Measurement · Comparing creators
How to compare influencer campaign performance
The creator with the most views, the highest engagement rate and the lowest cost may be three different people. Start with the job each creator was hired to do, then compare results on the same basis before deciding what to repeat.
Explore the guide| Measure | Creator A | Creator B |
|---|---|---|
| Views | 20,000 | 60,000 |
| Actions | 1,000 | 1,800 |
| Rate by views | 5% | 3% |
Start with this
Compare equivalent work at a similar age after publication, using the same metric definitions.
Show counts, rates, costs and content quality together. One leaderboard cannot answer every campaign question.
Separate what happened from why you think it happened. A useful comparison leads to a next test, not a guaranteed winner.
Decide what a successful creator contribution looks like
Write down the campaign objective before ranking creators. An awareness campaign asks a different question from a product demonstration, a library of content-only UGC or an affiliate offer. If creators were given different jobs, separate those jobs in the report rather than forcing everyone into one score.
Choose one main outcome and a small set of supporting observations. For example, an awareness comparison may focus on relevant exposure while using comments to understand the response. A demonstration may focus on whether the content answered the audience's product questions. Neither is automatically a sales comparison.
| Creator's job | Useful comparison | Evidence still needed |
|---|---|---|
| Sponsored distribution | Comparable exposure, response and cost | Audience fit, post age and whether paid distribution contributed |
| Product explanation | Accuracy, clarity and the questions people ask | A consistent comment sample and the agreed Creator Brief |
| Content-only UGC | Accepted assets, quality, rights and delivery | Separate results if the brand later distributes the assets |
| Affiliate work | Verified attributed outcomes and relevant costs | Transaction records, attribution rules and refund treatment |
Align the work and the reporting window
Start with a row for each deliverable: creator, platform, format, publication date, reporting date and agreed scope. A tutorial and a short product mention may have different purposes and production requirements. A newly published post has had less time to accumulate results than an older one.
Choose either a consistent age after publication or a clearly stated campaign snapshot. The first makes post maturity easier to compare; the second shows what the campaign had delivered by one date. If you use the snapshot, keep each post's age visible so it is not mistaken for an equal-window comparison.
Separate paid and organic distribution when the evidence permits. If one post was promoted and another was not, explain that difference. Keep platform and format differences visible too: the same label can have different counting rules across reports.
- Match the included engagement actions and denominator.
- Use one currency and a consistent cost basis, including the treatment of rights and other extras.
- Record unavailable figures and supplied private insights distinctly from public counters.
- Explain changes in the offer, brief, audience or publication timing that could affect the comparison.
Read the absolute response alongside the rate
A higher engagement rate can come from a smaller result. A lower cost per view can coexist with a higher cost per engagement. Put the raw counts next to the ratios so the reader can see which question each creator's result answers.
Illustrative comparison · two sponsored posts, same action set and reporting window
One creator leads on rate, another on volume
A has the higher action rate and lower cost per engagement. B provides more recorded views and actions, with a lower cost per 1,000 views. Costs use the same hypothetical scope, and currency ratios are rounded to two decimal places. Neither row proves better audience fit, positive sentiment or sales.
Together, these posts have 80,000 views, 2,800 actions and $1,800 of included cost. The total-action engagement rate is 3.5%; combined cost per 1,000 views is $22.50 and cost per engagement is about $0.64. Calculate these from matching totals rather than averaging the creators' individual rates.
| Measure | Creator A | Creator B |
|---|---|---|
| Included project cost (USD) | $600 | $1,200 |
| Recorded views | 20,000 | 60,000 |
| Engagement actions | 1,000 | 1,800 |
| Engagement rate by views | 5% | 3% |
| Cost per 1,000 views | $30 | $20 |
| Cost per engagement | $0.60 | $0.67 |
Compare the content and conversation as well as the totals
Review the work against the actual Creator Brief. Did it explain the intended use case accurately? Was the product visible when it needed to be? Was the agreed call to action understandable? Use specific observations rather than treating personal taste as evidence of performance.
Review comments using the same selection and classification rules for each creator. Praise for the filming is different from approval of the product. Product questions may reveal interest, confusion or a missing explanation. A large reaction count can include criticism, repeated replies or irrelevant discussion.
Keep audience relevance separate from volume. If geography or audience characteristics matter, use appropriately scoped evidence supplied for that audience; do not guess them from names, appearances or a handful of commenters.
Illustrative review note · hypothetical content
Turn a vague preference into a useful observation
Instead of writing that A made a better video, record that A showed the product being assembled and B showed only the finished result. If B's reviewed comments contain repeated setup questions, a reasonable next test is a clearer assembly demonstration. The observation does not establish that the format caused the difference in response.
Handle missing data and small samples fairly
If one creator supplies shares and saves and another supplies only likes and comments, their totals are not directly comparable. Use a common, named action set or show separate comparisons. Do not silently treat missing fields as zero or attach a precise rank to incomplete evidence.
A result based on very few recorded actions can move sharply when another action arrives. Keep the count and the post's maturity next to the rate, and resist describing a narrow difference as a stable advantage. A single campaign can help choose a test without establishing how a creator will perform in every future campaign.
If some deliverables are excluded from a calculation, list the coverage. The campaign total should not combine all costs with only the best-performing posts' results without saying that is what it does. Keep spend coverage and result coverage aligned.
Assess content-only UGC against delivery and intended use
When a creator makes content without posting it, their follower count or personal post reach is not the result you bought. Compare accepted deliverables, adherence to the agreed specification, useful variations, revision requirements and delivery timing. Keep what was originally agreed alongside what was supplied.
Check the usage scope before comparing prices. Assets with different permitted channels, durations or editing rights are not equivalent purchases. A lower production fee may still be the right choice, but it should be evaluated against the intended use rather than called cheaper for the same thing when the rights differ.
If you later test the assets in paid distribution, compare that test separately. Budget, targeting and placement can affect results, so an asset's reported ad performance is not solely a measure of the creator's contribution.
End with a decision and the evidence behind it
Use plain decisions such as repeat the format, test a clearer demonstration, collect the missing insight or reconsider the audience fit. Name the observation behind each decision and the uncertainty that remains. Avoid hiding missing data inside a single overall score.
When deciding whether to book a creator again, consider the next campaign's objective and the terms of the new work. Past results are useful context, not a promise. Keep the original comparison intact so a later test can show whether the change helped.