We talk a lot about optimizing our processes, but it's rare we get an inside look into how it's done. In this episode, Anna Lenik, Localization Team Lead at LetMeSpeak, tell us how they brought their localization workflow from Google Sheets to 24-hour turnaround times by focusing on vendors, tools, and timelines. Listen in to learn how it was done.

Audio

Text video

Key takeaways

• Prioritizing tools, guidelines, context, quality control, and development integration can help optimize localization processes.

• Establishing a dedicated localization team and permanent vendor relationships improves quality over one-off freelancers. Managing vendors directly instead of an LSP allows more control, though now a project manager assists with the growing workload.

• Transitioning from spreadsheets to specialized translation tools helped streamline LetMeSpeak's processes. The development team sets up the localization environment and selects content to translate within the tool, with English as the source.

• Optimizing timelines, such as delivering all updates within 24 hours, requires addressing issues like time zone differences and more.

• Involving the localization lead in product design gives perspective on the cultural and linguistic adaptations needed. Providing context like screenshots and string comments is also important for translators.

• Machine translation is avoided for UI due to context dependencies. Translators review and score machine translations to confirm the level of post-editing needed.

• Guidelines cover both technical and linguistic aspects like tone. They vary by language region (e.g. European vs. RTL vs. Asian).

About Anna Lenik

Anna is a user research and localization expert. When this interview was recorded, Anna was a Localization Team Lead at LetMeSpeak, a language-learning platform that helps people from 140 countries learn English. Currently, she owns a boutique user research agency to help companies connect with their user base in a profound way. To learn more and connect with Anna, visit her LinkedIn page.

Enjoyed the Localization Process Pod? In every episode, we'll be learning from one guest about the way they do localization for digital products. Subscribe here, on LinkedIn, or on the Localization Station website to get notified about the next ones.

Transcript

Michal: Hi everyone. I'm so glad to see you here for one more episode of the Localization Process Pod. I started this podcast to learn all the real hands on things that happen when companies get their products localized and we're five episodes in and we've already uncovered some incredible processes and advice from those brave enough to take the microphone first.

To If you haven't done so yet, make sure you give the previous episodes a listen. And if you have a localization process to share, send me a message. I would love to talk to you too.

If it's your first time here, then this is how it goes.

In every episode of the podcast, we meet with someone from a different company to grill them on what they're doing right and what they're doing wrong and how they're optimizing their processes for a better localized experience.

Today, I'm going to talk to Anna Lenik, Localization Team Lead of LetMeSpeak.

Anna will tell us about how they managed to remove bottlenecks and optimize their process to get localized content turnaround to 24 hours. If you've been listening to me for a while, then you may know I don't think speed or cost should be our sole metric for success, actually far from it. But there is still a lot to learn from this about making localization processes more efficient.

And I'm hoping you'll find this as valuable as I do. So hi, Anna.

Anna: Hi. Thanks for having me here.

Michal: Tell me a little bit LetMeSpeak and what you do? It's a really interesting approach.

Anna: So at let me speak, I'm Head of Localization and Education. My main responsibility is product development for any educational features.

It can be a challenge because in online, the motivation is really decreasing rapidly. That's why we have to think of better ways to teach people.

And I'm also a team lead in the localization team. We've got 16 languages. And users in 140 markets. So your localization flow, it's pretty large.

When I talk about myself, I usually just say that I am a linguist because that's my passion. I started my career in localization in 2012, so a long time ago.

A little bit about us. LetMeSpeak is a language learning... medium for everyone who wants to learn English. We've got a lot of apps integrated.

Like we've, we've got iOS and Android apps, a web app, and An NFT marketplace as well. So we have both users in Web2 and Web3 environment, and they all learn English together in this medium.

One of our unique concepts is that we give financial rewards for learning English. And that's what makes us unique in that market.

Michal: You say you have a few platforms, do you design the experiences themselves or is it more like a marketplace that people design experiences for LetMeSpeak?

Anna: No, we design everything and develop all the products ourselves internally. We've got a lot of users, but for now they use what is already available. We create everything like language learning materials and all the features connected with Web3: crypto exchanges, NFT marketplace, and whatever you can do with NFT and crypto on the marketplace and then the app.

Michal: You mentioned that you're doing both educational design, like in the product, and also localization. So how does that add up? It seems like two separate roles.

Anna: In the beginning, I was more a localization manager because we didn't have... established processes at all. When I came to the company three years ago we had two languages and it worked with makeshift processes established by someone else.

At that time, they didn't have a separate role for that. And then I came, we decided to scale a little bit more, so we needed more localization efforts. And now we only need to update and keep everything on track.

So now it's easier to manage.

But with education design, we've got a lot of new features basically every month. It's not so hard to work in both of these roles because we've got established processes already.

Michal: Yeah. Okay. So do you feel that being involved in localization gave you a different perspective on designing the experiences?

Anna: Yes, sure. Because when you work in localization, you just instantly become aware of what you deliver in different languages and in different cultures and how people react to it. You have to always keep that in mind when creating educational features or whatever product features you design, because all, the cultures and all the languages are different.

For example, if we talk about education design, you have to adapt your methods to the audience, because people have different native languages and they understand the grammar rules and exercises differently. For example, if we talk about RTL languages, like Hebrew or Arabic, the syntax is very different.

So you have to teach things differently for those users. The same in localization. You have to adapt everything to the culture, otherwise it just won't work for them. I think those two domains are very close because you have to adapt everything for culture.

Michal: Do you find that people in different cultures learn differently?

Anna: They actually do, but I think if we talk about online learning, the differences are not so distinct. If you teach online, you basically have to think of the syntax and the cognitive load for different types of learners, based on their level and based on their native language as well.

But it's also very hard to adapt. If we talk about localization. We usually adapt the learning content when localizing it. We've got special style guides for, like, grammar explanations. We always ask our linguists, what do you think is the best way to explain this concept to speakers of your language? How do teachers teach this concept in your language? And it's always different.

For example, if we talk about Chinese, they find special metaphors to explain certain concepts like tenses. And in Russian, we usually just use a lot of translation. So we try to adapt. So the copy is not the same in any language.

Michal: Do you find that, components of the experience, like for example maybe... if you give positive reinforcement, do you give it differently to people from different cultures?

Anna: Well, I don't think we do that now. But I think in the future we have to adapt things. All the marketing copy should be adapted, but sometimes it's like, too costly for certain markets, so you have to first analyze and do the research and then think of how much effort you want to put in that.

market.

Michal: Yeah. Prioritize.

Cause you talked about motivation and I was thinking, okay, I know that people here are motivated by different things compared to those from other cultures. And then... I'm guessing the more you adapt this to the culture, the more motivation you're actually able to achieve for each of these markets.

Anna: Yes, that's right, for example, if we talk about learning English, the motivation is very different in different countries. A lot of people want to learn English for moving to an English speaking country. That's why it's just important to emphasize different things in your marketing copy.

Michal: Yeah. I would imagine. Okay. So I want to ask you about your process in localization specifically.

Cause you said that you came at the beginning and there wasn't really a process, but then gradually you evolved it into something that can now more or less maintain itself. So you need a smaller team.

So tell me a little bit about that.

Anna: Okay, so when I came to the company, we didn't have any established processes. The only thing they used at the time was Lokalise for UI localization. We didn't have any vendors or freelance teams at the time. That's what I started with.

I think that team is one of the most important things in localization, because everything else depends on how qualified your vendors are, how qualified your managers are. The quality of the end product will depend on that. That's why I started from the sourcing and testing and everything. Now we have a wonderful team of mostly freelancers, but we've been working with some of them for more than a year. They know the product thoroughly, and they know everything about it.

So that's what I did.

Next came the tools. As many other companies, we also used spreadsheets in the beginning. Everyone has that step in their processes when they use spreadsheets.

Michal: Yeah. That's all we had, though. Wasn't it? It was, it was born out of necessity, I guess, because before we had of these tools, it was the only cloud based environment that we could use

Anna: So we used Google spreadsheets as well. And then we realized that they are not reliable for localization. So we started using more translation management platforms. And after that, we focused on streamlining the process, we wanted to deliver all the updates in 24 hours. And we kept on adding new languages, so the time zone difference was a problem as well. So we optimized, now we can deliver everything in 24 hours, basically. So that's how all the processes changed.

Michal: Let's talk about each of these. You mentioned people, vendors. You mentioned tools, and then you mentioned timelines.

So first of all, people, because you said you found your team and you started working with permanent linguists. Which I agree is a really important part of it.

Where did you find them? How did you get them to work with you permanently?

Anna: So at first, we started working with people from Upwork. Because it was the easiest, it's super easy to hire people. But what we found out in the process that the quality could be very low. We couldn't guarantee we deliver the best quality.

So we invested more time in hiring more freelancers. Who we could work continuously for a long time. And after that I also used Upwork and Proz, and also, I just Googled and used LinkedIn a lot to find the right vendors. So we tested them thoroughly. We just integrated a test processes with sample tasks that they had to complete before we had an interview. That worked like magic for us, we found a lot of very professional linguists who could do the work we have. And just... Continued working with them after that.

Now we've got 30 linguists in different languages. We've got two people for each language and we usually double check everything. So the first person is a translator. Then the second person reviews everything. Our quality just went up after we integrated that.

Michal: So you manage all the vendors yourself? So don't work with LSPs, you manage the freelancers on your own.

Anna: I used to manage all the freelancers on my own for a long time. But now we've got a localization project manager as well because now the workload is growing and now I mostly do strategic planning.

Michal: Yeah. And they manage the vendors. So that's great. And then you mentioned tools too. So you said you're using Lokalise for your UI translation.

Anna: Yeah, we do use Lokalise for UI translation. And we use everything it has to offer inside, like glossaries and also... machine translation. Sometimes when we need to do things quickly, we use it, and then our translators review it, but not so much in UI.

Michal: So why not for UI? What was your experience with that?

Anna: Because UI is very content dependent.

I think machine translation is just not the right tool for the UI localization. That's my personal opinion. I think the technology will develop and probably in the future, you'll be able to use it without any changes, but now it's more of a hassle. When you write or when you localize the UI copy, you have to keep the whole scenario in mind. Machine translation can't do that for you. That's why we don't use it for UI. Sometimes we do, but only for very simple things. Very simple titles when we have to deliver things very fast.

We do machine translation, but we never release anything with machine translation without review. We review it with the translator and only then we can release it.

I think machine translation works great for longer texts. For example, if you have a long article that is not too specialized, you can use machine translation, and then you need to review it anyway. And sometimes we just use it for educational content. When... It's not context dependent. Two years ago, I worked at a language learning game, and we didn't have a budget to translate the huge dictionary manually. So we used machine translation for the dictionary translation, and we asked our translators to review it. It saved us a lot of money because they didn't spend so much time translating every single word of that dictionary.

Michal: How do your translators respond when you send them machine translation for MTPE?

Anna: The first thing we do is we do the machine translation, and then we ask them to review it just very quickly. Just to assess it from one to five. And if it's four out of five, they just do the post editing. So... But when the quality is very poor... I think Hebrew, Arabic, Chinese, and all the Asian languages. Machine translations just can't cope with that. So we don't use it for all the languages as well.

For French, for German, I think for Spanish, it would be very effective use it for them. so it depends.

Michal: Yeah, absolutely. There's also, I think, sometimes other issues when you're using machine translation. Like in Hebrew, a lot of the times we have a gender issue because machine translation is trained on a lot of masculine form content.

Anna: Yeah, that's right.

Michal: And that's an issue on its own. Because, those sentences, even if they're correct. Changing them to non-gendered, it's like re-translating everything.

I think it's more of a question of prioritizing. Once you prioritize a language, and you teach the engine to handle that language correctly, then you will get significantly better results. But you do need to say, okay, I want to focus on that language now, and actually invest time to understand what it needs, and understand how to adapt that.

And I'm guessing there are languages that get the spotlight for now, because they have a lot more material and a lot more speakers. But it will change.

Anna: Yeah, that's true. Absolutely.

Michal: So let's get back to the process. When you localize. Who sets up the environment, who sets up the ICU syntax, who chooses the screens and imports them and exports them?

Anna: Usually the dev team. It's important that the dev team creates the strings because all the titles are integrated in the app.

We usually have the English copy as our source text. We first review it and check if it's correct, if it corresponds to the context and if everything is how it should be. And then we assign the translators from our team to do the localization of those strings.

Michal: What kind of context do you provide?

Anna: Oh, we try to provide a lot of context because that's... I think it's essential. So we screenshot everything from the app, or we also copy the screens from Figma and add them right in Lokalise as well.

We also write comments for basically every string, if there is something that is hard to understand from the screenshot sometimes. Sometimes you have to divide the strings into small parts and they are joined together in the end. So you have to leave a comment about that. So that's what we do.

We also have guidelines for our translators that they have to read before they start working with us. Like how they should deal with strings translation, how to use variables, how to translate the titles.

Michal: So about the guidelines. You mentioned that translators need to read them before they start. Do you mean before they start every project or is it before they start working with you?

Anna: No, basically, before they start working with us. Not every time. Sometimes we have to remind them of certain things because they just... Yeah.

but it's not every time. What we add every time is screenshots and small comments to strings that can be hard to understand.

Michal: Are these guidelines general for the company or are they language specific or locale specific?

Anna: They are specific to certain types of languages. Like European languages have similar syntax, so we don't create separate guidelines for every language. And we have specific guidelines for RTL languages, for Asian languages.

Some languages, we have certain instructions that we have to follow. It's usually with polite and impolite. So we have special that we want to emphasize.

Michal: Do you also give guidelines on branding and things like tone of voice, or is it just technical guidelines on gender and formality and syntax and stuff like that?

Anna: Yeah, we do. We've got a tone of voice file in Notion, where we state all the important points, how we want to address the user, and a small explanation: what our product is about, how people interact with the product and what they expect to see there, and also the level of formality we want to use and the forms we want to use for buttons and for different kinds of titles.

Michal: Do you sometimes get feedback from linguists on how that tone of voice fits their culture or the language?

Anna: Yes, we have a lot of feedback. In localization and in product development, nothing is set in stone. You can't create a file that will stay with you forever. You have to change it all the time. You have to edit things. You have to think of better ways to convey your message.

That's why we usually ask our freelancers for feedback. We also do UX research and we talk with the users. So we got some insights from them. And after that we adapt our tone of voice to what they expect from us. We also look at the competitors and see what they use in different languages, and also use these findings to continuously improve our documentation and how we localize.

Michal: Do you do UX research in all languages that you localize into?

Anna: I don't think so. We only use this for... Our big largest markets mostly. But ideally that's what I want to achieve. I think every user deserves to read quality copy. I think that we try to achieve the best quality, but of course we have to work and do more research in every language.

Michal: If you had gotten the budget now to do research in all of those markets. You don't have a lot of budget, but you do want to make an effort. So what kind of effort would you choose to get insights?

Anna: One of the most important things you can do is to use your translators as a source of expertise in that language. Or maybe use your team members as well, because I think in international companies, there are at least a few people who can speak different languages.

Ask them about the copy, go to whatever source you have... I'm also in a polyglot community, I'll just ask people to talk to me. I'll show them the screens, and they just give me a lot of feedback.

I remember once, when we wanted to launch in Japan, I just opened the Polyglots Facebook group and created a post. I had a lot of people who commented. We had a call and we showed him the product and we had a lot of insights after that. And it was basically for free. So you can always find people to talk to.

Michal: Yeah, definitely. And I like that you... instead of saying, oh, we don't have the budget, you said, okay, let's try something creative. Which I think is really great.

Anna: Yeah, it's very important to be creative. Sometimes you don't have the process you don't have the people. Sometimes you are just a one person team. So you have to think of creative ways to solve problems that come up, to get better results. That's what we do a lot in localization. We also try different platforms to hire freelancers. We tried everything.

Michal: Yeah, we should call it " localization hackers".

Anna: Yeah.

Michal: Try and find ways to do localization better if you don't have the budget or the time or the people.

Do you have an example of a time when you made an effort as a company, you changed something in the UI or you made an effort in terms of how you guided the translators, and you saw real measurable improvement in performance?

Anna: Let me think. if we talk about localization all your efforts pay off because you can cover more markets. I think when we localized to Chinese traditional and to Hebrew, we had the biggest return. If you find the right market you'll see the return. With the revenue, with downloads and with everything. I can't think of a specific example because localization is usually an iterative process. You come to the market with the first version of, localized copy, and then you don't see any return. But you do very small changes. And in a few months you see the revenue growing and you see the user base growing.

Michal: I think it's probably one of the hardest things about localization. Is you can't really say, Oh, we did this effort from A to B. And then we saw improvement in performance subscriptions or something else, because it is, like you said, a very, very long term process. And there are a lot of little tiny improvements throughout, which makes it really hard to measure and really hard to prove the value of it.

When you take a step back and you look at the longer process you can see there's definitely a significant improvement.

Anna: Yeah, absolutely.

Michal: So, okay, you mentioned that you managed to get all your requests answered within 24 hours. How did you make it happen?

Anna: We used a lot of project planning. We did kind of audit to see what the bottlenecks are. In the beginning we had a very long turnaround time, like three or four days for basically each language, because we had a distributed team of freelancers working from different countries. So there is a huge time difference. We didn't know how to manage that. So we just sat down and looked at the processes and we found out that we assign the translators too late for them to start right away.

Michal: Yeah.

Anna: So we found those bottlenecks and we just started working on them. At first, we started assigning the translations as early as possible. And it worked great right away. And then we talked with the project manager and we changed the release processes a bit. The strings were added in the beginning, so we had time to review them in English, add all the context and assign them to the translators at the right time. And we also integrated some planning tools. We use Notion for the project planning, so we created a board where our translators could see all the tasks and the timelines and the deadlines.

And we also just made sure they are available every day. We talked with them and discussed the workflows, and continued working with those who were okay with that, and wanted to work with us. Because it's important to be transparent in your communication.

I remember we had translators who were not so used to UI translation, but they did the other translations perfectly, like articles and marketing copy. So they ended up procrastinating on those stuff. So we talked with them. It was a lot of project management. And...

Michal: Yeah.

Anna: And we also just invested a little bit of our time in automation.

We started using Smartcat and Lokalise for different types of content. And we also used plugins for Figma to automate the workflow. So we just came to 24 hour turnaround time, you know, in 16 languages.

Michal: Yeah. That's really impressive. You said there were some people who were not so inclined to do UI translation. Do you find that it also impacted the quality of the UI content that they created?

Anna: Some of the translators are more proficient in certain areas. I think most of the translators do all kinds of translation work, but every person likes a certain type of tasks more than the other ones.

Michal: Yeah.

Anna: And now we keep on working with people who understand our product, who see the mission of our product, who are reliable in terms of deadlines and timelines and who just have a passion for their language.

It's a pleasure to work with people who are like that. They ask a lot of questions. They always want to know everything about the feature and you just have very interesting chats with them about certain features and intricacies of their language.

Michal: There are linguists who will do a lot of the work that you ask them to do, but the thing that they really do amazingly is marketing, or something else, maybe technical, maybe business, maybe medical.

So I'm curious why you chose to work with one or two translators per language for everything, rather than split the linguists by their specialties and the things that they're really passionate about.

Anna: We didn't come to that in the beginning, we had a lot of staff turnover in the process. In the beginning we had a small team of freelancers and they usually come and go and then you just find someone else. And after that we just found the right team for this kind of translation work. So they usually specialize in UI as well and in marketing copy. That's what we mostly have.

And we also have translators who work on educational content, so now the teams are kind of specialized because we've got a team of freelancers who work on UI, and another team that works on educational content.

Michal: I would imagine it even takes time to find those people because usually good translators are busy.

Anna: Yeah, it takes a lot of time. Another approach could be: you find people who really want to with you. And you teach them everything. But it takes a lot of time, a lot of effort from the manager.

Michal: Yeah, although I would imagine that if you invest in teaching your linguists, first of all, you get exactly what you need in terms of the training that they get. But also I'm guessing they are automatically more invested in working with you .

Anna: Yeah.

Michal: Because you gave them something as well.

Anna: Yeah, that's what we do a lot. We think of them as our team members, but just those who work part time. It's important that they feel that they are the part of the team. They are developing with us. So I think it's very important.

Michal: Yeah, I completely agree. So I have a question. Can you outline just the process looks, from creating the content in the source language and the design stage to having a completely localized feature out in production?

Anna: So in the beginning, we just brainstorm, we look at the user story, and then we interact with the designers with the product manager a lot to come up with the design that serves the right purpose. The English language copy comes up at the time.

When we approve the final copy, we usually try to gather all the context, the screenshots and all the comments and everything that we need to assign it to the translators.

After we review everything, we create the UI strings and add the terms to the glossary. We've got a large glossary of crypto terms. We need to sometimes think in advance how to translate them because there are no direct equivalents in many languages. So that's what we do. We create new terms. We talk with the linguists about how we can translate those terms. We do the research. We look at the competitors at this stage as well.

Sometimes when the features are very complicated, we have to talk with the developer team about the variables, about the plural strings, about the strings that are divided into different parts sometimes. So we have to look at that and talk with them about how we have to deal with them in different languages. Like Hebrew and Arabic and Chinese and all the others that don't use the Latin script.

So we try to gather and direct all this information to our linguists, so that they understand the context. And then we localize the copy, the QA team checks everything, visual bugs and functional bugs. Sometimes we spot the localization bugs at this stage or sometimes before the functional QA, we do the linguistic QA, but it usually depends on how big the feature is.

And after the LQA, we do the edits if they are needed. We correct everything and review everything and then it's launched. After the release, we also check it once again to see if it's okay. It's always double or triple checking. Localization is like that. You have to be ready for that.

Michal: Yeah, absolutely. So if you could do one improvement, whatever you want. What would you have optimized?

Anna: Well, I'm a big fan of automation in any process. As a strategist, I always think about how we can automate this. So I would integrate more automations in the process.

The last one we did is... We've got 16 languages and we got app store pages.

Michal: hmm.

Anna: And We've got like 10 screenshots for each of the languages. And for a long time the process was, we translated all the copy from the screenshots, and then the designers manually added them. We had to first copy and paste those strings, then the localizers had to change them and check them and... It was such a daunting task. No one wanted to update the screenshots anymore.

And it just left us behind. If you want to do marketing right, you have to update your assets. You have to do the experiments. So I just found the Smartcat plugin that helps you import everything from Figma to Smartcat.

Then it adds the translations from your Smartcat project to Figma. You don't have to copy and paste anything manually. I'm always looking for the ways to automate things. So that's what I would continue doing.

Michal: Yeah. Automation and optimization, right?

Anna: Yeah.

Michal: Just making things faster, I would guess, because even if it's not really automated, just having the busy work done for you can probably make everything much more convenient.

Anna: Yeah, and it makes your people happier sometimes, those who do these daunting tasks, like designers and localization manager. They are more satisfied with what they do because they don't have to do those routine and manual work that's very daunting

Michal: Yeah. Absolutely. Okay.

Okay. Thank you for taking the time to share your process at LetMeSpeak.

Anna: Thank you so much for having me here today. I hope it will be helpful for your audience. I would be happy to follow up anytime on the processes and on how the localization industry develops and talk about anything localization related. So that was great talking to you today. Thank you, Michal.

Michal: I also want to thank you, the person listening to this now, for taking the time to learn more about localization. Localization has always had this gray kind of lackluster image, but it's got such a huge impact on the world around us.

When we take the effort to make localization processes better and improve localized user experiences, we make a positive impact for users globally, and you are a big part of that. If you're as passionate about localized user experiences as I am, make sure that you join our community. There's a link on the Localization Station website, and it's where all the most interesting conversations happen.

And if you want to hear about new episodes of the Localization Process Pod... Be sure to subscribe on Spotify, on YouTube, or the Localization Station email list for updates. I was Michal Kessel Shitrit and this was the Localization Process Pod. Hope to see you again for another localization process review soon, and until then, have a great day!