KKBOX needed subscriber growth in Taiwan, and the team had already spent whole projects going after it from the familiar directions. Reduce churn among existing users. Win back the ones who had left. Both moved slowly, and neither moved much.
Then a telecom partnership put KKBOX inside a carrier's subscription package, and new registrations started arriving in volume from a single, clearly defined channel. We knew exactly when they arrived, through which plan, and what they did next. For a team that had been trying to read small movements in a large existing user base, a clean incoming cohort was the most measurable thing anyone had been handed in years.
These users also had a property worth building for. They had not compared streaming apps or read a review. KKBOX came with the phone plan. That meant their first session carried unusual weight: nothing before it had given them a reason to stay, so everything depended on what happened inside it.
Two opportunities sat on top of each other. A real chance to move the subscription number, and a well-defined population to test a new way of working against.
So we built a small cross-functional squad and gave it one job: run a full loop every two weeks. Propose, evaluate, build, validate, iterate. We called it KKBOX Cycle.
The loop worked because the lanes overlapped. While one idea was collecting data, the next was already in development, and the analysis from the first fed an optimization rather than blocking the queue.
So the first work was choosing.
