This is a fictional example of a Foundry Decision Review. The founder, the firms and the quotes are invented. The evaluation is real: the example input was run through the same deterministic policy a paid review uses, and what you see is what the policy returned.

The founder here has four encouraging conversations and is ready to start building. Read what the evidence actually justifies.

Fictional exampleInvented data, evaluated by the real policy. Not a customer's review.

The assumption reviewed

Owners of 5–30 person accounting firms will change their weekly workflow to use an automated deadline-tracking assistant.

Outcome: Iterate

Not yet. There is signal here, but not enough to build on, and not enough to stop.

Either a comparable-or-stronger alternative exists and is worth pursuing instead, or real uncertainty remains that more evidence could resolve.

Not: The idea failed.
Not: Nothing was learned.
Not: This is a stall or a non-answer.

Why

Real uncertainty remains that isn't yet resolved either way. More targeted evidence could resolve it.

Evidence quality audit

This sounds really useful. I'd love to see it when it's ready.
- Opinion · Weak
- Independent
- Current
- via outbound
- 22 Jul 2026
We'd probably sign up if it existed.
- Stated intent · Weak
- Related
- Current
- via referral
- 24 Jul 2026
Every January I lose a week rebuilding the deadline spreadsheet from last year's.
- Confirmed problem · Medium
- Independent
- Current
- via outbound
- 29 Jul 2026
Showed me a colour-coded shared spreadsheet that two staff update by hand every Monday.
- Existing workaround · Medium
- Independent
- Current
- via community
- 1 Aug 2026

Independent sources come from 2 distinct channels.
A disconfirmation attempt was recorded.
No repair attempt was recorded.

How far the evidence has climbed

PAIDWhat they paidnothing here yet
Money changed hands, or a contract commits them to pay.
COMMITTEDWhat they committednothing here yet
They put real time, people or budget behind it.
DIDWhat they didevidence here
They acted: built a workaround, shared real data or access, changed how they work.
SAIDWhat they saidevidence here
Opinions, stated intent, and a problem described in their own words.

The ladder explains the evidence; it does not decide. The outcome above comes from the full policy.

Contradictions

1 contradicting event recorded (2 Aug 2026). That is not yet repeated contradiction: the policy needs two events at least 7 days apart.

Falsifier status

If fewer than two of ten firms agree to connect a real client calendar within 30 days, this assumption is wrong.

Defined 15 Jul 2026; evidence collection began 20 Jul 2026. It was written down before the evidence was collected, so it can be used.

Not shown yet: what building would need

This risk level needs at least one item of observed customer behaviour (shared data or access, a workflow change, a resource commitment, or payment) -- not opinions or stated intent alone.

Not shown yet: what stopping would need

Listed so you can see how far the evidence is from either end. These do not count against you.

A real repair attempt is required before stopping -- one that actually changed pricing, ICP, message, or channel. None was recorded.
Stopping requires at least two contradicting events at least 7 days apart. The recorded events don't yet meet that bar.

Evidence limitations

A human review of this evidence is recommended before acting, though it is not required to reach this outcome.

Decision risk

Medium risk. The higher the risk of being wrong, the more the evidence has to show.

Cycle 1

This is the first evidence-gathering cycle for this assumption.

Next evidence step

Get one observed-behaviour item (shared data or access, a workflow change, a resource commitment, or payment) -- opinions alone won't clear this gate.

An example review, start to finish