Decision receipts: proving why a construction choice was made
A receipt is not a report. It is the reconstructable record of one decision: its scope, its evidence, the rules applied, and the exact versions in force.
Reports summarise. Receipts prove. A construction report tells you what the totals are today; a decision receipt tells you how a specific conclusion was reached, from which evidence, under which rule versions — and lets you run it again to check.
What a receipt has to contain
- Scope — exactly which elements the decision covers, addressed to a specific model version rather than a loose selection.
- Evidence — the extracted properties and classifications the decision consumed, with their provenance intact.
- Authority — which contract, recipe, material class and price book governed the result, each at a pinned version.
- Rules — the deterministic logic that ran, identified well enough to re-run it.
- Outputs — the resolved system, products, quantities, cost and carbon figures.
- Unresolved scope — what could not be decided, and why.
Why versions matter more than values
A price of 412 SEK/m² is meaningless on its own. A price of 412 SEK/m² resolved from price book snapshot v1, effective 2026-04-01, in SEK, for product version 3, is a fact you can defend. The same applies to recipes, material classes and classification rules: the value is only trustworthy because the version behind it is pinned.
This is also what makes change legible. When a price book is superseded, older receipts do not silently change. They keep pointing at the snapshot that was in force, and the difference between old and new becomes a visible delta rather than an invisible drift.
Append-only, not editable
Receipts are written, not amended. If a decision changes, a new commitment is recorded and the previous one remains readable. An editable audit trail is not an audit trail — it is a document with a version history nobody trusts.
Replay as a conflict-resolution tool
The practical payoff arrives during disagreement. A contractor questions a quantity. A client questions a carbon figure. A reviewer questions a substitution. In each case the answer is the same: replay the receipt. If the replay reproduces the figure, the conversation moves to whether the inputs were right. If it does not, you have found a real defect rather than a difference of opinion.
What a receipt deliberately does not do
- It does not smooth over missing evidence with a plausible default.
- It does not merge incompatible units, currencies or scopes to produce a tidier number.
- It does not treat a label from an authoring tool as authority simply because the label exists.
- It does not present a scenario or what-if exploration as a committed decision.
Each of those refusals costs a little short-term convenience and buys a great deal of long-term defensibility. That trade is the entire point.