How to create the perfect localization pipeline: a case study

by | Jan 24, 2023

From a broken translation process to publishing all languages in a single day

This is a success story about how we improved the translation process of a complex web site. Within 3 months, my client went from a turnaround time for the translation of a web page of over a month to 1 day. This means that one day after the web page was published in English, it was also published in the other languages.

Actually, the situation was even worse. When we started, the turnaround time was infinite because they had stopped translating all together. This delay in their time to market kept them from actualizing their offering, leading to significant delay in market penetration for countries that were to drive growth.

The translation process came to a full stop because:

  • it took a full time developer to publish content in all languages
  • formatting problems that originated in sending content from and to the translator were too time consuming to solve
  • the translations did not match the tone of voice of the source text and were often not even accurate – therefore no one believed it had value to have them
  • all updates were done manually and were done on an ad hoc basis by people from the organization – distracting them from their actual work

In their new situation:

  • The company publishes over 4x more content in all languages (including English) than before our improvement project.
  • Within a day of the last part of the webpage being done in English, the translations in all languages are published
  • They need no more project manager to deal with translations, all people in the process are self-supportive
  • The developer is not involved at all in the operational process
  • Translations are accurate and attract the right audience, just like the English version

Sounds like something you might like to have as well? Read on to see how we did this.

One method…

The method we created is called Lean Localization. It is an application of the well-known Lean methodology to localization processes. Lean came about in the Toyota Production System, and was happily taken over by many companies for all kinds of processes, Tesla being one of them.

The key take-away? Go for flow and improve what MUST be improved, not what CAN be improved. Define your problem and find the root cause of your problems. This means: don’t just add extra steps before you understand what the underlying issue is. A simple example: don’t add extra QA if you don’t know why a translation is not to your liking.

So how do you find out what must be improved? An extremely simplified version of what we do is described below.

…two major changes

In Lean Localization, you start with two very practical essentials that make all the difference in approaching localization processes.

  1. Unbatch the content – work with smaller pieces
  2. Collaborate with translators directly: optimize the whole instead of suboptimization

Unbatch the content: work with smaller pieces

One of the tenets of Lean is single piece flow, or one piece flow. The key to this is to define the smallest piece that has value to the customer. In Lean Localization, we are looking for the smallest publishable piece. These pieces are going through the entire process one by one, from content creation to publication of all languages.

However, in the case of most translation projects, including my client’s, the exact opposite is happening. Content is usually batched before it is sent off to a translation vendor. This means that, from a Lean perspective, one of the biggest wastes in the translation industry is… waiting time. We wait for a whole website to be perfectly done before we send the whole site to an agency. We then wait for everything to be translated before the content is reviewed. And then we wait for everything to be reviewed before we try to publish it, all in one go.

Coming back to my client, at this moment of trying to publish everything in one go, my client found that, for basically every project:

  • Formatting was broken so the content couldn’t get imported – or with very strange results
  • The design did not leave room for the length increase of most translations
  • The translation was not to the liking of anyone in the company

This led to enormous rework on all three aspects. With the obvious consequences of delayed deadlines, overtime and lots of workarounds that were not sustainable, and that made the same thing happen again when updates to the page had to be translated.

So we made a major change and we took one page of the website to be our “piece”, aka the MVP – the minimal viable publication. We ran these pieces through the process and for each cycle, we monitored and analyzed the problems that we ran into. We made small process or technical improvements on the root causes of the problems every day.

Now this may be rather obvious, but working in smaller pieces has three big advantages: you have smaller problems, you learn from them faster and therefore you end up with fewer problems.

This means it enables you to actually make a deadline without having to solve last-minute issues! And you will experience fewer and fewer issues as time goes by.

Work with translators directly

In order to create flow and obtain continuous improvement, you need to have transparency of every step in the process and access to all people working in it. You need to be able to investigate why mistakes are made in translations and communicate the feedback to them directly.

This is why we decided to collaborate with freelance translators directly instead of going through an agency. For the first part of the project, they engaged directly with people from the sales department. The direct contact between people interacting with customers and the translators led to more trust that localization could actually benefit their purpose, and gave the translators more guidance of what was expected from the translations.

Eventually we were able to remove the review step completely and this improved the speed of publishing by at least a week – as these people actually have other jobs than reviewing translations.

Additionally, by including the translators in the process, they became an incredible source of information for what had to be improved in order for them to create the best possible translations. Acting on their feedback, for instance on the source text and technical imperfections, was the fastest way to address the most critical productivity and quality issues, which significantly reduced the cost of localization.

What we learned

Although a lot of improvements were already implemented in this first phase – and translated pages were actually published – the main goal was to discover the REAL problem and solve it in order to create flow. So this is what we learned – and more importantly – could prove and show with actual data.

  • For this website, a page was still too big and too complex a piece. The philosophy of Contentful was not followed when creating pages, and that was the major source of complexity.
  • This also led to problems that originated in the way the content was formatted, in order to obtain the right lay-out
  • The rest of the problems originated in the way the pages were designed, which allowed too much freedom – leading to too much variety and unstructured execution.


The solutions we came up with were no rocket science, and they had been thought of before. Back then, however, these were all ideas that seemed way too complex to implement.

Removing all waste in the process before deciding on the technical solutions made these solutions truly obvious and simple to implement. Therefore buy-in from higher management was not an issue anymore. The value was clear, both for authoring and translation, and the improvements were prioritized.

Here are the main major improvements that were implemented on the basis of our findings:

  • The client was working with Contentful, which is an extremely powerful tool – if you follow their philosophy. They work with chunks of content that can be reused, “entries”. We moved to entries as the smallest piece, both for content creation and translation. The content creation department now delivers entries to the translation process, which are translated and published into Contentful within a day. When all entries are done, the page is assembled for all languages at the same time.
  • We banned all HTML from the content in Contentful and solved necessary formatting with rich text
  • We created a library of reusable design components, which were tested for localized content before they were formally approved. Now, zero rework is needed after translation
  • We moved away from the integration that was supplied by Localize – the translation tool, and built a small tool that fit our approach. It is a plug-in in Contentful which allows you to send the content directly from the entry.
  • We implemented some pretty awesome and sophisticated processes for updating both the English and the foreign languages.

The current state

In the current situation, the client is experiencing zero issues when it comes to translation. No review is necessary. Smaller pieces are created and sent to the translators directly after they’re done. This means that when the entire page is put together, all parts have already been translated, and all languages can be published at the same time – without waiting time.

They do not need me – or anyone – anymore; the system works on its own.


Working in one-piece-flow – the key to Lean Localization – has enabled my client to set up a process that was crystal clear and that led to enormous savings in terms of hours – but mostly opportunity costs.

Removing all waste and exceptions from the process made it really simple to put together a robust integration between the several systems. All issues have been removed, no more project management is needed. Because the process is so smooth, they manage to keep the best translators committed to them, leading to the quality translations they need to build the trust with their clients they need.

Our role in all of this? We set up and visualized the current process, designed the ideal process, found and onboarded the translators, helped them out with temporary project management, analyzed the root cause of the issues that arose, came up with the solutions to the problems and helped design the new integration. And probably a few more things…

Most importantly, we reported the outcomes and proposed necessary changes in such a way that essential decisions could be made on the basis of data instead of good faith and our expert role.