Product Discovery

You might say

Don't have AI start coding features yet. Let's do product discovery first to see whether users actually care about this problem.

The exploration process of collecting user and business evidence to validate whether a problem is real, valuable, and feasible before committing to full development.Suppose a team wants to build a prompt-result comparison tool. Discovery turns 'creators need comparisons' into a testable assumption, then gathers evidence from recent real stories and a low-fidelity task. The evidence may support continuing, narrowing, changing direction, or stopping. Discovery reduces early uncertainty; it does not guarantee product success.

Know first

User Story User Flow PRD

Product Discovery Phase

How product discovery validates hypotheses before writing code

Example research recordPrompt-result comparison tool

There is an idea, but no user evidence yet

Value: Creators frequently need and will use result comparison
Usability: Beginners can understand two-column differences and choose
Feasibility: The same input can produce comparable results reliably

“I think users will like it” does not belong in the evidence column.

Discovery produces an evidence-backed next step, not automatic approval.

How a product idea becomes a testable decision

Before: the team treats a guess as user demand: The team assumes creators need prompt-result comparison and immediately asks AI to build accounts, dashboards, and settings. Shipping those features proves delivery, not that target users face the problem or would replace their current workaround.

Write risky assumptions, then gather behavioral evidence: Separate value, usability, feasibility, and business constraints. For example: creators manually compare results and this repeatedly causes rework. Reconstruct the last real event and workaround, then observe one selection task with a paper or low-fidelity prototype. Each assumption needs observable behavior; compliments alone are insufficient.

The outcome is a changed decision: If evidence supports only quick visual comparison, narrow the concept to a two-column view instead of expanding a full workspace. If the problem does not appear, pause and reassess. Continue, narrow, redirect, or stop are all valid outcomes when the evidence explains the choice.

Quick checkChoose the best answer

Research shows users manually compare results side by side, but they currently care only about spotting differences; nobody uses collaboration or multiple exports. What should happen next?

Narrow the scope to two-column differences and saved results, then keep observing task completion and repeat useBecause comparison matters, finish collaboration, multiple exports, payments, and an admin dashboard togetherBecause users did not endorse the full concept, conclude that no comparison problem exists and discard the research

You can say this to an AI Agent

I want to build a prompt-result comparison tool for independent creators. Do not build the full product or high-fidelity UI yet. List the riskiest value, usability, feasibility, and business assumptions; ask what evidence already exists; then design interviews around recent real events and one low-fidelity task. Define observable signals and whether each outcome means continue, narrow, or pause.

Learn next

1User Journey→2PRD→3MVP→

Recommended tool

Web fine-tuning tool · Selector Pro

Select page elements for precise AI edits

Click a button, text, or card, note the change, then copy its exact location to AI.

Open Selector Pro website