Light as a feather: How Smartcat helped streamline localization of the Flo app

Where healthcare applications are concerned, the accurate translation of medical terms is paramount in ensuring that users feel as comfortable as possible when using the service. This is especially true when the application serves a delicate and intimate topic, such as tracking a user’s menstrual cycles.

Initially created as a period tracking application in 2015, Flo has evolved into an AI-powered women’s health product and now the world’s leading one-stop health product for women during their entire reproductive life cycle: from first periods to menopause, from pregnancy to young moms. Flo was the most downloaded app in the App Store’s Health & Fitness category in August 2019 and boasts 100 million users from all over the world.

We sat down with Alexander Markevitch, localization team lead at Flo Health Inc., to discuss their approach to localization and how Smartcat fits in it.

“Ensuring proper quality of translations has always been one of the major challenges we have faced when developing Flo,” says Alexander. “Before we started to use the Smartcat Freelancer Marketplace, we lacked a constant pool of freelance translators/vendors for us to draw upon. We handled all our localization needs via Upwork, which meant no certainty that the same person who worked with us one day would do so the next.”

“We handled all our localization needs via Upwork, which meant no certainty that the same person who worked with us one day would do so the next.”

What kind of tasks do you use Smartcat for the most? Which additional tools do you use?

We use Smartcat for translating articles and localizing the Flo user interface. We use Serge to push and pull resources to and from Smartcat. For LQA, we unload XLIFFs and then check them in QA Distiller.

How did Smartcat help streamline your translation processes?

Before using Smartcat tools, our translation process consisted of 38 steps — Smartcat helped us reduce this to six! It simplified our process for onboarding new freelancers, and our overall efficiency got a huge boost due to the platform’s Vendor Management features, the Marketplace, and the open API.

“Before using Smartcat tools, our translation process consisted of 38 steps — Smartcat helped us reduce this to six!”

What gains in productivity did you see?

In fact, our productivity went through the roof in just the first month of using Smartcat. We were able to pump out 300 new articles in the first month, while previously we could only manage 40–50 monthly.

“Our productivity went through the roof, growing by 500% in just one month of using Smartcat.”

What about quality?

In addition to the aforementioned efficiency gains across the board, we were glad to see a large reduction in the number of complaints concerning non-gender-neutral language and an upswing in general consistency. Now we are also able to quickly fix any interface errors.

How do you measure success? Which metrics do you track?

We use the control pointcuts system to measure success. In the first two weeks of our cooperation with new vendors or freelancers, we check a random sample of 30% of the total volume of texts created by the new contractor, using a third-party reviewer. If a vendor successfully completes the first pointcut, we make a second measurement a month later, using a sample of 15% of their work. If the second pointcut is also successful, we take a final measurement after three months using a 5% sample. This allows us to monitor the quality of work produced as fairly, efficiently and accurately as possible.

Vendor quality control cheatsheet from Flo Health Inc:

  • Check 30% of a new vendor’s translation
  • In one month, check another 15%
  • In three more months, check another 5%

Another metric we use to judge the success and efficacy of our localization efforts is by calculating the ratio between cost and number of words translated. It is also useful to monitor user reviews — sometimes users vocalise complaints or make their own recommendations regarding improvements to translations. Naturally, the less of these we receive, the better the job is being done!

What do you think is changing in the translation industry, positive and negative?

The main negative we have picked up on is that the industry seems to be moving more and more towards post-editing of machine translation (PEMT). While this can save on time and costs, it can also lead to poor-quality translations — something we cannot afford in our industry.

On the positive side, as advancements in translation and localization technology are finally beginning to be embraced and embedded in technical processes, we are able to make our work as seamless as possible.

“As advancements in localization technology are finally beginning to be embraced, we are able to make our work as seamless as possible.”

Thank you for this talk, Alexander! Good luck further developing Flo!

Thank you, too.


Vova Zakharov
Vova Zakharov Smartcat’s former editor-in-chief, Vova loves g̶l̶o̶b̶e̶t̶r̶o̶t̶t̶i̶n̶g̶ staying at home with his family, playing some good old metalcore, and talking to self-aware robots.