Discovery & Research · master area
Assumption Surfacing
Naming what we believe but haven't witnessed. Three states — witnessed, inferred, not witnessed. Each cycle moves assumptions from one state to another. The cycle that doesn't surface assumptions runs blind into the gaps it doesn't yet know it has.
Owners: PO Phase it lives in: Before We Build → After We Build The corpus principle this enacts: Witnessed, not described.
Where it lives in the chain
- Before We Build · Assumption surfacing — the canon
- After We Build · The Model Update — where assumptions are moved between states
The three states
| State | Meaning | Carries which risk |
|---|---|---|
| Witnessed | We've sat next to the person doing this. | Lowest — but witnessing once is not forever. Re-witness if the context changes. |
| Inferred | We're reasoning from related observation, or industry knowledge, or what the brief from a similar moment told us. | Medium — explicitly named so we know what we're carrying. |
| Not witnessed | We believe this is true. We have not seen it. | Highest — most production surprises live here. |
How to do this
The Initiative Brief and Feature Brief both carry a "What we have NOT yet witnessed" section:
Not witnessed:
- The teacher mid-session, switching languages.
- Graders who have less than 1 year experience.
- Sessions on tablets.The cycle's discovery either witnesses some of these, or the team ships knowing these gaps exist — and the chain-aware label observation-mismatch is on the watchlist for the signal reading.
After the signal reading, the model update closes addressed items and adds new ones:
Before: "We have not observed someone with no Hebrew reading ability."
After: "Closed — May 2026. New: user switching languages mid-session, seen in usage data."What good practice looks like
A team that surfaces assumptions catches observation-mismatch before shipping — "We assumed graders work in linear sequence; we haven't watched that; let's witness before we design." One session of witnessing might change the entire flow design. That's discovery doing its job.
A team that doesn't surface assumptions ships features confidently, then learns from production what the brief should have anticipated. The cost is the wallet bug pattern: one line of code, six weeks of trust to rebuild.
Related crafts
- Observation / Field Research — the upstream practice
- The Model Update — where assumptions are moved post-cycle
- Initiative Brief — where assumptions are first named