Proof and claims

Published: 2026-09-12 · Last reviewed: 2026-09-12 · Reading edit: 2026-09-12

The draft says your product saves teams 40% of their time. The evidence is one task that took less time in a small test. Before that sentence goes on the website, decide what the result really supports. This guide helps you make useful claims with enough context for a buyer to judge them.

Create the claim register

Use this guide before publishing outcome numbers, comparative statements, customer endorsements or AI capability promises. Give marketing an operational review process and involve the appropriate legal or technical reviewer where needed. Advertising obligations differ by jurisdiction and subject; this is not a universal legal approval checklist.

Claim type	Useful support	Common mismatch
Capability	Reproducible test on the stated version	Demo succeeds only on a curated input
Customer result	Permission, baseline, period and method	One customer’s result presented as typical
Comparison	Current, equivalent conditions and scope	Different packages or historical prices
Forecast	Explicit assumptions and uncertainty	Estimated benefit presented as achieved
Endorsement	Accurate experience and required disclosures	Incentive or relationship hidden from readers

Review one claim end to end

1. Capture the exact wording and placement. Include surrounding visuals and the audience. “Up to” does not automatically repair an unsupported number.
2. Write the interpretation a reader is likely to take away. Ask a person outside the launch team to paraphrase it. Review both the literal and implied promise.
3. Attach the evidence. Record source, owner, date, method, sample, version and permission to use it. Separate measured observations from modeled estimates.
4. Narrow the wording to the evidence. If a result concerns one workflow in one deployment, retain those conditions. Put material qualifications where people will notice them, not only in a distant footnote.
5. Assign approval and expiry. Technical owners verify capability; customer owners verify permission and context; the relevant reviewer checks regulated or comparative claims. Expiry should follow the evidence’s volatility.
6. Track every published use. When evidence changes, update ads, sales decks, partner materials and web pages together. Retiring a claim means removing its distribution copies, not merely changing the master deck.

Worked example

A fictional customer reduces a task from 100 to 60 minutes in a ten-task pilot. That is a 40% reduction in measured task time under the pilot’s conditions. It is not automatically a 40% labor-cost saving, company-wide productivity gain or expected result for every customer.

The team publishes the pilot context with permission, explains the tasks and excludes an unrelated setup period explicitly. It does not hide failed runs. A wider claim waits for evidence from representative work and a clear measurement method.

Copyable review record

Exact claim and locations:
Likely interpretation:
Evidence and reproducibility:
Population, baseline, time window and exclusions:
Permission and relationship disclosures:
Approved wording and visible qualifications:
Reviewer / owner / expiry:
Correction or withdrawal procedure:

Release decision

Approve, narrow, investigate or remove. If evidence does not exist, change the promise before publishing. Track unsupported-claim corrections and expired evidence as maintenance work. Do not turn “a human reviewed it” into proof that a numerical or legal assertion is sound.

Try it with your own work

Choose one headline on your website. Find the evidence behind it and write down exactly who, what, and which period it covers. Revise the headline if it promises more than that evidence shows.

Sources and scope

• US Federal Trade Commission: advertising and marketing basics explains the US expectation that advertising be truthful, non-deceptive and supported. Consult applicable local requirements for your audience and claim.

The example is fictional; any numbers illustrate the method rather than a benchmark. Adapt the worksheet to your own situation.

What to read next

Apply the register to a case study, comparison page or product launch.

Chapter guide · All playbooks

---

Copyright © 2026 Ivan Xu. All rights reserved. See the copyright and reuse terms.

Canonical source: github.com/weilun88313/B2B-Playbook