What you will build: Build one evidence trail that a colleague can check without reconstructing your analysis.
What makes a claim traceable?
A traceable claim identifies its source records, relevant period, definitions, calculation or interpretation, and review history. A reviewer can follow that trail to understand how the finding was produced. Traceability helps inspection; it does not independently establish that a source is true or a conclusion is sound.
Continue with the service example. The defensible claim is “84 of 120 responding accounts reported resolution during this period.” Beside it, show the 200-account eligible population and 60% response coverage. Then retain the source snapshot and the rules used to reach those counts.
Build the trail in the order a reviewer needs it
- Write the exact claim and its audience.
- Identify the population, period and source version.
- State the metric definition and exclusions.
- Show the calculation or reviewed interpretation.
- Provide an authorized path to the supporting evidence.
- Record corrections, reviewer and release version.
A data dictionary supplies definitions and mapping rules. It does not create provenance automatically: the system must also retain source references, version information and the operations applied to the data.
Numbers and themes need different evidence
For the 70% rate, retain the numerator, denominator and the 120 eligible response records used in that calculation. Check that one account has not been counted twice. For a claim that setup instructions are a recurring barrier, retain the coding definition, analyzed response set, supporting passages and review decisions.
A quoted comment illustrates a reported experience. It does not prove how common that experience is across nonrespondents or show what caused a change. State which responses were analyzed and how the theme was formed.
Show the right evidence to the right audience
An internal reviewer may be authorized to inspect case-level records. A board or public reader may need an aggregate result and a de-identified illustration. Do not publish names or source passages simply because they make a trail more concrete.
Where feedback was collected anonymously, preserve that boundary. The evidence trail may lead to a response batch or aggregate table rather than an identified person. Define who may inspect each level and test exports as well as on-screen access.
When monetary valuation needs more work
If a report uses SROI or another monetary valuation, preserve the outcome, valuation source, rationale, adjustments, time period and calculation. Show how different justified assumptions affect the result. A clear audit trail does not make a poorly chosen financial proxy appropriate.
This is optional depth, not a requirement for every report. The Impact Measurement & Reporting course provides a separate path into valuation after the basic evidence and reporting work.
Practice and continue
Choose one finding from your test set. Give a colleague the claim, definition, calculation and an appropriately authorized source trail. Ask them to reproduce it and identify one uncertainty. Fix the trail where they get stuck. Next, adapt the method to the kinds of records your organization actually manages.