The screens above make the solution look direct. The hard part was the model underneath. When I joined Pocket FM in October 2024, listeners bought coin packs and spent them to unlock episodes. That PPV economy generated most of the revenue.

The business needed a subscription model, but Premium could not behave like a cheap replacement for coins. The constraint was firm: PPV revenue had to hold. A generous plan would collapse the coin economy. A restrictive plan would stall conversion.

Users in Tier 2/3 India expected a subscription to unlock the catalogue. Our plan did something narrower: 60 mins of daily free episodes per show plus ad-free listening. After that, paid and free users still met the same coin wall. I had to make extra access feel worth paying for without making coins feel obsolete.

Users kept asking for subscription, so the expectation was clear: more access, less friction, and a value story they could understand in the middle of listening.

There was no prior subscription product. I was starting from zero.

After V1 shipped, our user researcher Rachit Jain ran a qualitative study on R2+ cancellations. He interviewed former subscribers who had used the product and still churned. I was a named DRI on the research plan, sat in on calls, and observed user behaviour.

Rachit helped us see V2 differently. Users saw an unexpected bank debit and cancelled out of anger. They had lost trust in consent, not interest in the product. The phrase that came up repeatedly: "Bina puchhe paisa kaat liya." Money taken without consent.

I reframed the V2 brief around trust repair at renewal, debit, and cancellation moments. I turned that research into three V2 decisions.

Before locking the model, we ran experiments across every dimension of what a subscription could be: full app access, 120-min daily unlocks, 90-min, catalogue models with a fixed set of free shows, extra coin bundles. Every option broke something. A generous plan would crater PPV revenue. A restrictive plan would stall conversion.

I held the design through that ambiguity. I pressure-tested each construct against two questions: does the user understand what they are getting, and does the business survive if they do? I kept the 60-min daily unlock because users could understand it and PPV could survive it. From that point, I made the model legible, trustworthy, and worth paying for.

PPV revenue held throughout. We shipped subscription without touching the coin economy. That was the constraint from day one, and the reason the early design decisions mattered.

Repeat behaviour mattered most: a third of Indian subscribers have re-subscribed at least once, and repeat subscribers drive nearly half of India subscription revenue.

Parallel to subscription work, I took ownership of accessibility across both platforms. The app had no baseline. I audited from first principles: WCAG requirements, touch target sizes, ALT text coverage, VoiceOver and TalkBack behaviour. Documented the gaps, prioritised by severity, and shipped fixes in phases across iOS first, then Android.

I scaled the audit with a Figma plugin called A11Y Assistant. It automates VoiceOver label generation using AI, analyses layer structure, naming conventions, groupings, and interactive regions, then suggests context-aware accessibility labels designers can apply with one click. Shipped internally in March 2026, published on the Figma community.