Most product failures are not engineering failures. They are value failures: we built something competently that nobody needed. In high-uncertainty markets, the cost of being wrong is dominated by how long it takes to find out.

Discovery is the discipline of buying information cheaply. It is not a phase that ends when delivery begins. It is a set of loops that run alongside delivery.

Start with the riskiest assumption

Every initiative rests on a stack of assumptions about desirability, feasibility, viability, and usability. Teams often test the easiest assumption — usually usability — because it produces attractive artifacts. Instead, rank assumptions by uncertainty × consequence and test the top one first.

  • Desirability: Do people care enough to change behavior?
  • Viability: Does the business model survive if this works?
  • Feasibility: Can we build and operate it at the required quality?
  • Usability: Can people succeed without training?

Design the cheapest decisive experiment

A good experiment is not a prototype demonstration. It is a question with a decision attached. Write the decision before you write the test: "If fewer than 30% of interviewed admins can complete the workflow unaided, we stop and revisit the onboarding model."

Concierge tests, landing-page probes, narrow internal betas, and fake-door tests can all be valid. The right method depends on the assumption, not on the team's aesthetic preferences. A five-day research sprint is often cheaper than a five-week build — and far cheaper than a five-quarter commitment.

An experiment without a kill criterion is just a demo with better branding.

Keep engineering in the room

Discovery that excludes engineers produces elegant concepts with impossible operating costs. Include a senior engineer in the problem framing, not just the solution review. They will spot data-model constraints, integration risk, and support burden that product research will miss.

Timebox the uncertainty

Set a discovery budget in days, not weeks. If the team cannot reduce the critical uncertainty within the budget, the answer is not "keep researching." It is to narrow the bet, change the customer segment, or stop. Sunk-cost pressure is strongest when the team has been exploring for a long time without a decision.

Write the learning down

Discovery produces decisions, not just insights. Keep a one-page record per bet: the assumption, the test, the result, the decision, and the next review date. This artifact is worth more than a folder of interview notes because it preserves the reasoning that future teammates will otherwise have to reconstruct.

The goal is not certainty. It is a faster, cheaper sequence of honest encounters with reality.