As content becomes the number one factor driving the global economy, time-to-market is shrinking more than ever. This means that, unless your company works exclusively with one market, you need to be able to launch your content in all the languages of all your markets as quickly as possible.
How can companies do this? With continuous localization. Essentially, it’s a process where any new content you produce — whether it’s a string for an app, text for a blog post, or a tweet — gets automatically sent to a vendor or vendors responsible for its translation.
In this article, we will take a look at how you or your language service provider (LSP) can enable continuous localization for your content with various alternative approaches.
Disclaimer: For the purpose of this article, “language service provider” can mean either an external company or a department within your own organization. Almost all considerations below apply in both cases. We believe this article will also be interesting for “external” LSPs who want to educate their would-be customers on the topic of continuous localization.
Process and Challenges
The main steps in a continuous localization flow are as follows:
Set up: First and foremost, the initial activity to set up a continuous localization process should be reasonably simple.
Push to LSP: There are plenty of repositories, CMSs, and other platforms where content can be stored. Once a string is in it, it should be automatically sent to the LSP.
Mobilize: Once the content is with the translation provider, it should be sent for translation. Ideally, this should happen without the provider’s asking questions or your making additional requests.
Do it: The translation must be good, and that’s typically a matter of trust because the customer doesn’t usually speak the languages that the content is translated to.
Sync back: Once translated, the content should be placed correctly in the repo/CMS structure, ideally with no need for manual input by the original creator.
Security: Not really a step in the process, but rather a challenge that must be met throughout: Your confidential information must stay secure to prevent possible breaches and leakages.
With these and other challenges in mind, let’s compare some of the options to build a continuous localization flow.
The Options
We’ll consider four popular options:
Manual localization. This is not continuous, but worth including in this comparison to show you what companies sticking to the “good old ways” are losing out on.
Various “platform-like” translation agencies such as Gengo, Steppes, Unbabel, etc. We’ll call them “Unbabel-like” for brevity.
A generic, cloud or desktop, translation tool such as Memsource, Matecat, or Trados + a translation management system (TMS) such as XTRF or Plunet.
Smartcat, either standalone or with a TMS.
Manual localization
This is the “old-school” way of doing things, with stuff sent back and forth over email or instant messaging.
Set up | Nothing to set up here. You just do things as you’ve always done them. |
Push to LSP | You have to download the content from the repository yourself and send it to your LSP. |
Mobilize | There’s no way the LSP can get translators working on your strings right away. This is especially true for small jobs that both the LSP and translators spend a disproportionate amount of time on. |
Do it | The LSP will perhaps select every translator manually to find the best match for your job, so that’s a plus. On the other hand, if they don’t use at least a CAT tool with a memory of your past translations & terminology, quality will be compromised. For larger pieces of content, not using technology tools will also hurt productivity. |
Sync back | The LSP will have to manually compile the translations, send them back to you, and you’ll have to somehow upload them to your repo. If that isn’t bad enough, multiply this by the number of languages you have — you’ll have to repeat it that number of times. |
Security | Although emails are considered a more or less secure means of communication, the sheer number of peer-to-peer communications involved in the process increases security risks. |
Other issues
The human factor means that you will sometimes find strings in the wrong languages or even forgotten.
The problem is worsened if you have versioned repos and have to attribute specific content to specific versions.
The overall time it takes makes overnight delivery impossible: Most likely, it will be days or weeks.
The LSP is likely to ask for a higher rate because handling small jobs means more overhead.
Takeaway
The conclusion is obvious: it is not possible to enable continuous localization with a manual approach.
Unbabel-like translation agencies
Such agencies present themselves as “online platforms”, where you can quickly upload your content, possibly via an API, and have it translated without any human interaction between you and the platform.
Set up | Provided the agency has an open API, the setup is fairly simple. However, you still have to find a way to automate the use of the API. You can do this with tools like Zapier, but it still takes some software/localization engineering knowledge, and won’t suit more complex configurations. |
Push to LSP | Once configured, the process becomes more or less automatic, again, provided you have the tools to push content as required. |
Mobilize | Translators are usually chosen automatically from a large pool using the “early bird catches the worm” principle. |
Do it | Different platforms have different approaches, but, in general, the overarching problem is that quality falls victim to a “random” choice of translators. |
Sync back | This is easy too: Once a translation is done, it gets pushed back to your repo via API — if you have the tools to handle such API requests. |
Security | Although some of these platforms make translators sign NDAs — others don’t, — the anonymity and randomness of the ultimate contributors put your content at a high risk of leakage, which can be critical for, say, major app updates. |