Ankit runs engineering at Flo Labs, the company behind Shopflo and Sortment. Shopflo is the checkout layer for a large set of e-commerce brands in India, it processes a high volume of transactions a day and cannot afford even a couple of minutes of downtime. Sortment is their newer product, an AI marketing platform for lifecycle communications, still early but growing fast. Both ship every week. Both run on small teams.
Challenge
Unit tests, API tests, all of that was solved. The thing Ankit could never get to stick was UI level end-to-end testing.
“End-to-end tests have been something we have been trying to automate, but never could really make them work beyond two to three release cycles. They would get outdated and the team would have to again spend a lot of effort.”
Anyone who has tried to build an in-house Playwright suite knows the shape of this problem. You write the tests, they pass for a release or two, then the UI shifts, selectors break, and suddenly you are spending engineering hours fixing tests instead of shipping features. After a couple of cycles of that, most teams quietly let the suite rot. For a lean team shipping weekly across two products, that is not a sustainable trade.
Solution
Ankit found Empirical through another founder at a VC event. We have been working with Flo Labs across both products since.
The setup is simple. Empirical’s forward deployed engineers own the test suite. Flo Labs ships code, tests run, and when something needs to be added or fixed, the team tags Shopflo engineers on Slack. Maintenance is not the problem anymore.
On Sortment, the team operates with a single QA engineer. That is only possible because Empirical handles the recurring regression coverage, freeing the QA to focus on new feature validation.
On Shopflo, the platform has gotten broad. Different brands want different things, so over the years the surface area has grown to a point where manual coverage is not realistic. Empirical now runs a wide test suite on every commit, which gives the team a clean sanity check before anything goes out.
Results
Here’s what changed:
The migration test
The clearest signal of whether end-to-end tests are actually working is what happens during a big migration.
Shopflo’s backend was originally Python Django. Over time, it has been migrated to Java services. Normally, that kind of migration means rewriting your test cases alongside the code, which doubles the work and stretches the timeline.
“We just switched over APIs, asked Empirical team to take care of certain nuances and we were able to see that before and after the migration, things are working smoothly. It gave our team quite a lot of confidence boost before releasing such large migrations.”
The tests did not need to be rebuilt. The team validated the migration by running the same suite before and after, and shipped it.
Why the human layer matters
QA man hours on redundant testing dropped by 50%. That is the number Ankit tracks, and it is what keeps the team lean while shipping weekly across both products.
Ankit is clear that automation alone is not the full answer. With a high volume of commits going out every week, things break. AI can handle some of it, but not all of it.
“There is a lot of work required to ensure that the test cases are working week on week. AI can do a part of it, but there is still a human layer involved. That is where the Empirical team really helps us.”
This is the part most testing tools miss. The cost of end-to-end testing is not writing the tests, it is keeping them green as the product changes underneath them. Empirical’s forward deployed model is built around that maintenance layer.
Advice for teams considering Empirical
We asked Ankit what he would say to engineering leaders weighing this. His answer was direct.
“I feel there is no better way of doing end-to-end tests for your software than Empirical. If you want to keep your team lean, this is the way to go.”