Day 28. The four week-one artifacts have stopped being aspirational and started being load-bearing. Here is what has to be true by then — and what it means if any one of these is missing.
The bar is concrete, not vibes
A month in is the first point at which a SolutionWright engagement has to show its work in a way an outside reviewer could check in an afternoon. Not a pretty deck. Six items, each one either present or absent, each one tied to the /trust-receipts-standard.
If a vendor at week four can't list these on demand, the engagement is running on momentum, not evidence.
1. One measurable change, in production
The first observable change has shipped and is being used by your team (Class C — configured, written down, and visible to both sides). It is small on purpose. The point is not the size; the point is that the loop from scope memo to deployed-and-measured change has been closed once, end to end, while the trust is still cheap to repair.
If week four arrives with nothing live, the scope was too big, the hand-offs are wrong, or the engagement is stuck in slide-production mode. We say so and rescope.
2. Six receipts in motion
The scope memo, evidence-class baseline, falsifier list, and exit-plan draft from week one are now joined by two more: a measurement log and a decision log.
- The measurement log records what we actually observed against the metric the scope memo named — baseline value, current value, who took the reading, when.
- The decision log records every meaningful choice made since kickoff with the reason, the alternative considered, and who signed off.
All six live in your shared folder, in plain text or markdown, version controlled. They are written so that someone you hire next year, who has never met us, can reconstruct the engagement from these files alone.
3. The falsifier list has actually moved
The falsifiers we wrote down in week one were not decoration. By week four, each one has a status:
- Still standing — the falsifier holds; the hypothesis has not been refuted and has not yet been confirmed.
- Refuted — the falsifier triggered; we were wrong; here is what we are doing about it.
- Confirmed-for-now — the predicted change occurred; we have moved the underlying claim from U to A and named the next thing that would knock it back down.
A refuted falsifier at week four is not failure; it is the system working. A frozen falsifier list at week four — one where nothing has moved in either direction — is the actual failure, because it means nothing testable has happened.
4. The client is running the ledger themselves
This is the one we care about most. By day 28, someone on your side — not us — is the person who updates the measurement log and the decision log. We will have drafted the entries in week two and walked your owner through them in week three. By week four, the pen is in their hand and we are reviewing, not writing.
The reason is structural. A ledger written only by the vendor is a marketing document. A ledger written by the client, that the vendor co-signs, is a record. Only one of those survives the engagement ending.
For the longer argument on why the ledger has to live with you, see /measurement-honesty-for-ai-projects.
5. The exit plan is no longer hypothetical
The exit-plan draft from week one had blanks in it — placeholder names, estimated costs, untested hand-off steps. By week four those are filled in (Class C). Specifically: the named successor vendor or in-house owner has been told they are named, the cost estimate has been pressure- tested against at least one real quote or salary band, and the hand-off steps have been rehearsed at least once on the smallest live artifact.
If the exit plan is still hypothetical at week four, you are locked in without knowing it.
6. The week-four review is dated and on the calendar
A one-hour review, with the six receipts open on screen, where the question on the table is: continue, rescope, or stop. We schedule this on day one and we keep it on day 28 regardless of how the month went. The meeting exists so that "keep going" is an active decision, not a default (Class E — the working norm in evidence-disciplined practice documented in /trust-receipts-standard).
What you do with this
Print the six items. Hand them to the next vendor who pitches you a month-long engagement and ask which ones they will produce by day 28. The ones who flinch are telling you something useful.
If you want this posture for your own project, the /workshop is where we work out whether the shape of your problem actually fits this discipline. Not every engagement does, and saying so up front is itself one of the receipts.