- Localization Workflow Highlights
- Developer engagement with localization stops at providing source content
- Quality of localization matters a lot, yet it competes more with speed of delivery
- Increasing trust in the accuracy of machine translated content.
- Embedding Localization deeper on CI/CD flow
- More automations around localization workflow
- Closing notes
This is the second blog in a two-part series, in which we discuss how 11 companies are leading the way with modern localization workflows. In our previous blog, Modern Localization Workflows: A Deep Dive into 11 Real-World Cases, we broke down the notes, feedback, and patterns we observed from our year of researching and observing how customers are localizing. In this second part, we highlight some more interesting workflows from customers.
Read on for some specific approaches that customers shared during the interviews and highlight the ones that we found to be most interesting.
Localization Workflow Highlights
Developer engagement with localization stops at providing source content
This is a derivative observation, based on the preferred way of integrating TMS with the main application (that is API) as well as on the way the final localized content is embedded back to the application. Customer interviews and observations have revealed that automation of the localization process is a MUST when the localization steps reach the engineering department.
Our observations also revealed that companies tend to create internal tools that automate content and context extraction. This means that specific requirements related to localization quality and translation flow are set up from the beginning and certain code automations are developed. An alternative to these automations would be for the developers to manually extract or create the additional context needed on an as-needed basis, which is not the best usage of developers’ time.
Developers want to spend their time adding value to the core application, not providing one-off localization notes
It is safe to conclude that developers want to spend their time adding value to the core application they maintain, not getting sidetracked by localization support questions to accommodate other teams. All the additional tooling and automations are there to ensure developers focus on adding value to the core application.
Quality of localization matters a lot, yet it competes more with speed of delivery
63% of the customers interviewed value the quality of translations. This finding is also reinforced by our interactions with customers who pool or crowdsource translations. These customers are optimizing for speed with their pooling approach, and subsequently prioritize applying fixes for sentences or phrases that may have distorted meanings once localized. It’s clear that the need for growth and faster expansion to new markets competes neck-in-neck with the need for localization quality.
Increasing trust in the accuracy of machine translated content.
From customer behavior within Transifex, we have seen a growing trust in the quality of machine translated content. This is evidenced by the increasing number of workflows where MT-produced content is marked as “fine” and approved.
Two of the interviewed customers are considering MT as a valid solution: one is actively using MT in the localization workflow, and the other is considering MT as an valid option once it becomes sophisticated enough to take into account intricacies such as context and tone of voice.
Additionally, we can see the need for fast delivery from how localized content is getting published. In the cases listed in this section, content is published and improved progressively. This means that the value gained even from “halfway there” or “almost correct” localized content is more than if the localized versions were not available at all. Again, we see speed of delivery win over the perceived quality of localization. That’s why we believe that speed of delivery and improved MT quality will make automated workflows mainstream in the near future.
Here is how a fully automated workflow could work, walking an actual customer’s workflow leveraging existing Transifex features:
- Set up an automated localization workflow with company’s MT of choice
- Leverage in-platform features to protect variables and placeholders*Ensure translation quality by using glossary terms translations
- Report localization issues to fix problematic translation in weekly or monthly batches
* Note: Transifex always protects variables and placeholders when content is sent for MT. Also, when using Amazon MT you can feed your glossary in Amazon as entities to consider while MT translates your content.
Embedding Localization deeper on CI/CD flow
An interesting observation from the interviews was the intention to move forward with localization flow starting earlier in the feature development process. In most cases, localization starts right after a feature development is complete. Two of the interviewed companies also mentioned that localization starts in parallel with feature development, where developers are responsible for sharing new content for localization while working on code. This means that a feature’s public release will have at least two phases: one for the source language and a subsequent one when the localized versions of the feature are complete. A possible alternative is to delay the feature’s public release until all global content is complete, essentially delaying the release.
During our discussions, we saw an emerging trend of starting the localization process even earlier than development, during the design phase. Specifically, 54% (6 out of 11) of the companies interviewed shared the intention to move to a “localization from design” workflow.
This localization workflow entails having the product and design teams work together closely during the design phase of a feature, in order to finalize all screens and copywriting. The final designs are then handed off to both developers and localization teams to start working in parallel. This way, the public release of a feature can be achieved faster and be more complete, including all global content.
On this note, we recently launched a Figma Plugin to help our customers move towards this emerging localization trend.
More automations around localization workflow
If there is one thing that stands out from all the interviews, it is the need for automations around localization. These include automations around how content is delivered, how content enters the TMS, and how localized content leaves the TMS to feed back into the applications using it.
Optimizing the way content enters and leaves a TMS will free up a lot of resources and make localization seamless and easy to integrate in any development workflow. In turn, this will enable further growth of companies that are not currently localizing their content. Because of this, we have invested in a new solution called Transifex Native, making the connection between development and localization fast and easy and fast.
With a solution for content input and output (content I/O) from a TMS under way, companies will be able to integrate faster with localization tools and build their global content repository to manage all their content, regardless of what language it is in.
With this problem solved, the next areas of improvement come in the localization stages that follow. Specifically, localization teams are asking:
What happens after content is added in TMS and localized content before leaving the TMS? What are the tools needed to manage fast and accurate localization? Will these steps be fast enough to keep up with the pace of development and design of new features?
The answer to these questions lies in the suite of tools that help automate or speed-up global content management and creation. Such tools include:
- Fill-ups, that use your existing translations through Translation Memory or trained MT engines, or generic computer assisted translations through MT to automatically translate new untranslated content.
- Glossary and context, to help translators deliver better translations
- Quality assurance tools, to help the localization managers quickly evaluate delivered work and identify issues.
This list can expand more to include more items around people and team management, automated workflow tasks and many more.
Interviewing our customers is a great way to understand how different companies are using localization and, more specifically, how they are using Transifex. It is great to collaborate with teams who share the same drive for innovation, constant improvement, and new opportunities!
From these customer interviews and daily interactions with our customers, it is clear that localization is an active priority in their development cycles. Both the individuals we spoke with and the companies they represent are interested in moving forward and moving fast; and are adjusting their workflows towards success and growth. This emphasizes the active need for innovation and agility from the localization industry.
This year, we have taken great strides towards contributing to this industry innovation and agility. Earlier in 2020, we took our largest step towards localization innovation with Transifex Native — a tool we built to change the global content game. With Transifex Native, you can now manage all your global content in one central place and save time on deployment. To learn more about modern localization workflows and how they are being disrupted by Transifex Native, visit www.transifex.com/native.
With our latest Transifex Native solution, you can now manage all your global content in one central place and save time on deployment. To learn how you Transifex Native can help you make localization a seamless part of the development lifecycle, visit www.transifex.com/native.