Problem and goal
Users can export any report from any core Semrush tool. That’s the first and only time the majority of users interact with reporting. It looked like this:
We needed to scale and add new scenarios — which was impossible because of the extremely cluttered interface and seven-year-old legacy code.
Process and result
In the early stages of discussion with the PO, we decided to shift from a very misleading one-button scenario to a multi-format export hub — making it clear and scalable. I made a quick rough draft — enough to keep the conversation going right away and move it one level down, from the concept to what it actually does:
The code was structured around a one-button scenario. Together with the team, we took a deep dive into backend and frontend mechanics and how we could realistically change them, while staying within the constraints of the A/B test. It took us six one-hour grooming sessions. One thing we needed to measure turned out not to be measurable — our A/B platform couldn’t count specific items inside a GA4 ecommerce event. I took it to the team who build it, and they added support before we launched.
Most of the decisions from the draft survived into the final design. The main one: every export scenario became a thing of its own, instead of one more toggle in a wall of toggles. And AI Summary, our highest-converting paid feature, got its own place in the top settings block instead of sitting buried in that same wall.
I ran five hallway tests on a clickable prototype of the new structure to check that people read the split into separate actions the way we intended. All five did — including three who had never opened reporting before.
In the A/B test, export conversion rose from 58% to 70%, and the use of AI Summary grew 3.4 times.