01 — Rank Assumptions
Assumptions are fine; invisible assumptions are not.
Early-stage teams operate in fog. That is normal. The trap is treating artifacts as proof: a proto-persona becomes the user, a design sprint becomes validation, a problem interview becomes confirmation because the questions were built to hear the pitch back.
Assumptions should be explicit and ranked—which ones, if false, kill the wedge. Research at validation stage is not a phase you finish. It is the compass that tells you whether build should start.
Proto-personas are valuable when labeled as hypotheses with owners and expiry dates. They fail when they become wallpaper in every deck—users you never met, quoted with confidence.
02 — Proxy ≠ Market
Proxy users test the team, not the market.
Internal stakeholders, friendly experts, and synthetic stand-ins can align a room. They cannot substitute for the person who will live with the workflow, absorb the risk, or pay. Proxy sessions are useful for coherence; they are dangerous when cited as demand.
If your discovery plan ends at people who already agree with you, you are not de-risking—you are rehearsing.
Field research in consequential domains means watching work under real constraints: interruptions, policy, liability, handoffs. The assumption trap closes when the team can write: we would stop if we saw X.
03 — Hunt Disconfirming
Problem interviews fail when they confirm the slide deck.
Good interviews hunt disconfirming evidence. They ask what people do today, what they have tried, what they will not do, what would make them distrust you. Bad interviews collect adjectives about your concept.
The test is simple: did the last three sessions change the wedge, the kill criteria, or the build order? If not, you bought therapy, not research.
The trap shows up in healthcare incubation work the same way: storyboards align the room long before field evidence rewrites the wedge. Confusing alignment with proof is the failure mode in every domain.
