Operations
Part of Reading campaign measurement with a sceptical eye
Campaign measurement checklist: claim and chart gates before release
A campaign measurement checklist for the release stage: the claim gate, the chart and access gate, and how to sign off each one before publishing.
What to take away
- Run these gates after the analysis and before any number leaves the team.
- Approve evidence and wording together, one sentence at a time.
- A chart must carry its meaning in text and without color alone.
- A campaign can be safe to monitor internally and still lack the evidence for an external claim.
- A failed material gate blocks the affected use until it is fixed or disclosed.
A campaign measurement checklist creates a reviewable record for launch, analysis and publication. These gates cover only the last step, release, where a number becomes a claim and a chart becomes something other people must be able to read and act on.
The claim gate
Every factual sentence needs a record behind it before it is published, whether it appears in a deck, a dashboard note or a client email. The FTC's policy statement on advertising substantiation explains its expectation that advertisers possess a reasonable basis for objective claims before dissemination.
Legal application depends on the facts, so treat the policy as a working discipline rather than a verdict. Approve the evidence and the wording together. The overclaims this gate stops are catalogued in campaign measurement mistakes, and each one starts as a sentence nobody checked against its record.
- Match each factual sentence to the record that supports it
- Distinguish delivered, observed, attributed, modeled, surveyed and incremental values
- Show population, period, absolute result and the relevant denominator
- State uncertainty, exclusions, data loss and material changes
- Remove causal, comparative or universal wording the design cannot support
- Have a qualified reviewer challenge the claim independently
Wording is where claims drift. A result observed in one population gets reported as true for every audience, or a modeled value is written as if it were counted. If the record says modeled, the sentence says modeled. A comparison to a peer figure is a claim too, and it has to pass the tests in campaign measurement benchmarks first.
The chart and access gate
A chart is a claim in picture form, so it passes the same review as a sentence. A text alternative has to serve the purpose of the chart, and the measurement owner remains responsible for the data and the interpretation, whatever the charting tool does with them.
Item / Check
- Chart
- Title, units, population, period, scale and source are visible
- Alternative
- Equivalent finding and material limitation are available in text
- Color
- Meaning remains clear without color alone
- Table
- Headers and reading order are usable
- Correction
- Owner can replace values and notify recipients
- Archive
- Brief, data, logic, review, decision and version are retained
Section508.gov guidance on color states that color can accentuate content but cannot be the only way information reaches viewers. Check the chart with the color taken out, and if two series become indistinguishable, add labels or markers before it goes anywhere.
Complex findings do not need long alternative text. They need a concise summary in words plus an accessible data table, so a reader using a screen reader gets the same finding and the same limitation as everyone else, and nobody has to decode a trend from a picture.
Corrections and the archive
A published figure can turn out to be wrong, and the correction path has to exist before that day. The owner named in the table can replace values and notify recipients, so write the recipient list when the report is released, not when the error is found.
The archive holds the brief, data, logic, review, decision and version, which is what lets a reviewer rebuild the figure later. Before the first release, confirm the measurement setup so that the archive records something worth keeping rather than a dashboard nobody can reproduce.
Sign off twice
Sign the checklist twice: once for operational fitness and once for publication fitness. A campaign can be safe to monitor internally while still lacking the evidence for an external claim. The two sign-offs can have different reviewers and different outcomes, and neither replaces the other.
Record exceptions with an owner, a temporary control, a deadline and the uses that stay prohibited. An exception without a deadline is a permanent waiver, so give it one. Review depth should follow risk, as set out in campaign measurement best practices.
Mark each item pass, fail, not applicable or excepted, and note where the evidence sits, who reviewed it and when. Stop the work when a mandatory privacy, security, data-quality, accessibility or correction condition fails, even if the rest of the score looks favorable.
