An interviewee says a new workflow “made the week much easier.” The draft becomes “The software cut weekly administration by 50%.” Somewhere between the conversation and the headline, a qualitative impression has turned into an unsupported measurement.
A claims and evidence register makes that change visible. It connects each proposed statement with the relevant source, its limits, any calculation and the decision about publication. A writer can keep it in an ordinary document or spreadsheet. The important part is that another reviewer can follow the reasoning without relying on the writer’s memory.
Give consequential statements an identifier
You do not need a ledger for every linking sentence. Start with numerical results, comparisons, attributed quotations, descriptions of the customer’s process and statements likely to be reused as standalone proof.
Assign a short ID such as C01. Record the proposed wording, source reference, source wording or factual detail, time period, scope and current decision. For a recorded interview, include the recording version and timecode. For a document, include the file version and relevant section. A link to an entire shared folder is not enough to locate the evidence.
Keep exact quotations visibly separate from paraphrases. A tidy sentence written by the agency is not automatically something the customer said. If you propose edited wording for approval, label it as proposed wording rather than backdating it into the transcript.
A completed register
Fictional exercise: a customer describes introducing a shared scheduling process at one location. All statements and sources below are invented; none are evidence of a real customer outcome.
| ID and proposed statement | Source and context | Decision | Follow-up |
|---|---|---|---|
| C01: “Scheduling became easier for the coordinator.” | Interview v1, 12:10; coordinator describes seeing requests in one place | Candidate paraphrase, limited to that person’s experience | Customer checks the meaning in draft 1 |
| C02: “Administration fell by 50%.” | Supplier’s initial brief; interview provides no measured comparison | Exclude from draft | Reconsider only if a defined, supportable calculation is supplied |
| C03: “The team used the process at every location.” | Interview v1, 18:40 explicitly discusses one location | Reject this wording | Use “at the location discussed” if relevant |
| C04: A named quotation for a sales slide | Candidate quote at 21:05; website story permission is being discussed | Hold the slide use | Confirm final wording, attribution and sales-slide use separately |
The register preserves useful material without trying to rescue every proposed claim. The absence of a percentage does not prevent a clear process story.
Treat calculation and attribution as different checks
If a result is supplied, ask what was measured and how. Record the baseline, comparison period, population, exclusions, units and source owner. A measured change may still have several contributing causes. Do not assign all of it to one product simply because the story concerns that product.
For a hypothetical example, a customer may provide logs showing 30 requests in one period and 20 in another. That establishes a difference in recorded request counts under the supplied definitions. It does not alone establish time saved, a product effect or a result every customer can expect.
Only include a calculation when its inputs and meaning can be reviewed. Where the evidence supports a narrower statement, write the narrower statement. Where the comparison is not suitable for publication, leave it out instead of replacing it with a dramatic but vague claim.
Review quotations in the context where they will appear
A quotation may be accurate word for word yet misleading after its qualification is removed. Keep the preceding question and relevant surrounding passage available to the reviewer. Note whether the participant was describing a particular team, period or situation.
For US-facing endorsement work, the FTC’s endorsement guidance explains that endorsements must be honest and cannot make otherwise unsupported advertising claims. A customer quote is therefore not a substitute for checking the claim. The register is a practical editorial record; it does not itself establish that every jurisdiction’s publication requirements are met.
Keep evidence status separate from permission status
A statement can be well supported but not cleared for the proposed use. It can also be approved by a stakeholder while still lacking an adequate basis. Give those questions separate fields.
In the fictional register, C01 may progress from “candidate paraphrase” to “meaning confirmed,” while the final named use remains pending. C04 stays on hold for the sales slide even if the website article is ready. Record the actual asset version and permitted context rather than attaching a single permanent “approved” badge to the customer.
Use the approval map to identify who can resolve each open decision. Do not infer permission from a job title or from the fact that someone attended the interview.
Hand the register over with the story
Before release, compare the final copy against the register. Check headlines, captions, pull quotes and thumbnails as well as the article body. A cautious paragraph cannot repair an overstated headline that travels on its own.
Keep removed statements and their reasons so they do not return during a later edit. Give Sales a reuse handoff identifying which passages can support which tasks. The register earns its place when it helps a colleague make an accurate new asset without reopening the same factual questions or broadening the customer’s account.