Product thinking and discovery
Cheap discovery tests
Fake doors, concierge, Wizard of Oz and prototypes — spend days, not months, to see if an assumption holds.
Match the test to the assumption. The spirit is the same across names: spend hours or days, not months, before you burn delivery capacity.
A week of concierge that kills an idea is a win. A quarter of ML work on an untested desirability bet is the failure mode.
Concept
A cheap test is the smallest next step that could invalidate a riskiest assumption. Write the kill criterion and the date before you start.
Walking through the toolbox
| Test | What you do | Good for | Weak against |
|---|---|---|---|
| Prototype / mock | Clickable Figma or paper | Usability, clarity of value prop | Real willingness to pay |
| Fake door | Button or ad to something that does not exist yet; measure clicks; explain | Demand / desirability | Overpromises if you do not message carefully |
| Concierge | Deliver the outcome manually for a few customers | Value and willingness to pay | Scalability / feasibility |
| Wizard of Oz | UI looks automated; humans behind the curtain | Feels-real UX + value before automation | Long-term cost; disclose when appropriate |
| Landing page / waitlist | Explain the job and offer signup | Messaging and rough demand | Fake email collecting |
| Technical spike | Thin vertical slice | Feasibility | Desirability |
Fake door: add "Suggested rate" behind a button on invoice create for 10% of users. On click, show "We're testing this — want early access?" and log the click. If click-through is 1%, desirability is weak. If 25% and people complete a short form about their niche, continue. Do not leave fake doors that look like broken product without an honest explanation.
Concierge: for ten customers, you personally send payment-link emails and gather what they charged. Learn the job and the words they use. Only then automate.
Kill criterion example: fewer than 10% of activated users click within two weeks → drop this solution and try another under the same opportunity.
Cheap tests burn trust if mishandled. Fake doors that look like broken buttons train users to ignore your UI. Wizard-of-Oz that pretends to be full AI forever is not fine; "early access, humans in the loop" at signup is. Concierge that over-delivers manually creates expectations automation cannot meet — frame it as a pilot.
When a solution survives, scope an MVP that still targets the riskiest remaining assumption — not a v1 wish list.
Common mistakes
- Leaving fake doors up with no explanation. That is a broken button.
- Concierge that spoils the eventual product. Set the frame: manual for now.
- Spikes only. Feasibility without desirability is how you build unused pipelines.
Check yourself
You need to know if freelancers want AI pricing. Prototype or fake door?
Fake door (or an interview probe). A pretty mock will not tell you demand.
A Wizard-of-Oz research agent has a human writing briefs. What do you tell signups?
That it is early access with humans in the loop. Do not pretend it is full AI.
Key takeaways
- Match test to assumption: fake door for demand, concierge for value, spike for feasibility.
- Write kill criteria and a date. Honest framing protects trust.
- Survivors become thin MVPs, not wish lists.
Further learning
Next, run a week end-to-end in a week of discovery.
Related
Continuous discovery and assumption testing
Weekly customer contact, opportunity solution trees and cheap tests before building.
Riskiest assumptions
State what must be true for an idea to work, then test high-hurt and high-uncertainty beliefs first.
MVPs and scoping
Building the smallest thing that tests the riskiest assumption, and iterating.