The survey question that predicts churn

We ran NPS surveys for eighteen months. We had dashboards, cohort breakdowns, and a steady stream of 6s and 7s with no context. Then we added a single Sean Ellis question and everything changed.

The question: “How would you feel if you could no longer use FeatBudy?” with options: Very disappointed / Somewhat disappointed / Not disappointed.

Why this question is different

NPS asks users to predict their future behavior. The Sean Ellis question asks them to imagine a loss. Loss aversion is a more reliable predictor than satisfaction. Users who say they would be “very disappointed” are your retention core. Users who say “not disappointed” are telling you they have one foot out the door.

• Benchmark: >40% “very disappointed” correlates with strong PMF
• We sat at 34% before optimizing our onboarding
• After onboarding changes, we hit 51% — churn dropped 18% the following quarter

When to ask it

Timing matters as much as the question. We trigger it on day 14 — after users have had enough time to experience core value but before their trial expires. Asking too early gets aspirational answers. Asking after churn gets nothing at all.

The moment a user loses your product is when they tell you what it was actually worth to them.

What to do with “not disappointed” responses

Follow up immediately. These users are in the churn window. Ask what value they expected that they did not find. The answers are almost always actionable — a missing integration, a confusing workflow, a pricing mismatch.