Strategy question

Cursor uses a credit-based pricing model. How would you redesign pricing to reduce user confusion?

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Pricing design that solves a real, named user complaint, confusion, rather than a generic pricing overhaul.

Clarify the specific confusion: users cannot predict how many credits a given task, like a large refactor versus a small edit, will consume before running it.
Propose upfront cost estimation: show an estimated credit cost before executing an agentic task, similar to a price preview, so users can decide before committing.
Propose simplifying the underlying unit: consider whether credits should map more directly to a familiar unit, like requests or time, rather than an opaque computed score.
Propose a usage dashboard showing real-time remaining credits and a clear breakdown of what consumed them recently, reducing the feeling of an unpredictable black box.
Define success: reduction in support tickets about credit confusion, and reduction in negative sentiment specifically mentioning pricing in reviews or feedback.

Targets the actual named problem, unpredictability before committing to an action, with a concrete pre-task cost estimate rather than a full pricing model rewrite.
Proposes making the underlying unit more intuitive, tying credits closer to a familiar concept like task type or time, rather than an opaque scoring formula.
Adds a transparent usage dashboard as a trust-building mechanism, addressing the feeling of losing credits without understanding why.
Sets a measurable target, reduction in pricing-related support tickets and negative reviews, directly tied to the stated problem of confusion.

Proposing to eliminate credits entirely for a flat subscription, which may not fit the variable cost of different AI tasks and ignores the actual ask.
Ignoring the need for upfront cost visibility before a task runs, which is the core source of confusion.

How would you estimate credit cost accurately before running a task whose complexity is not fully known in advance?
What would you do if simplifying the pricing model reduced revenue from heavy users?