Decide the review before the launch
Evaluation begins in the brief. Agree the primary objective, reporting window and available evidence before anyone publishes. A campaign intended to explain a complex application technique should not be judged only by immediate purchases. Equally, a sales-led brief should not quietly substitute a large view count when transaction evidence disappoints. Record any change in objective so the final report does not rewrite the original assignment.
Separate fulfilment from effectiveness. Delivering the agreed posts, meeting the publication window and applying the agreed disclosure are operational outcomes. Reach, meaningful viewing and website visits describe audience response. Incremental revenue or a change in brand consideration requires a different evidential standard. A useful review can recognise excellent delivery while acknowledging that the commercial result is uncertain.
Put definitions next to the numbers
For every metric, record the source, extraction date, measurement period and definition. Reach generally aims to describe distinct accounts or people within a system, while impressions describe exposures; neither should be casually substituted for the other. Video view definitions and replays vary across platforms. Check current platform documentation rather than assuming that similarly named metrics are directly comparable.
Engagement rate is especially easy to misread. State which interactions are included and which denominator is used. Interactions divided by followers answers a different question from interactions divided by reach. A percentage without its denominator is not a complete result. Distinguish organic distribution from paid amplification and avoid adding platform reach figures together as if the same person could not appear in more than one audience.
| Evidence | Supports a statement about | Does not establish alone |
|---|---|---|
| Published asset record | Delivery against scope | Audience attention |
| Platform reach and viewing | Reported distribution and consumption | Unique cross-platform exposure |
| Tracked purchases | Sales credited by a defined system | Incremental sales caused by the creator |
| Well-designed controlled study | Estimated effect under stated assumptions | A universal result for future campaigns |
Read tracked sales carefully
A tagged link or creator code can help connect an action to a campaign, but it cannot observe every influence on a purchase. People change devices, return later through search, share codes or encounter several advertisements. Codes can also circulate beyond the intended audience. Describe the attribution window and model, and distinguish placed orders from completed, non-returned purchases.
Google Analytics 4, a platform ad manager and an ecommerce system can report different totals because they use different identities, time zones, consent conditions and attribution rules. Reconciliation should explain these differences rather than selecting the largest total. Do not ask a creator to supply customer-level personal data merely to validate a campaign when aggregate reporting is sufficient. Privacy choices are part of the measurement environment, not a defect to work around.
Keep costs and comparisons honest
Explain which costs are included in an efficiency measure. Creator fees, production, product seeding, shipping, agency support and paid media can materially change the interpretation. If reporting cost per thousand impressions, state the cost boundary and the source of the impression count. Return on ad spend compares attributed revenue with advertising spend; it is not automatically profit or a complete return on the partnership investment.
Use relevant comparisons. A short product reveal, a long tutorial and a paid retargeting advertisement do different jobs. Compare similar formats and conditions where possible, and record changes in stock, landing pages, promotion periods or distribution. Historical averages can provide context, but a small or selectively chosen sample does not create a reliable benchmark. Avoid ranking creators as if platform volatility were entirely under their control.
Close with decisions, not just a score
Review qualitative evidence alongside the totals. Did viewers understand the demonstration? Were recurring product questions answered? Did comments expose confusion about shade availability or the sponsorship? Summarise themes using a stated method, and avoid treating sentiment from a handful of comments as a population estimate. Preserve examples only when necessary and handle personal information proportionately.
The Media Rating Council’s standards and IAB measurement resources can help teams ask more precise questions about advertising measurement, but referencing a standard does not make a campaign report accredited. End with a short account of what was delivered, what was observed, what could not be established and what should change next time. Where causal impact matters, consider a properly designed experiment with qualified measurement support. Do not retrospectively label a convenient comparison group as a controlled test.
“The system attributed purchases to this code” is narrower—and often more defensible—than “this creator caused every one of these purchases.”