In the series: Lean Localization Questions
One of the fun applications of ChatGPT is that you can see how people usually answer questions. Well, at least up to two years ago. So when I asked ChatGPT how to prevent translation errors, it came up with quite a few excellent answers (such as: “Encourage feedback and collaboration”) but it also told me the following:
“Perform thorough QA checks: Perform thorough QA checks on all translated content, including proofreading, checking for accuracy and consistency, and reviewing for cultural appropriateness.”
This is what has been preached for decades, mainly by translation agencies: use the 4-eye principle, review every single piece of content and your translated content will be great.
However, doing QA checks on every piece of content can lead to an enormous increase of costs, and even worse, time to market. And still there is no guarantee that the human being who is checking another human being will correct every mistake – or won’t add any.
So this could mean that your customer is waiting for that new feature in your product that is going to reduce your churn rate dramatically, because someone is checking the translated content, going back and forth between people, discussing who’s to blame, etc. etc.
Or maybe even worse, a translation that was perfectly fine to begin with that is waiting for days for a review – and then minor changes are made that add no value to the customer but sprout from just another human being’s opinion.
All this can lead to major delays in the time to market of your product or content. So in that sense, review steps can get in the way from providing value to the customer.
No wonder that one of the Lean Localization Questions is: “How do I increase the time to market by eliminating the review step, and still deliver the quality I need?”
First, let’s look at what Lean is all about. It’s an approach that aims to create flow and value for the customer by eliminating waste in the process. To make our lives easier, 8 types of waste have been defined by Toyota so we know where to look in order to make sure every step actually adds value to the customer. See the infographic above.
As you can see, there are two types of waste that are of interest here: Defects and Overprocessing.
Defects refers to the rework that needs to be done because of the output not being up to spec, overprocessing refers to wanting to deliver higher quality than the specs tell you.
In that sense, unnecessary reviews are considered waste, because it adds no value, leads to waiting time and people making changes that are not necessary. On the other hand, if at the end of the process the output does not meet the requirements, having to do the rework is a killer for the velocity of your process.
The end result must provide value to the client and we want the user experience to be equally good in all languages. But in terms of Lean, the review step can never be the way to ensure quality, because we still have the waste of fixing defects and the overprocessing of a review that was never necessary.
So how do you eliminate the “overprocessing” of review and not end up with more waste because of unexpected rework. In other words: how do we get to a “first time right”?
The core of the answer is: use the review step to improve your process, and not your output. So instead of believing that review is the way to ensure quality, see it as a temporary step that serves the process improvement and can be made redundant by structurally removing the root causes of the occurring defects.
Here are the 6 steps to eliminate the review step from your process and still deliver a high quality user experience to your users
- Mind shift: think of quality and defects in terms of your customer experience and what is important for THEM. Not what is important to a translator, a reviewer or an agency, or even the localization manager.
- Create one piece flow, where you run smaller pieces through your process
- For EVERY defect your reviewer finds, that falls into the defect category, run a root cause analysis and improve the process so that the defect cannot happen again.
- Design a channel for user feedback that feeds right into an “update” process. This can also be done by applying business metrics, like conversion or support calls, to the several language versions.
- After you have improved your process in such a way that at least 80% of the cycles are without defects, and zero critical defects occur that need instant rework, remove the review step and move to the “First publish then edit” approach.
- Design an update process that has a very low cost of change, in terms of time and money, so that defects that occur are fixed at a lower cost than reviewing every translation. Run a root cause analysis and improve your process for every defect that occurs.
Scary? Well, maybe. However, if you think of the opportunity costs that arise from having a review step in every cycle, and the costs of having a mistake in the translation in 20% of the cases, that is then fixed within a few hours at low costs and no effort, you might reconsider.
There are three prerequisites for making this approach successful:
- Collaboration: Direct contact and a good relationship with translators so you know what they need to deliver work without defects
- Automation: without automation your cost of change will be too high
- Integration: all subprocesses must be included in this design, such as content creation, publication and the actual translation, or you will not have access to solving the root causes of your defects.
In short: just bluntly skipping the review step might just be overdoing it a bit. Just like in software delivery, you don’t want to serve your customers with crappy content. However, with well thought out processes, the review step can indeed be considered “extra-processing” and can be removed.