Bullseye Cohort

A method for iterating a borderline product toward product-market fit by finding the small sub-segment that already loves it, working out what that segment shares, and expanding it — rather than optimising for the average user. Developed in practice by Jag Duggal and the product team at Nubank.

The problem with the average

When a new product’s overall Sean Ellis score is borderline — say 40% answering ‘very disappointed’ against Nubank’s 50% gate — the instinct is to chase the average user with broad improvements. That tends to produce a product inoffensive to many and beloved by none. The bullseye method inverts this: the signal is not the average, it is the peak.

The method

1. Find the bullseye. Within the borderline base, locate the cohort whose score is already very high — at Nubank, ≥70% ‘very disappointed’.
2. Reverse-engineer what they share. Identify the concrete attributes that distinguish the cohort: how many products they have registered, which behaviours they exhibit, which features they touch.
3. Expand the bullseye. Build the few things that move more of the base into the high-scoring pattern, then re-measure.

Worked example — the Payments Assistant

Paying bills in Brazil runs across four payment rails of varying reliability; a missed bill can put a black mark on your credit. Nubank’s Payments Assistant launched with a borderline Sean Ellis score (~40%). But a small cohort with four or more commitments registered across at least two of the four rails scored 70%. The shared attribute revealed the work: make it easy to onboard many bills across many rails in one go, so automation takes over and the benefit becomes a fundamentally different experience rather than an incrementally better one. Over roughly fifteen months of dozens of small iterations, the product went from struggling in the hundreds of thousands to over ten million monthly actives.

The same move worked on Ultravioleta, Nubank’s high-income rewards card: the cohort that spent enough to clear the monthly fee waiver loved the surrounding benefits, which pointed at who the product was really for.

Why it works

It pairs naturally with the discipline of refusing to scale before PMF. A high-scoring cohort is proof that the product can be loved by someone; the job is then narrow and concrete — make more of the base look like that cohort — rather than vague and additive. It also gives the product manager a defensible reason to hold the line against pressure to scale prematurely.

Source

See Jag Duggal on Building Fanatical Customers, the Sean Ellis Score, and Nubank's Product Strategy. Related: Product Market Fit Engine, Category Design, Reference Customer.