What wasn't working.
Mobile top-ups, a core digital goods offering, were hard to discover inside the product. The user journey had grown organically over years and buried the purchase flow behind navigation nobody had revisited.
The underlying technology stack carried enough debt that shipping and testing changes was slow. There was no reliable way to know whether a change actually improved anything, so most decisions were made on instinct.
What we actually did.
I redesigned the user journey to surface digital goods earlier and remove the navigational friction sitting between users and the purchase flow.
In parallel, I scoped a modernization of the technology stack specifically to unblock experimentation, then introduced A/B testing as standard practice rather than an occasional exercise.
Once the journey and testing infrastructure were in place, we added Gift Cards as a new revenue stream on the same rails, instead of starting a separate build from scratch.
Where the interesting calls were.
I fixed discoverability before adding a new product. A new revenue stream on top of a broken funnel would have underperformed and hidden the real problem.
I framed the stack rework as an enabler tied to a business outcome, not a side project for engineering. That made it easier to prioritize against feature work.
I launched Gift Cards only after the testing infrastructure existed, so the new line could ship and iterate quickly instead of waiting on its own foundation.
The numbers after we shipped.
Digital goods revenue increased by 30% after the redesign removed the friction that had been burying the purchase flow.
Gift Cards launched as a net new revenue stream on infrastructure that already existed, cutting the time to ship.
A/B testing became an established practice the team kept using well beyond this project.
Things I'd carry into the next case.
Technical debt work is easier to justify, and to prioritize correctly, when it is tied to a specific business capability it unlocks.
Fixing the funnel before adding features to it compounds. Everything built downstream benefits from the same fix.
Testing infrastructure pays for itself once, then keeps paying on every feature that follows.