Concept

Product Market Fit Engine

conceptpmfproduct-market-fitsuperhumannubankmeasurementsurveymulti-source

Product Market Fit Engine

A systematic method for measuring, segmenting, and improving product-market fit, developed and popularised by Rahul Vohra at Superhuman. The framework makes PMF a tractable, iterative process rather than a binary judgment.

The Sean Ellis question

The engine is anchored on a single survey question, first articulated by Sean Ellis:

‘How would you feel if you could no longer use this product?’

Respondents choose from: very disappointed, somewhat disappointed, or not disappointed.

The target signal: 40% or more of respondents answer ‘very disappointed.’ Companies above this threshold have generally found strong PMF; companies below it have not.

The three steps

1. Measure. Survey all active users using the Sean Ellis question. Calculate the percentage answering ‘very disappointed.’ If below 40%, do not proceed to growth — improve first.

2. Segment. Among respondents who answer ‘somewhat disappointed,’ separate those for whom the main product benefit resonates from those for whom it does not. This is done by asking a follow-up question about the main benefit they get from the product.

The critical insight: ignore the ‘not disappointed’ group entirely. Their feedback points away from PMF — it represents users who are fundamentally not the target. Acting on it is the most common PMF mistake.

3. Improve. Generate a roadmap from the two high-signal groups:

  • Ask ‘very disappointed’ users what they love most. Use those answers to identify what to double down on.
  • Ask ‘somewhat disappointed / main benefit resonates’ users what would make the product better. Use those answers to identify what is blocking conversion.

Allocate roadmap roughly half and half: deepen what fans love; overcome the specific objections of the target segment.

Why this roadmap is guaranteed to increase PMF score

The algorithm is self-reinforcing by construction. Doubling down on what fans love strengthens retention and word of mouth among the already-converted. Addressing the objections of the ‘almost convinced’ segment directly increases the conversion rate of the next wave. Both moves raise the ‘very disappointed’ percentage.

What this prevents

The common alternative — surveying all users and building toward the median feedback — tends to produce a product that is inoffensive to many but beloved by none. The PMF Engine explicitly filters out feedback from users who are not in the target segment, preventing the roadmap from drifting toward an unmeetable average.

At Nubank — the score as a hard scaling gate

Jag Duggal, CPO of Nubank, runs the Sean Ellis question as a near-hard gate: the company rarely scales a product until it clears the threshold, and product managers are culturally expected to say no to scaling until it does. Two adaptations stand out:

  • A 50% threshold, not 40%. Nubank raises the bar because Brazilians answer surveys more optimistically and politely than the global average — a deliberate cultural calibration of Sean Ellis’s original number.
  • The bullseye-cohort method. When the overall score is borderline, Nubank does not chase the average; it finds the small sub-segment already scoring ≥70%, reverse-engineers what that cohort shares, and expands it. The Payments Assistant went from ~40% overall to 10M+ monthly actives this way.

Duggal pairs the gate with two further instruments: NPS post-launch (Nubank skews 70s–90s; a one-to-two-point dip triggers alarm) and a churn watch to ensure the product is not a leaky bucket — because the word-of-mouth growth loop only compounds if acquired customers are retained. The operating principle: never scale a small problem, or you turn it into a big mess. As a lead designer’s phrase, now a review standard, puts it — ‘good enough isn’t good enough; is it great enough?‘

Source