Measuring creator campaigns, and why provenance matters

A report that says a campaign reached a certain number of people looks precise. It tells you nothing about who produced the figure, how it was collected, or how far to trust it. Two campaigns can show the same number and rest on very different evidence.

Provenance is the answer: next to every figure, a label that says where it came from. It does not make the number bigger or smaller. It lets the reader decide how much weight to give it.

These are the labels used in MOWJ. Whatever tool you use, the distinction is worth keeping.

Verified platform data

Figures taken from a social platform's own reporting and checked. Usually the strongest evidence, when it is available.

Imported data

Figures brought in from a file or an export. As reliable as the file, so keep the original.

Creator-reported

Entered by the creator. Useful and often accurate, and still a claim made by the party being paid.

Brand-reported

Entered by the brand, for example results from its own website or store.

Tracked

Counted through a tracking link or code attached to the campaign. Good for actions such as clicks and conversions.

Estimated

Calculated or assumed. Acceptable when labelled as such, misleading when not.

Choose measures after you choose the objective, not before. Each objective points to a different set of numbers.

• Awareness: views, reach and impressions.
• Engagement: engagements, and the engagement rate once reach or impressions are known.
• Traffic: clicks, and the cost per click once spend is known.
• Leads, sales and installs: conversions, sales and revenue, and the cost per conversion or return on ad spend when both sides are known.

Reporting every possible measure for every campaign is a way of hiding the one that matters.

Ratios are only as good as both of their parts. Engagement rate is engagements divided by reach or impressions. Cost per thousand impressions is spend divided by impressions, times one thousand. Cost per click is spend divided by clicks. Cost per conversion is spend divided by conversions. Return on ad spend is revenue divided by spend.

If one input is missing or estimated, the result is too. A sound report shows a derived measure only when its inputs exist, and carries the weaker label forward.

• Which measures will be reported for each deliverable.
• Who reports each one, and by when.
• What evidence supports it: a platform screenshot, an export, a tracked link report.
• How long after publication results will be collected.

Settling this in the brief or the agreement prevents the most common argument in creator marketing, which is about what the numbers should have been.

Do not add a verified figure to an estimated one and present the total as fact. Keep the labels visible in the report. If you must summarise, summarise by label: what is verified, what is reported by the creator, what is tracked, what is estimated.

Record the date of each observation. Audience figures change, and a number without a date is hard to interpret later.

Suppose a brand runs a campaign to send people to a product page. The creator reports views and reach from their own analytics. The brand tracks clicks through a link. The brand's store reports orders and revenue.

That is three sources and three labels: creator-reported for views and reach, tracked for clicks, and brand-reported for orders and revenue. Cost per click can be worked out from the brand's spend and the tracked clicks. Engagement rate can be worked out only if the creator also reported engagements, and it inherits the creator-reported label.

A report that lists each number with its label lets a reader see at once which figures are firm and which rest on someone's word.

• Treating reach and impressions as the same measure.
• Quoting a total that blends verified and estimated figures.
• Comparing engagement rates calculated with different denominators.
• Leaving out the date a number was observed.
• Treating attribution as certain when it is a judgement.

Every metric in MOWJ carries a provenance label and a date. Derived measures appear only when their inputs exist. At launch there is no automatic connection to social platforms, no scraping and no auto-posting, so results are entered or imported by people and labelled accordingly. Creator audience numbers on profiles are creator-reported.