-speaking world and the dispersal of the Chinese-speaking diaspora, the simplified characters were not accepted everywhere: the simplified characters were adopted predominantly in the People’s Republic of China as well as Singapore and Malaysia, while Hong Kong, Taiwan and Macau continued using traditional characters.</p><p>Though we weren’t quite ready to launch in China, the combined Hong Kong + Taiwan market is 30 million people. Moreover, both territories have advanced software capabilities, technological development and widespread internet coverage, while English ability was assessed as moderate by the <a href="https://www.ef.com/wwen/epi/">English First Proficiency index</a>. Though both territories are in a complicated space geopolitically (and this was most definitely a debate we wanted to stand aside from), from a purely economic and business standpoint, it made sense.</p><p>The language therefore came onto our radar thanks to our internal partners, and through further research developed into something worth considering as a future direction of growth. The deciding factor was that the expertise gathered from this experience would help us use the language as a stepping stone for growth in China, and after considering the costs and return on investment potential, in January 2020, the decision was made to proceed with localization.</p><p><strong>2. Who will carry out the translation? Selecting a Partner</strong></p><p>With the commitment to localize complete, the next step was finding a translator/proofreader team. We outsource our localization to <a href="https://neuma.ee/en/">Neuma translation agency</a>, which made a tender competition. In this instance, we had a couple of extra resources we could draw on: we have a number of Chinese employees who had contacts in the translation industry, and Taiwan-based partners who also suggested a translator team with whom they had previously worked. We invited all of these different teams to enter themselves into the competition too.</p><p>Once we had received the “expressions of interest” and quotes, we made a shortlist. We developed a test project based on a small, recently-launched microservice and one small web page, and sent these to the teams on the shortlist. For us it was important to assess how well the translators worked on both technical texts, as well as creative content. Given that Pipedrive has a very friendly, joking and slightly colloquial tone, we assessed for tone of voice and readability as well as accuracy (although given the cultural norms in different markets, this isn’t completely inflexible — we listened to our translators in other markets if they argued this wasn’t a good idea). We had 2 evaluators independently score the test tasks, one of which was an internal employee from Taiwan and the other was an external vendor who didn’t enter the competition (in order to avoid bias).</p><p>I am aware that certain unscrupulous vendors in creative industries try to use test tasks in an ethically-dubious way. I myself once applied for a translation job in competition, was awarded the contract, and when I set to work on their website noticed that the test tasks (for which I and the other applicants hadn’t been paid) had been published — Free content!! (except it became apparent the vendor was unethical in other ways too — needless to say I didn’t stay in that contract for too long). At Pipedrive we definitely don’t do this!</p><p>Based on the evaluation feedback, we moved forward but there were still a few more steps before the winning team was selected. I’m a strong believer that vendor-client relationship is about more than output or rates, but also about compatibility and collaboration. We want translators with whom we feel that we can work effectively, and most importantly, with whom we can communicate and who feel free to give us honest feedback. It’s not simply a case of selecting the team with the best rates or the best quality, because if you find later down the road that you can’t work effectively together, this is a failure of the recruitment and selection process. Maybe it takes a little longer, but at the end of the process, everyone is confident that they’ve entered into a successful long term partnership, rather than the cheapest or fastest.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tQJX0dueFTMUoc32NipVxA.jpeg" /><figcaption>The island of Taiwan has a population of 23 million and a well-developed economy</figcaption></figure><p><strong>3. What/How to translate your product?</strong></p><p>With the partner selected, it was finally time to dive in and start translating! We use <a href="http://www.crowdin.com">CrowdIn</a> to manage our localization process, and here the advantage of using such a platform becomes clear. We decided to start by translating our product, and managing this in CrowdIn was as simple as going into the Project Settings, selecting the additional language and, as if by magic, the content had duplicated itself and a new language had appeared in the homepage interface with 0% translation progress, a blank slate into which our translators could begin their work.</p><p>One of the additional advantages with CrowdIn is the ability to integrate directly with your product to provide <a href="https://support.crowdin.com/in-context-localization/">in-context translation</a>, and we took advantage of this to allow our translators to see exactly how the texts looked in situ. Our guidelines when translators begin is not to use the main CrowdIn repository but to begin with contextual translations —open the app, look around, play with different features and translate the terms on the pages you can see. In this way, the most important “core” terminology and jargon, which may be unclear when removed of context, will be complete plus the translators get to know the tool, its navigation and visual interface, and more importantly, its value to customers.</p><p>So with the team set up, now it was time for us to sit back and relax with a margarita, right? Not quite. The translation process took around five months, and most of this time we spent answering queries from translators and providing screenshots. Starting with the main framework of the product gave us a quick, visible sign of the progress being made, but as the team buried deeper into individual microservices, the texts became more complex and specific, and trickier for them to reproduce. I always characterise the Localization team as the bridge between translators and developers, and here our product expertise came especially handy. As the project progressed, we were called to answer questions about ever more obscure texts and edge cases in more specialised features, but the end result was a well-translated product ready for the Taiwan/Hong Kong market.</p><p>Once the product translations were complete, we turned our attention to our web experience. As a web-based Software-as-a-Service product, the web is the gateway to the Pipedrive product. Our customers literally can’t use Pipedrive without visiting our web, so we regard our Minimum Viable Translation as including both web <em>and</em> product. This meant that we could queue the development task for activating the new language for product, and direct our translators to start working on our web experience, and even though the product activation was done fairly quickly, we were able to do a “silent release” of the new language inside the Product only. A few eagle-eyed customers noticed and changed their language, but by and large it passed completely under the radar (or at least we didn’t receive a deluge of complaints that the Chinese language was missing on the web).</p><p>Using CrowdIn, combined with Pipedrive’s codestack logic, meant that activating the in-app language is a relatively painless task. We add the language to our Translator client, download the completed translations from CrowdIn where the new language is already present, and then it’s just a case of adding the language to the picker and running a Quality Check to ensure everything is working ok. The Web, on the other hand, turned into another story altogether…</p><p><strong>4. Roadblocks on the Information Superhighway</strong></p><p>When we began the process of translating our Web assets to Traditional Chinese, we ran into a roadblock which was both an opportunity and a massive headache. The tribe responsible for the development and engineering of Marketing assets (of which our public web is one) were in the process of relaunching our Web Content Management System and overhauling our existing web content. This was wonderful because it meant that the developers only needed to add the new language when they were setting up the new CMS (rather than reconfiguring the old one), and the content would automatically begin the send and fetch with CrowdIn. Additionally, as each page underwent its content refresh and the new content was sent to the Translation Management Platform, the Chinese translator team would simply work in parallel with the other language teams, and the Web Publishing team would just publish Chinese alongside all of the other languages when the refreshed content was ready (the only difference being that the Chinese team were translating the page from scratch, while the other teams were only translating the refreshed elements).</p><p>Seamless….except that the CMS overhaul was expected to land in some 3–4 months, and the refreshed content another 2–3 months on top of that, which meant the full language launch would be another 5 to 7 months on top of the 5 months it had taken us to translate the product. From a business perspective, this didn’t make a lot of sense: We had already invested a significant sum in the product translation, and without a webpage this new asset we had just invested in would be sitting dormant. In fact, it wouldn’t even be dormant, as there would be a running cost of maintaining Product translations while we were waiting for the web CMS and content relaunch to be completed.</p><p>Rather than sit around and wait, at Pipedrive we like to come up with solutions, so a compromise was found: we would assemble an elite, strike team to create a single Chinese landing page summarising all of the key information from our other essential, high-traffic pages, and hook this up to our flow for sign-up (for new users) or login (for existing users). In addition, the page would serve as a resource for our Taiwanese partners to contact potential customers.</p><p>With the help of our Conversion Specialist and the Web Publishing team, we used existing content blocks and visual assets to avoid burdening the designers with temporary content and the developers working on the CMS mission, and we stitched together a Frankenstein’s monster of a page; a beautiful hero block, an in-depth pricing block, some testimonials, a “Why use a CRM?” block, and regular sign-up buttons and CTAs (Calls to Actions). I have to say, I love working with my colleagues, but it is when problems and projects like this emerge that you truly appreciate how intelligent, talented and incredible the people working at Pipedrive are. We slightly misused Optimizely and set up the page as a test, directing visitors from Taiwan and Hong Kong to this special page, and voila: we could start to convert customers from these markets and make use of the new asset!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VHFQJQnjYw2NKX2fJQqgJA.png" /><figcaption>An early wireframe of the “Frankenstein” Chinese landing page</figcaption></figure><p>Ultimately, this compromise made everyone happy:</p><ul><li>The developers could continue working on the CMS relaunch without having to worry about building assets for a new page or, even worse, doing the same job twice by adding a new language to a soon-to-be-deprecated CMS.</li><li>The Web Publishing Team had to do a little bit of double work, creating a short term landing page that would be deprecated after a few months, but by recycling assets and content blocks from existing pages, the process was fairly straightforward (or so it seemed to me. If you ask my colleague, maybe she will disagree!)</li><li>The Global Growth and Partners were happy that they could start using this asset to help customers in Taiwan and Hong Kong to sign up immediately, without having to wait for up to 7 more months (taking the lead time to an entire year) for the asset to go live.</li><li>And the Localization team avoided using our budget servicing a dormant asset and could start demonstrating return on investment immediately.</li></ul><p>Then, once the content refresh and CMS missions had landed (six months later), we simply turned off the test, closed the page (which was emotional, like sending your child off for their first day at school), and a shiny, completely translated Traditional Chinese web experience was ready for the customers.</p><p><strong>5. Converting the New Language into Growth</strong></p><p>It was great to translate our product and Web, but putting it live simply gets us to the starting line. As Alex Ferguson once said, “You sign for Manchester United—and that’s when the hard work begins!” and so it is with localizing a language. We had a product; now was the time to start winning customers.</p><p>Our growth strategy for launching new markets depends on conditions in the target area, but in this case, the decision to add the language was largely Partner-based, so once we launched we were quite happy to hand it over to them. That said, there were a couple of steps we took to ensure the maximum impact.</p><p>Firstly, we sent emails and an in-app message to existing customers to let them know that the new language was available. We then carried out a PR exercise to generate wider visibility in the market, and an onboarding flow that ensured our new customers were being properly welcomed and introduced to the product by our partners.</p><p>The impact of adding a new language is definitely not immediate, but we have enough data to know that it causes significant jumps in conversion rates, as well as having a positive impact on churn, so we track the performance at a number of milestones (quarterly, six month, and annual impacts) but with a heavy dose of expectation-setting. However, when we evaluated the 12-month impact, we were extremely happy to see the outcome — a 50% YoY growth, far outstripping other localized, non-localized and East-Asian markets. Such results tend to be sustained, so we are looking forward to continued success in this language over the coming years!</p><p>Additionally, the number of customers switching and using the new language validated our assumptions about how necessary the language was. We had launched Swedish 6 months earlier, but the proportion of users in Taiwan using the Chinese-language app had rapidly outpaced those in Sweden using Swedish.</p><p><strong>6. Conclusions</strong></p><p>Taiwan and Traditional Chinese may not have been an obvious choice for a new language to add to Pipedrive. Certainly, we identified other markets with more favourable conditions, greater growth potential and higher customer ceiling, but the decision has to include other internal stakeholders. Without consulting other interested parties, it would have been very easy to miss what turned out to be a very positive business opportunity. The results have been really pleasing.</p><p>The process was both straightforward and complicated (if that’s even possible). As I’ve mentioned before, using a Translation Management Platform like CrowdIn is going to make your life a lot easier, and so it was here — adding the language, onboarding the translators and communication between our team and the translators was all very seamless.</p><p>We ran into a roadblock when we got to the activation phase of the process, but with a little creative thinking and teamwork, we came up with a compromise that got the language working for us without having to wait for the blocker to clear. Pipedrive is full of really intelligent and talented colleagues, and working on this project together with so many of them, from Growth to Web and Developers to Marketers, always has me walking away feeling full of gratitude for how lucky I am to work in such a place.</p><p>And the results…well, they speak for themselves :-)</p><h3>Interested in working in Pipedrive?</h3><p>We’re currently hiring for several different positions in several different countries/cities.</p><h3><a href="https://www.pipedrive.com/en/jobs/open-positions?utm_source=medium.com&amp;utm_medium=ownedsocial&amp;utm_campaign=pipedrive-engineering">Take a look and see if something suits you</a></h3><p>Positions include:</p><ul><li>Product Designer</li><li>Product Manager</li><li>Software Engineer in DevOps Tooling</li><li>Backend, Full-Stack, iOS Engineers</li><li>Infrastructure Engineer</li><li>And several more</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6673c5690616" width="1" height="1" alt=""><hr><p><a href="https://medium.com/pipedrive-engineering/how-we-launched-traditional-chinese-6673c5690616">How we launched…Traditional Chinese</a> was originally published in <a href="https://medium.com/pipedrive-engineering">Pipedrive R&amp;D Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[2+2+2: The Political Landscape in Estonia after the 17th October 2021 Local Elections]]></title>
            <link>https://medium.com/@david.edwards_42657/2-2-2-the-political-landscape-in-estonia-after-the-17th-october-2021-local-elections-232d57bbd44b?source=rss-67973d56daea------2</link>
            <guid isPermaLink="false">https://medium.com/p/232d57bbd44b</guid>
            <category><![CDATA[estonia]]></category>
            <category><![CDATA[eastern-europe]]></category>
            <category><![CDATA[politics]]></category>
            <category><![CDATA[politics-and-elections]]></category>
            <dc:creator><![CDATA[David Edwards]]></dc:creator>
            <pubDate>Mon, 18 Oct 2021 11:54:22 GMT</pubDate>
            <atom:updated>2021-10-18T11:54:22.939Z</atom:updated>
            <content:encoded><![CDATA[<h3>2+2+2: The Political Landscape in Estonia after the 17th October Local Elections</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*P_DTxDdNF3vQ-b-_DqVWQg.png" /><figcaption>Source: ERR</figcaption></figure><p>The results are in and the outcome looks like….a continuation of the themes from the 2019 General Election. The party system continues to fracture: in the national elections 10 years ago, there was a 2+2 structure of Reform, Kesk, the Isamaa predecessor party and Social Democrats. Today, there is something closer to a <strong>2</strong>+2+2 landscape emerging: 2 “Parties of Power”, with their own fairly broad electorates and appeals + 2 “Rising Star” parties, EKRE and Eesti 200, relative newcomers whose support seems to be growing + 2 existing smaller parties, Isamaa and the Social Democrats.</p><p>The main headlines from yesterday’s vote were:</p><ul><li>The biggest winner of the night were the local election alliances, highlighting above all that local elections are about local issues and cautioning against drawing conclusions about national elections.</li><li>The two main parties continue to be Kesk (Centre) and Reform , who were edged in terms of the total outcome (142k/24.4% to 101k/17.3%), but in the usual battlegrounds, the usual outcomes prevailed.</li><li>There were no major breakthroughs for the 2 “rising star” parties. EKRE (Estonian Conservative People’s Union) posted a solid, if unspectacular result, once again as the third largest party, while Eesti 200’s performances in Tallinn and Tartu were encouraging that they can achieve a breakthrough into the Riigikogu in the next national elections, but struggled to make a broader impact across the country.</li><li>Isamaa and Sotsid continue to struggle, but aren’t quite completely faded yet. Both had broadly disappointing results, but locally retained a few key power bases.</li><li>Green parties, including the Party for the Future, failed to make any breakthroughs</li></ul><p>The opposition party of former Prime Minister Jüri Ratas won the largest number of votes, and Mihhail Kõlvart retained the key post of Mayor of Tallinn, the capital city. Their power bases in the Tallinn suburbs of Lasnamäe, Mustamäe, Kopli (Põhja-Tallinn) and Haabersti, as well as the suburb town of Maardu, appear to be intact. These were good indications that they will continue to be electorally successful.</p><p>There are 2 major concerns for Kesk. Firstly, they lost their majority in Tallinn and will need to enter into a coalition to govern. Will the erosion or attrition of their support in the capital city affect their chances going forward? More importantly is the problem of Ida-Virumaa, and especially Narva. They returned victories in the towns of Sillamäe and Kohtla-Järve, but their performance was decimated in Narva by the Katri Raik list in punishment for an ongoing corruption scandal. For Kesk, the big question is: will those voters who defected to the Katri Raik list return for the national election, or will they return votes for her Riigikogu party, the Social Democrats?</p><p>The 9,000 votes for Katri Raik’s list were double the result of the Social Democrats in Narva last time around, and how they could use that support in 2023. The two red spots on the electoral map showed victories in Hiiumaa and Võru, but despite fielding candidates and attracting votes across the country, they were only able to gather 5% of the vote nationally. With a less crowded field in 2023, it is unlikely they will fall below the electoral threshold, but this was a mixed evening’s results for the party of Indrek Saar.</p><p>Another party hoping for a boost after a slightly-worse-than-expected result is Isamaa. The popular logic is that they have been outmanoeuvred on the political right by EKRE, but the truth is slightly more complicated. The light blue spots on the landscape; Muhu island, Jõgeva, Kohila, Elva and Viljandi vald, demonstrate that the party still holds appeal — they are not a spent force just yet. Nevertheless, they are involved in a tussle on the right, and appear to have become the second conservative party of choice behind EKRE, who seemed to enjoy a good night without breaking through the dominance of the main two powers.</p><p>There were a few dark blue spots on the electoral map, significantly Pärnu, though the dark blue of Põlva and Räpina masks the fact that they won these district by only 16 and 54 votes respectively. The key for EKRE is that they appear to have broad national support and returned deputies in virtually every electoral district. This would seem to indicate the potential for a positive result in 2023, especially if they can continue to eat away at Isamaa support, though this is not necessarily a given, or the important issues for EKRE voters aren’t addressed, namely restrictions related to the pandemic, fears over erosion of an Estonian way of life and the widening rural/urban split.</p><p>The other “rising star” party, Eesti 200, gathered enough support to be encouraged ahead of the national election in 2 years time, but were unable to turn any districts turquoise. Their best performances, in Tartu and Tallinn, underline their appeal as a party for the cosmopolitan, educated electorate, but to make a more significant breakthrough, they need to create a broader appeal. Nonetheless, their agonising 2019 result, narrowly missing out on the election threshold, seems unlikely to be repeated in 2023.</p><p>Internationally, environmental parties have seen a rise in performance, especially in Germany where the Greens won 14.8% of the vote to become the third biggest party. In Estonia, the Greens just about gathered 1% of the vote, while the Party of the Future achieved only 282 votes. The major international trends of a political debate dominated by the climate doesn’t appear to have reached Estonia.</p><p>And what of the party of the Prime Minister, Reform? A big win in Tartu, some districts in Tallinn and around, but third place in the national result with 17.3% — there seems quite some room for improvement. They were the biggest loser from the swing to local election parties as the largely yellow electoral map from the national 2019 elections turned to local greys last night. Will they be punished by pandemic fatigue, especially if, as predicted, new restrictions will be introduced to tackle one of the world’s worst coronavirus outbreaks?</p><p>In the end, these elections didn’t give many answers, but did serve to clarify some of the key questions ahead of the national elections in 2023. The electoral map doesn’t seem significantly different from 2019, notwithstanding Kesk’s troubles in Narva, but rather the trends that were identified back then have continued. Ultimately, in two years time, we will be looking to know:</p><p>How will the Narva corruption scandal affect Kesk in Ida-Virumaa? Will their erstwhile voters migrate to Katri Raik’s Sotsid, and help boost their result?</p><p>Where will those who voted for local election alliances turn? It seems like these were most likely Reform voters who wanted to focus on local issues, but it’s not necessarily certain they will return to Reform, is this the electoral battleground where other parties can edge higher?</p><p>Eesti 200 and EKRE are increasingly seeming like an electoral dichotomy. One draws support from a narrow, cosmopolitan, university-educated demographic, while the other draws broad support from across the country, especially from conservative (with a small c) and rural districts. Although they are at different stages of their maturity in the political system and facing different challenges, the question for 2023 is how far can they continue to develop their electorates and grow their performance?</p><p>Can Isamaa and the Social Democrats re-energise their performances? Both seem to be facing radically different existential fights. On the one hand, Isamaa are fighting against EKRE for the sole voice of the patriotic, conservative right, and are losing ground to the newcomer. They will need to shore up those areas where they still have significant support and start to appeal broadly. Social Democrats are fighting against a much more broader problem experienced by Social Democrats around the world, of eroding support and migration of the working class vote to populist issues. In the same way as Isamaa, Sotsid will need to focus on building a strong rapport locally while also pushing back the tide of international political trends, demonstrating why social democracy in a broader sense should be entrusted with electoral support over populist policies.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=232d57bbd44b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Do I Measure Localization Impact?]]></title>
            <link>https://medium.com/pipedrive-engineering/how-do-i-measure-localization-impact-53915c703f9b?source=rss-67973d56daea------2</link>
            <guid isPermaLink="false">https://medium.com/p/53915c703f9b</guid>
            <category><![CDATA[translation]]></category>
            <category><![CDATA[business-growth]]></category>
            <category><![CDATA[internationalization]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[localization]]></category>
            <dc:creator><![CDATA[David Edwards]]></dc:creator>
            <pubDate>Wed, 23 Sep 2020 13:02:30 GMT</pubDate>
            <atom:updated>2020-09-28T08:24:08.688Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IP_6hMFzBbSjl5jE4J8kbw.png" /><figcaption>Image from Local Measure</figcaption></figure><p>One of the questions I’m asked most frequently when discussing localization, really cuts to the heart of one of the biggest misconceptions around what we do. <strong>What is the impact of Localization, and how do you prove your return on investment?</strong></p><p>It’s a really important question to ask, a question all managers should bear in mind when assessing their teams. A question all leaders should be asking when weighing whether or not to invest in something: <br> — How do I measure<em> </em>the impact of an activity? <br> — How do I define success, and how do I know if it’s been worthwhile?</p><p>The reason why I’m asked this question so often is that <strong>localization isn’t a process that delivers a clearly-defined, directly measurable impact</strong>. There isn’t a single business lever or metric that localizing or “localizing better” affects. There is no single statistical Performance Indicator that tweaking your localization output will directly impact, nor is there a barometer that allows you to clearly see the impact for the business of translating/localizing your product.</p><p>For others, translating your product into a new language is regarded as a “low-hanging fruit” or an express route to growth. If I go to the market with an English-only app, my customer base is limited only to people who are comfortable working in English. Alternately, if I choose to add a language and localize my product into Danish, my reach has now been extended to people who are comfortable working in English or Danish, so logically I can infer that my company will grow in Denmark and I can monitor growth as my KPI, right? In fact, this is where the misconception lies, because <strong>the only <em>direct</em> impact of my activity is that my potential customer base has grown</strong>. I haven’t actually done anything to get those customers to use my product yet...</p><p>Put simply:</p><p><strong>Localization does not mean Growth</strong></p><p>Yes, translating your product greatly enhances your growth potential, but it’s up to the rest of your operations to take advantage of this extended reach. My beautifully localized product won’t reach any customers in Denmark, for example, if I don’t carry out any marketing in Danish. If I choose to offer my product but not my website in Danish and I don’t carry out any localized marketing… well, my customer base has been extended a little. I’m able to now reach Danish-speaking customers who are comfortable browsing the web in English, but wouldn’t be comfortable using an app on a day-to-day basis in English, <em>which is slightly better than nothing</em>, but it’s still not going to make a large impact to your bottom line.</p><p>This therefore makes localization a difficult beast to manage, and presents a problem for any business leader considering localization or extending the dimension of their business into markets where their primary language isn’t spoken.</p><p>If localization doesn’t automatically translate into growth, if I don’t know exactly what the potential will be in advance, if I can’t justify or monitor the business impact, and if it isn’t possible to keep a close eye on return on investment — how do I know if it’s worthwhile doing at all?</p><p>A solution requires a slight change in how the team and its impact is visualized. <strong>Instead of thinking of the localization team as having a <em>direct impact</em> over <em>specific</em> levers, the team instead has an <em>influence</em> over a <em>wide range</em> of metrics, </strong>right down to (but far beyond) the bottom line of your company. <strong>Localization amplifies or enhances the performance of the company by extending its reach into markets and areas that can’t be reached using only one language, but it doesn’t do much on its own</strong>. It is therefore necessary to monitor a whole suite of performance indicators, while at the same time acknowledging that Localization is not responsible for any one of them. Complicated, right? Let’s take a look at an example.</p><p><strong>Thinking out Loud?</strong></p><p>Let’s use a metaphor. Imagine Ed Sheeran standing with his acoustic guitar and playing his music. In this metaphor, your product is the song he’s singing, and your customers are the people who hear it and stop to listen. If he’s busking on the street without an amplifier, people are going to pay attention, but the reach isn’t going to go very far, basically only as far as his voice will carry. If he wants to step up and start playing in bigger venues such as a bar or a club (read: enter a new market), he can continue playing with just his acoustic guitar, but the people at the back aren’t going to hear him so well. That’s like entering a market without localization. You’re in a bigger room with a bigger capacity, but your voice isn’t any louder — you’re not reaching any extra customers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KvUO18AEvKFc4WrPoQ085g.jpeg" /><figcaption>Can you hear what he’s singing? Time to invest in an upgrade! (Copyright: Heather Buckley)</figcaption></figure><p><strong>Localization then, is like plugging a guitar into an amplifier</strong> — now that everyone at the venue can hear, more listeners are reached — the more languages you add, the bigger and better the amplifier so the bigger the venues you can play (ie the bigger your potential customer base). Before you know it, the song (your product) is still the same, but you’re on the headline stage at Glastonbury with hundreds of thousands of fans, and none of it was possible without localization.</p><p>But what was the impact of buying an amplifier? If you’ve upgraded your sound system to accommodate Wembley Stadium but you’re still busking on the High Street — the potential and capacity to play to bigger audiences is there, but you haven’t actually done anything to grow your fan base — you’re just playing very loudly (and probably expensively) to the same customers, probably much more loudly than you need to. If the venue is empty, it’s not necessarily the amplifier’s fault, it’s projecting the music as it should do, it’s just that it’s not being heard. The deficiency lies elsewhere, either people don’t like the song (your product) or they haven’t heard about you yet.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/578/1*_eyMK4Mj2aMQF7JF1hWXBA.jpeg" /><figcaption>If you’re not growing in a market, don’t wreck your amps like Pete Townsend — it may be something other than localization</figcaption></figure><p>This is where marketing and other activities come in. You still need to get radio play, cultivate Spotify and Youtube channels, do chat shows, movie cameos and all the rest of it if you don’t want to be playing to an empty room. It’s fine being able to accommodate Danish-speaking customers or Ed Sheeran fans, but they still need a level of brand awareness before they start to make an impact in your bottom line.</p><p>The sound system? Simply having the amplifier doesn’t grow his fanbase, it just allowed him to play to bigger crowds. At the same time, <strong>none of it would have been possible without the amplifier</strong> — he’d be standing on stage at Glastonbury, but only the front few rows would be able to hear him. The amplifier then, enabled him to deliver on all those other activities, bring the music to more people, allowing him to perform at bigger and bigger venues and eventually build a global fanbase.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*wtY15DvL0IHywZnCFs1KFQ.jpeg" /><figcaption>Localization didn’t write the music, but none of this would have been possible without it (Copyright: Alamy)</figcaption></figure><p>On its own, the amplifier’s impact isn’t that noteworthy, but the music, exposure, walk-on part in Game of Thrones — none of these other activities would have been able to deliver the benefit of it. <strong>Localization is thus an essential element combined with your product, marketing and other activities, which on its own doesn’t do much, but without, it’s impossible to grow to anywhere near your full potential.</strong></p><p>There’s no shortcut: to really get the most from the pool of potential customers, you’re going to need deliver an entire suite of localized assets — webpages, marketing materials, and onboarding materials. To <em>really</em> make the most of the investment, everything you do in English needs to then be transferred and available in the target language. <strong>Every extra activity adds to the impact, but every asset you opt-out of restricts the performance of your business in the target market.</strong> What’s more, there’s no way of knowing exactly how much it negatively affects. The return is proportional to the scale of your commitment — the more you opt-in, the more you get back.</p><p>There is, however, one experiment to easily visualize this impact. Run two different Google ads (or any other campaign) for the same length of time — one only in English, and the second, an identical campaign with Danish added (or whichever target language). Now you can measure the impact difference in the overall click-through rate and visitor numbers, and you can see the added visitors from the second market or country. The exact impact of localization will be immediately visible as a number, a boost in your reach and we’re no longer dealing with “potential customers.” The translation of the campaign has amplified its message and reach in a quantifiable way.</p><p>Let’s go back to the original question, then. <strong>How do I incorporate Localization into my business, and how do I measure its impact?</strong></p><p>Start by visualizing the customer’s journey from start to finish. Go beyond just your product and consider how a customer discovers that you exist, right through to the very end when their money arrives in your account. Then at every step, consider how you can reach customers who have a different first language, or live in a target market where they speak that language. It becomes slightly more work, but the growth potential makes it worthwhile, and by following the principles I laid out in my previous article on <a href="https://medium.com/pipedrive-engineering/building-a-localization-process-that-translates-into-results-f683419860fa?source=friends_link&amp;sk=e6e49c44265c231bda8944d70adc3ff7">scalability</a>, the added workload isn’t actually so high.</p><p>Organizationally, from my experience, the best-performing and most impactful Localization Teams either work directly with every part of the business (in a similar way to your Design Team for example), or in close collaboration with some kind of International Operations team. In a way, your translators are like a high-caffeine energy drink or espresso for your revenue — you’ll be able to perform without it, and localizing will definitely have an impact (even if you can’t pinpoint precisely how much or where), but simply drinking an espresso and waiting for something to happen isn’t going to achieve any results directly.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_tO4yWohpracd143NhvDYg.jpeg" /><figcaption>Full of energy, now time to get going!</figcaption></figure><p>At a risk of laboring the point: Imagine your product as the lead singer/songwriter in your band — this is what the fans come to see. Your Marketing team is the lead guitarist, turning the amp up to 11 and busting out the explosive solos and catchy riffs that draw the listeners back and distinguish it from everyone else. The Localization Team are like the bass guitarist. You can’t measure how much better the final product is with them, but you’ll have an inferior song without them and you’ll definitely notice that they’re not there. With an Internationalization team as drummer, completing the rhythm section, you’ve removed all the limitations to growth and opened up the path to global success.</p><p>To measure the impact? Take all of the levers and metrics you would examine for the overall health of the business, apply them to the single market and compare before and after. In theory, as your potential customer reach has expanded, localization will provide a small boost to each activity you would normally perform. But remember, the responsibility ultimately lies with whichever team is performing the activity. That amplifier gave you the extra listeners, but you still need to play the song through it for anyone to be able to hear it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=53915c703f9b" width="1" height="1" alt=""><hr><p><a href="https://medium.com/pipedrive-engineering/how-do-i-measure-localization-impact-53915c703f9b">How Do I Measure Localization Impact?</a> was originally published in <a href="https://medium.com/pipedrive-engineering">Pipedrive R&amp;D Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building a Localization Process that Translates into Results]]></title>
            <link>https://medium.com/pipedrive-engineering/building-a-localization-process-that-translates-into-results-f683419860fa?source=rss-67973d56daea------2</link>
            <guid isPermaLink="false">https://medium.com/p/f683419860fa</guid>
            <category><![CDATA[development]]></category>
            <category><![CDATA[process]]></category>
            <category><![CDATA[translation]]></category>
            <category><![CDATA[localization]]></category>
            <category><![CDATA[translation-services]]></category>
            <dc:creator><![CDATA[David Edwards]]></dc:creator>
            <pubDate>Wed, 20 May 2020 13:03:05 GMT</pubDate>
            <atom:updated>2020-05-20T13:03:05.411Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/338/1*FwmY_XHCQBnLEMeEU2e-cA.png" /></figure><p>Building a localization process from scratch or implementing a localization process into an existing piece of software can seem like a daunting prospect. Maybe you want to expand your growth and decided to dip your toes into a new market with a different language, or perhaps you acquired a product with a very different language profile— either way, going from 0 to a fully localized product is something that you are likely going to need to break down into steps.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*47n6Vu34zdnKAJ88liFCDA.png" /><figcaption>Look at all those languages! But implementing and maintaining a localization process requires a lot of work and preparation</figcaption></figure><p>Before you embark on your localization adventure, remember that adding languages is not a “quick-win” and it isn’t going to magically deliver thousands upon thousands of new customers to your bottom line. It can take some time before those shiny languages on your language picker begin to turn into results, but with a stable base and steady investment, it has the potential to connect you to new clients. Getting your localization process right is therefore the first step on a long journey, but as with most things, getting those first steps right can make the difference between success and failure</p><p>At Pipedrive, I can identify 4 key principles in our localization process, and whilst our solution isn’t perfect, I think it works pretty well. Like any process, there is always a need for a balance between speed, cost, and quality. Getting your process right requires considering all three and finding the balance that suits you best.</p><p><strong>Outsource</strong></p><p>My first recommendation is to outsource. I’ve heard examples of companies taking a Do-It-Yourself approach to localization, such as asking employees to localize the product into their mother tongue. Whilst this has the advantage of saving costs, there are major disadvantages to this approach. Firstly, simply being a native speaker of another language doesn’t necessarily mean that you’re going to become an expert translator overnight. Translation is a skill that’s developed over many years, people conduct Masters and PhD studies into the art of translation, so simply asking your new software developer or support agent to translate your product into Serbian isn’t necessarily going to produce something that Serbians are going to be happy with, regardless of how well they speak their first and second languages. Secondly, in the long term, you’re going to have trouble maintaining the translations if your temporary Serbian translator goes back to their regular job and no longer has time to localize all those cool new features they’re developing.</p><p>The solution is therefore to outsource. Fortunately, there are loads of translation agencies, with expertise and experience in sourcing excellent quality translators from whatever language you choose. Remember, it’s not possible to know everything, and that’s fine — the best solution is to find someone who does, and can help you out at a price you can afford.</p><p>Finding the right localization partner can also be a minefield. There are numerous things to consider, such as cost and location, but ultimately it is up to you to pick the one that feels like the best fit. There might be a cheap agency, but if they’re based in a vastly different timezone, you’re going to have communication problems that may affect speed and quality. Speak to as many companies as you can to compare rates, size and your relationship, then pick accordingly.</p><p>Once you’ve found someone to translate your software, you then need to consider <em>how</em> they’re going to carry out the translation. For this, I would also recommend outsourcing. There are a lot of localization platforms on the market, from Lokalise to CrowdIn, and the expertise of these platforms is worth leaning on. Sure you can build your own in-house portal, but when it comes to plurals, gender, variables as well as import/export, remember an established localization platform has identified and solved all of these problems before, so invest in their expertise and enjoy the ride.</p><p><strong>Automate</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*0qIbASYdpDEvWpMd" /></figure><p>Once you’ve got a translation agency on board and you’ve signed up to a localization platform, you’re going to have to consider how to get your texts onto the platform. I’ve seen companies complain that their localization became out of date very quickly, and the main reason is that they didn’t automate. Sure, you can get the intern to upload and download new strings manually, but at some point they’re going to have to get coffee or they’ll find a new job, and suddenly you realize your app’s Spanish texts haven’t been updated since Limp Bizkit were cool (they were cool at some point, right?).</p><p>This problem of maintenance can easily be solved by enlisting our trusty friends, the robots — no, not Kraftwerk, but an automation script that sweeps through your code, automatically uploads new texts and downloads translations into your product. The good news is your localization platform usually has a well-integrated API (remember, they were solving these problems back when you were at university watching the Lord of the Rings and Hobbit trilogies back-to-back and trying to figure out how to pass Sociology because the lectures were at 10am and you overslept the entire course) so it shouldn’t be super-complicated to implement. It also frees you up to focus on delivering more complex projects with a greater impact. You may have to invest a little bit of time, but spending some more time building a long-term solution that stops you from worrying about localization is worthwhile, and believe me, your intern will thank you — just think of all the extra coffee they can bring you if they don’t have to worry about doing this manual stuff anymore!</p><p><strong>Manage</strong></p><p>So you’ve set up your localization process to automatically send and fetch to your chosen platform, and your agency has found some amazing translators who are working on your software in between their real passion, translating obscure poetry or trying to improve on the authoritative <em>War and Peace</em> translation by researching archaic, lost meanings. Job done, right? Not quite.</p><p>Now is the time for the real hard work to begin. Start by opening up a three-way communication channel between your developers (see my <a href="https://medium.com/pipedrive-engineering/localize-this-dceb57ca69d8">previous article</a> on considering localization when writing code), your translators (who have zero experience of software development), and your customers, the guys who pay you for the end product. Your localization process is not going to be perfect immediately, nor are the translations regardless of how good your translators are, so you’re going to need to tweak, learn and improve where you can. The good news is that you’ll see the gradual improvements the closer and closer you get to perfection, and at some point you’ll have an (almost) self-managing ecosphere of (almost) perfect translations. This will free you up to…</p><p><strong>Scale</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/561/0*VhefsyowoTdyD3pi" /><figcaption>Start with the essentials, then build and expand your localization horizontally and vertically as your company grows and your process matures</figcaption></figure><p>Start from the beginning, start small, and when you’re ready, move to the next stage. This could mean either adding a new language to your portfolio (extending horizontally), or finding new content (extending vertically). Probably the best place to begin is with your actual product, as there’s no point localizing your website into Arabic if your product can’t support Right to Left, unless one of your hobbies is infuriating Egyptians (in which case I would recommend finding a new hobby because this isn’t a very nice thing to do). Once you’ve got your product localized, you can start to think about the next area that will demonstrate results and target that. Maybe it’s marketing texts or maybe it’s your knowledge base — ultimately, you need to decide for yourself based on your own interpretation of impact and results, but given your experience it should be less painful than the first time.</p><p>The other option is to begin expanding your language portfolio. If this is something your translation partner has experience doing it shouldn’t be too demanding, and unless you’ve got one super-talented linguist translating into multiple languages it shouldn’t have a significant impact on your delivery time. Do try to be careful about over-extending too quickly into too many languages as there’s nothing worse than adding one language to discover you’ve enraged their neighbours and fierce rivals by not adding another. Also, vertical scaling becomes more complicated and more expensive as you have to write exclusions. As with most things in life, the key is balance and priorities — having a product with 100 languages but marketing only in English probably isn’t going to fuel your growth as much as you were hoping, and those investors who you promised such incredible growth to may not be so keen to wait for speakers of those languages to find your product themselves.</p><p>There is no right or wrong answer when it comes to setting up a localization process. Each decision has benefits and disadvantages. It’s up to you to weigh those up and make your decisions accordingly.</p><p>The four principles I recommend to establish a successful localization process are <strong>Outsource</strong>, <strong>Automate</strong>, <strong>Manage,</strong> and<strong> Scale, </strong>but this is based on <em>our </em>experience and interpretation of what the best solution was for us regarding speed, quality and cost. But maybe your needs are different — maybe you want to personally vet your translators, or keep everything manual so that you have oversight on everything going to the translators. Either way, at every step, consider what your priorities are, and what impact each decision you make is going to have, but I guarantee that seeing users discover your product on the other side of the world will make the whole thing worthwhile.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f683419860fa" width="1" height="1" alt=""><hr><p><a href="https://medium.com/pipedrive-engineering/building-a-localization-process-that-translates-into-results-f683419860fa">Building a Localization Process that Translates into Results</a> was originally published in <a href="https://medium.com/pipedrive-engineering">Pipedrive R&amp;D Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Is it possible to learn a new language using ONLY Duolingo?]]></title>
            <link>https://medium.com/@david.edwards_42657/is-it-possible-to-learn-a-new-language-using-only-duolingo-bd8920bb4721?source=rss-67973d56daea------2</link>
            <guid isPermaLink="false">https://medium.com/p/bd8920bb4721</guid>
            <category><![CDATA[duolingo]]></category>
            <category><![CDATA[hawaiian]]></category>
            <category><![CDATA[language-acquisition]]></category>
            <category><![CDATA[language-learning]]></category>
            <category><![CDATA[language]]></category>
            <dc:creator><![CDATA[David Edwards]]></dc:creator>
            <pubDate>Wed, 01 Apr 2020 14:49:56 GMT</pubDate>
            <atom:updated>2020-04-01T14:52:05.295Z</atom:updated>
            <content:encoded><![CDATA[<h3>Is it possible to learn a new language using ONLY Duolingo? Language learning in the time of COVID-19</h3><p>I decided to embark on an experiment with the popular language learning app Duolingo, to see the extent you could master a new language with only the help of flashcard prompts and a friendly green bird…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/958/1*nZqmOJ4Qluo92-gWXmsAJA.png" /><figcaption>Honolulu’s Iolani Palace, in the distinct Duolingo style</figcaption></figure><p>About three weeks ago, though I didn’t quite realise it at the time, I was lucky enough to be on a vacation right at the moment before the world changed completely. It was our second visit to Hawai’i, and even more than the first, I was starting to grow an interest in Hawai’ian history, and in particular the Hawai’ian language.</p><p>Hawai’ian language is everywhere in Hawai’i. It’s packaged for tourists with the ubiquitous “Aloha” (hello) and “Mahalo” (thank you), as well as “E komo mai” (welcome) and “A hui hou” (see you soon/til we meet again) if you pay slightly more attention. One is completely surrounded by cultural and environmental phrases, from the frequent requests to practice “ ‘āina” (respect), and names of places, animals and foods. Street names are mostly Hawai’ian, as are the names of towns and cities. Pearl City, for example, is squeezed between Waipi’o, Waipahu, Waimalu and ‘Aiea. Hawai’ian music draws heavily on Hawai’ian language, though it is also a hybrid of Western instrumentation and structure with enchanting Hawai’ian melodies.</p><p>I started to dig a little deeper, though, and began to realise that the language isn’t quite as healthy as it seems. I noted, for example, that nowhere on O’ahu, Mau’i or Kaua’i did I hear Hawai’ian language being spoken, even amongst groups of Hawai’ians. That struck me as odd, given that in the UK I’d visited a pub in Bala in which the entire audio milieu was Welsh (with English swearwords, of course), and that I’d been so entranced by a Gaelic conversation in Oban, Scotland that I’d bumped into the two ladies talking, who both also apologised in Gaelic. But upon further investigation, it seems this was observation about Hawai’ian was not unusual — according to the US census in 2011, there are only 2,000 first language speakers and 24,000 fluent speakers left, of whom most live on Ni’ihau, the westernmost Hawai’ian island, which is more or less completely closed.</p><p>Leaving Hawai’i, I resolved to do more; to learn about this beautiful but endangered language. I read about Samuel Elbert and Mary Pukui, the titans of modern Hawai’ian linguistics. I bought their 1979 <em>Hawai’ian Grammar, </em>and I was amazed by new concepts such as duplication, and how the language only has 13 letters including the ‘okina (commonly known as the glottal stop, or glo’al stop), and how it incorporated English loans despite having only 8 consonants. I was interested in morphology between Hawai’i and other Polynesian languages such as Maori, and I was amazed that when, in 1778, Captain Cook landed as the first Westerner to visit Hawai’i, he brought a Tahitian who was able to communicate with the Hawai’ians despite the fact they had been isolated for over 500 years.</p><p>But as I returned to Estonia, I went back to work and my interest started to wane, fading into one of the those eccentric holiday memories you put in the cupboard along with your suitcase and suncream. And then fate intervened…</p><p>I’m an experienced language learner. I speak German, French, Finnish, Estonian and Russian, along with Czech and Slovak, which I learned and promptly forgot, and my native English. I also have different learning experiences and motivations. Like every English student, we were forced to learn French, as well as German at school. I learned Czech and Slovak at university, and by the time I studied Finnish for my Master’s, language learning apps were starting to become a thing. Gone were the days of carrying that little blue pocket-sized “vocabulary book” or buying hundreds of physical flashcards, dropping them on the floor, spilling coffee on them and eventually throwing them out en masse when you move house. By the time I reached my PhD, market was mature and there were a number of different options available.</p><p>The language I regard as my most successful language learning experience was Estonian. It represented the perfect set of circumstances — as part of my pre-PhD, I was required to spend a year learning, amongst other things, Estonian language, so I had time, focus, and intensive language study. During this period I dedicated hours to uploading vocabulary in droves onto Memrise, and way before I’d ever heard of Charles duHigg, I built myself the habit of going to the campus coffee shop every day at 15:00, buying a giant cup of coffee, and simply sitting, learning vocabulary. This was accompanied by two classes a week, 2 intensive three-week language schools, regular conversations, both formal and informal, the driving motivation of <em>needing </em>to master the language in order to complete the PhD, and of course, the dopamine hit of shocked Estonians who had never heard of a foreigner learning their language (I have many such stories). Within 18 months, I was translating fiction and conducting research interviews in Estonian. I recognise how fortunate I had been because when I came to learn Russian, certain elements were missing, and the results were less striking. I’d achieved proficiency, even mastery, but the language journey was (and still is) uneven, with corresponding outcomes.</p><p>When I began learning Russian, though, a friend recommended a new tool: a free app, which had apparently revolutionised language learning, called Duolingo.</p><p>Duolingo consists of pre-loaded language courses with a pretty impressive 35 languages to choose from for English speakers. The lessons are structured around themes such as family, food, or weather, each theme consists of 5 levels, each containing 3–5 lessons. Each lesson takes about 5 minutes, and is formed of 15–25 questions based on recognising phrases, constructing sentences using building blocks, and eventually translating sentences entirely, in both directions. The app is free, but it makes money by showing you adverts between each lesson sub-section (I still haven’t “killed any goblins,” I still don’t want visibly clearer skin and I still haven’t signed up to a course to improve my mastery over the “power of attraction,” in case you were wondering), while there are also incentives to pay for and a “Duolingo Plus” that offers other benefits. You’re given hearts which you lose if you make a mistake (don’t worry, it usually recognises if you’ve just made a typo), and you can replenish these with “gems” which you can either buy or earn by completing lessons, or wait five hours per heart. If your hearts run out, you lose the progress in that lesson and need to either replenish or wait.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RQ_rgaZdLIf4IgOW75LnjQ.png" /><figcaption>The Duolingo interface with Crowns, Streaks, Heart, Themes, Lessons, XP</figcaption></figure><p>In addition, there are all sorts of gamification elements, which seem mandatory for building an app these days, including experience points, badges, learning streaks and leaderboards (though I’m clearly quite cynical about all of this, I still religiously watch the leaderboard to check my progress).</p><p>Honestly, when I learnt Russian, I was pretty disappointed with it, so I endured for a little while before deleting it. But in Hawai’i Duolingo hovered back into view. Scroll beyond the big hitters such as German, Chinese and Spanish, and you’ll discover some smaller languages such as Swedish or Hebrew. Keep going, and a whole new world opens up. Invented languages such as High Valyrian from Game of Thrones, Klingon from Star Trek, or Esperanto are listed alongside endangered or indigenous languages. Being from the UK my interest was piqued by Irish, Gaelic and Welsh, all of which are available, as well as Navajo and Hawai’ian. For me, it’s amazing that Duolingo regard saving endangered languages as being part of their mission, especially as technology can be such a potentially transformative influence on languages. Intuitively it is easy to think of technology such as the internet enhancing the spread of dominant languages and speeding the demise of minority languages, but it also has the potential to save and revive such languages. It was already clear that Hawai’ian suffered from a lack of educational resources even in Honolulu, and in Estonia there was absolutely nothing, but with the power of the internet it was possible to learn even here. If I wanted…</p><p>3 days after we returned, the company I work for instructed us that we were to work from home. 5 days after we returned, the whole country went into lockdown. Amongst the flurry of “how to survive at home” guides, one recommendation was to keep yourself occupied and take up a new hobby, and the “what if…” seed was planted. Still tanned from our vacation, still slightly bleary-eyed and jet-lagged from the 30 hour journey and 12 hour time difference, I pulled that not-yet-dormant interest from the cupboard and asked myself the question: <em>What if</em> I downloaded Duolingo again? Would it possible to learn Hawai’ian from scratch using <em>only</em> Duolingo?</p><p>This seemed like an interesting experiment, to learn a language in isolation without any other resources. I was stuck at home so I had a lot of time, and there was no better opportunity to build a new habit and no excuses for breaking it. I decided to do it. 2 weeks ago, I reinstalled the app and began.</p><p>I’ve currently completed nine themes out of thirty, and I’m ready to give some initial observations. The first positive is that it is pretty impressive to be able to sit in Estonia, in coronavirus lockdown, learning Hawai’ian language. Above all else, I’m still amazed that I have this opportunity.</p><p>By virtue of the fact I’ve completed 9 lessons, I can conclude that I’m making progress. Each lesson builds on structures learnt previously, and I can recognise words when they are repeated in later lessons. In that sense it’s clear that learning using Duolingo has delivered noticeable benefits in building a base vocabulary in the target language. I’ve also hit upon a number of “tangibles,” things I can point to as having learned: Imperatives and plurals, for example. It seems like verbs don’t conjugate (though how this translates into tense and aspect, I’m still waiting to find out), and I discovered in lesson 1 that Hawai’ian has some form of vocative. Also, because Hawai’ian doesn’t have the letter D and every consonant needs to be followed by a vowel, my name is actually Kawika in Hawai’ian. I could now imagine my football teammate in 2006 who was learning Czech and used to yell “Davide!” when he wanted my attention, would actually have been yelling “E Kawika!” if he’d been learning Hawai’ian. Tangible!</p><p>The technique of learning a language using <em>only</em> Duolingo gives a unique colour to the language learning journey. The sense is that I’m learning a language in the same way as language pioneers, ie through repetition, without explanation or context, and slowly accumulating layer upon layer of linguistic knowledge. In the absence of grammar explanations, my brain is having to fill in the gaps and build grammar rules concerning the structure of the language on its own. On the one hand, that’s quite fun and inspiring. On the other, though, it feels like I’m being treated like a child and being forced to acquire language like a child, and that’s frustrating. As an experienced language learner, I’ve identified what my most efficient linguistic acquisition technique is: my Estonian method. This Duolingo-only method is stalling or slowing down my progress. Now that’s not necessarily Duolingo’s fault, rather it’s a flaw in the method, but at the same time it’s still annoying. Science, of which linguistics is one, is a process of building upon the work of those who went before you, but here I have no help or guiding hand, which has led to repetitive errors which even further decelerate the learning experience.</p><p>A number of examples spring to mind, but syntax seems a good place to start. Hawai’ian word order consists of Verb Subject Object (as opposed to English word order of Subject Verb Object). Because I have learnt this through learning phrases, this hasn’t been explained in Duolingo. For all I know, that could be wrong, but actually I read it in Pukui and Elbert’s <em>Hawai’ian Grammar</em> in Vancouver airport. I remember at the time thinking “wow, that’s pretty crazy!” It’s lucky I did read that sentence, because without it I’d be struggling, but even so, interrogatives and particle phrases emerge and change word order, and I haven’t quite picked up the pattern yet. Maybe it’s because I’m only nine lessons into a 30 lesson course, but I would really appreciate some clarity on this point.</p><p>The problem is compounded when things like this happen:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9vKStentQt2vDSdlG9ewLg.png" /><figcaption>“You are correct” — there’s the endorphin rush</figcaption></figure><p>“The people don’t want it” behaves exactly as expected, according to the grammar rules I’ve built myself.</p><p>Negative Marker + Verb + Subject:</p><p><em>‘A‘ole makemake ka po’e</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yI9-2QTn_gjCgfK8XB3Uvw.png" /><figcaption>The rule I created isn’t quite right</figcaption></figure><p>“I don’t want it,” however, doesn’t behave in the same way. The grammar rule I constructed: Negative marker + verb+ subject would suggest</p><p><em>‘A‘ole makemake au</em></p><p>but Duolingo marks this as incorrect. The correct answer is:</p><p><em>‘A‘ole au makemake</em></p><p>with the pronoun <em>au </em>sitting between the verb <em>makemake </em>and the negative marker <em>‘a‘ole. </em>In fact, whenever a personal pronoun is used, the word order goes Negative marker + Pronoun + Verb</p><p>Why? Without an explanation, it’s impossible to know</p><p>The case is similar with the verb “to be.” I’m used to functioning without it, Russian doesn’t have it, at least in the present tense. I have identified in Hawai’ian that there are numerous ways to express “to be” using word order and particles such as <em>Aia </em>and <em>He</em>, but again, without any explanation, I’m somewhat in the dark. I know that this causes errors, which eats up my hearts and slows down my progress, but I can’t <em>fix</em> it, it needs to come naturally, and all the while it costs me XP points and drives me down the leaderboard.</p><p>Hawai’ian seems to be full of particles such as <em>he, ‘o, o </em>and <em>i</em>, and while I can recognise that they are performing some linguistic function in the sentence, using just a flashcard app, it’s not immediately clear <em>what</em> <em>exactly</em> that function is. The particle <em>i</em> seems a good example. It usually appears in the second half of sentences, so I can infer that has some kind of effect on the object of the sentence. But in some instances it disappears, and I can’t identify the difference, for example in <em>kēia lā </em>and <em>i kēia lā</em>, which both seem to mean “today.” Similarly <em>o </em>sometimes appears and sometimes performs a possessive function, but sometimes isn’t present and I can’t figure out why. Furthermore, the English prepositions <em>in </em>and<em> on </em>seem to be translated as both <em>i </em>and<em> ma </em>and used interchangeable; it’s not clear what the difference is: both <em>i lumi moe</em> and <em>ma lumi moe</em> are accepted as correct translations for “in the bedroom.”</p><p>Though this lack of clarity is a result of my Duoling-only method, Duolingo itself isn’t helping. I ought to be learning right now. I have two more lessons to complete before I finish today’s quota, but I’m stuck. I’ve run out of hearts, for some of the reasons outlined above, and the “practice to replenish hearts” function is broken. The explanation for hearts given in Duolingo’s Help Center is that they deliberately discourage users from excessive progress, which can damage language acquisition. They are certainly slowing down my language acquisition, to the extent that it is currently a handbrake. Theoretically I could pay for Duolingo plus, which would give me unlimited hearts, but it would no longer be a free app. Actually, I quite like the idea of “you’ve made too many mistakes, now slow down and review what you’ve already learnt,” but when the function is broken, it doesn’t work.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wozLb9mF-zSf1uSsub2Zdg.png" /><figcaption>Trying to reload some hearts, something went wrong…</figcaption></figure><p>As a language professional who encounters machine translations regularly, one of the limitations of technological language learning aids is how arbitrary it can be. Whilst Duolingo has done a pretty good job of safeguarding flexibility when it comes to translations, it hasn’t completely erased it. Whilst it was probably user error which marked my translation of “It is raining in Hilo” incorrect in favour of “Hilo rains,” more broadly speaking I find that attempts to render translations naturally in English are occasionally punished. In the absence of wider explanations, it’s not possible to know whether I violated a specific rule, or whether the software made an arbitrary decision. Time will tell, I suppose.</p><p>After one third of my experiment, what are my initial observations? Honestly, I’m feeling really excited about my progress. It’s a fun linguistic challenge to pose to your brain — let’s dive into this world with no context and see how far we can go. On the other hand, I’m desperate for grammar explanations, especially given this language is so different from English, or Estonian or Russian. I need greater context about what is wrong and why, but I also very much appreciate working at my own pace. If Duolingo offered some inline grammar explanations for a few dollars extra, I would snap it up immediately. I could supplement my learning by reading Pukui and Elbert at the same time, but in the interests of the experiment, I will persevere without.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kO_XZ5Cjn11UBIQa5qpKXg.jpeg" /></figure><p>It definitely feels like Duolingo would work better in a more fully rounded language learning ecosystem, though. The app is a limited tool for drilling language and vocabulary. It probably wasn’t designed to replace the other elements of the language learning experience. Indeed, I can recognise that the drills that form the basis of the Duolingo experience would be a fantastic addition, reinforcing vocabulary and grammar learnt elsewhere and accelerate a learner’s general progress. In isolation, it feels like harder work than learning a language actually ought to be. Whatever is maddening currently about the platform would be significantly less maddening if this wasn’t my only gateway to the language, the only way I can make progress, and if I also had access to a teacher who could clarify the grammar points, validate my hypotheses and practice my speaking.</p><p>This is not to say that learning a language through Duolingo can’t yield very real outcomes. My vocabulary in Hawai’ian is growing, and my interest has been sustained. I started recognising linguistic cognates when watching The Fast and the Furious: Hobbs and Shaw (forgive me) and The Rock began talking in Samoan to his family. That was a very real lightbulb moment, and I definitely felt the endorphins from positive reinforcement washing over me. Certainly some skills are improving, and though my grammatical awareness is improving, I can’t shake the feeling that I’m honing one very specific skill — the ability to recognise and reproduce specific texts. Within a wider language learning framework, Duolingo would probably fit in better as a complementing factor; on its own, I’m learning to be a parrot, and I’m sceptical about my ability to independently construct sentences outside of the Duolingo environment.</p><p>The inability to add one’s own content, for me at least, is a limitation and especially when I was learning Russian, led to a disconnect between what I was learning on the platform and what I was learning elsewhere. On the other hand, I recognise that for language learners with less time to craft their own language journey, saving hours uploading words and simply focussing on learning them, is probably a blessing. In that sense, we’re very fortunate that dedicated linguists took the time to programme Hawai’ian language into Duolingo for us to take advantage of.</p><p>This thought allows me to appreciate the bigger picture. Duolingo have performed a very noble service by helping to preserve this language. They’ve invested their own time and money, an investment from which they will probably never yield any direct monetary dividend, into preventing an endangered language from becoming fossilised. In reality, they’ve given it a very significant chance of revival in the 21st century, and they’ve broadcast that chance so that curious language learners on the other side of the world can explore this bewitching language on their own terms. In a cynical world, I’m less inclined to view this as a marketing stunt, and leaning towards the idea that this is an app built by dedicated linguists who have a passion for spreading and preserving languages. Whatever limitations the platform has regarding language acquisition outcomes in isolation, this wider accomplishment most definitely needs to be applauded, and complemented by traditional language learning methods, there is no reason why Duolingo can’t be a very useful tool on a learning journey for those without the time to invest in crafting their own.</p><p>I will post an update at the end of the journey in a few months time to reflect on the final outcome and compare with Memrise, but until then <em>Mahalo</em>!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bd8920bb4721" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>