What to measure in a creator campaign
Most creator-campaign reporting measures the thing that is easiest to collect rather than the thing the campaign was for. The useful discipline is to decide what question you are answering before you look at any number, because each of these metrics answers a different one and none of them answers all three.
A view is not one thing
Each platform counts a view differently, and the differences are large enough to make cross-platform totals meaningless if you add them naively. Before comparing anything, find out what the number you are looking at is counting on that platform — the thresholds differ, autoplay is treated differently, and a repeat view may or may not be included.
The practical consequence: compare a platform against itself over time, not against another platform at a point in time. A campaign that “did better on TikTok than YouTube” may simply have been counted more generously.
Where the number comes from matters more than the number
A view count typed into a spreadsheet by the person being paid for it is not a measurement. Ask any platform or agency one question: what is the source, and can you show it. If the answer is a screenshot of someone's analytics, you have no audit trail and no way to tell a strong campaign from a well-screenshotted one.
On this platform the count is read from each platform's own API through an account the creator has verified, refreshed while the campaign runs. That is not a claim about honesty; it is a claim about where the number comes from, which is the only part you can check. How fake views are spotted covers the rest.
Engagement rate is the most misleading one
It is a ratio, so it moves when either side moves. A video that reaches far beyond its creator's followers — which is the outcome you were paying for — will show a lower engagement rate than one that reached only the people who already follow them, because the denominator grew faster than the numerator. Optimising for engagement rate on a reach campaign means optimising against yourself.
It is a reasonable health check on a single creator's typical output. It is a bad way to compare creators, and a worse way to judge a campaign whose point was distribution.
Three questions, three different numbers
- Did it reach people? Verified views, by platform, over time. Not impressions, not reach as the platform defines it, and not the sum of counts across platforms.
- Did it land? Retention and completion on the individual posts — where the graph drops tells you whether the opening worked. This is the number that improves the next brief.
- Did it do anything for the business? Whatever you measured before the campaign started, measured the same way afterwards. If that was not set up beforehand, this question cannot be answered retrospectively, and no amount of platform data will fill the gap.
The attribution problem, stated honestly
Short-form video is mostly an upper-funnel channel watched by people who are not about to click anything. Last-click attribution will therefore under-credit it, often severely, and a UTM report will tell you a campaign did nothing when it moved branded search and direct traffic instead.
What works better is unglamorous: hold the campaign period against a comparable period before it, watch branded search volume and direct sessions, and accept a range rather than a figure. Anybody offering precise per-view revenue attribution for a feed video is selling certainty that does not exist.
What to set up before launch, not after
Decide the one question the campaign exists to answer, and write down how you will know. Record the baseline for whatever you will compare against — branded search, direct traffic, signups, whatever it is — for a period at least as long as the campaign will run. Agree what the floor is: the result below which you would not run it again.
That last one is the discipline most campaigns skip, and it is why so many get described afterwards as “good brand awareness”. A number you decided on in advance is a result. A number you interpret afterwards is a story.
The report worth asking for
Whatever you are buying, a campaign report should let you answer one question without trusting anybody: where did each number come from. A report that is a chart with no source is a design exercise.
- Per post, not just totals. A total hides the distribution, and the distribution is the finding — one clip carrying the whole campaign means something very different from twenty clips performing evenly.
- Each post's link. So you can open it, watch it, and see that it is still up. A clip that was removed after it was counted is a thing that happens.
- The source of the count. Platform API, scraped, or self-reported. These are not equivalent and should not be presented as one column.
- When it was last refreshed. View counts keep moving; a figure with no timestamp cannot be compared with anything.
- What was rejected and why. The clips that did not count tell you whether the brief was clear, which is the part that improves the next campaign.
What not to put in the deck
Combined view totals across platforms. They are counted differently; adding them produces a number that is larger than any of its parts and means less.
“Estimated reach” or “potential impressions”. Follower counts multiplied by something. It is not a measurement of anything that happened.
Equivalent media value. A figure invented by multiplying views by a rate you did not pay, to produce a number you did not earn. It survives because it flatters the person presenting it.
Common questions
Should I pay per view or per post?
Per view ties spend to what arrived; a flat fee is spent whether anything was watched or not. The trade is predictability — a fee is a known cost. The comparison is here.
How long should I wait before judging a campaign?
Longer than feels natural. Short-form video has a long tail: a clip can sit flat for days and then be picked up. Judging at 48 hours mostly measures how fast a platform's recommendation system made up its mind.
What about cost per thousand views?
It is the right unit for comparing campaigns against each other, and a bad unit for comparing channels, because a view here and an impression elsewhere are not the same event. What moves it is covered in what determines your CPM.
Can I measure sentiment?
Comments are worth reading and are not worth scoring. On a campaign of any size they are too few and too noisy to produce a number, and reading twenty of them tells you more about the brief than any sentiment metric will.
Do I need the creators' own analytics?
For retention, yes — it is not public. For views, no, and you should prefer a source the creator cannot edit.
The short version
Pick the question first. Verified views for reach, retention for whether it landed, and a pre-registered business metric for whether it mattered. Treat engagement rate as a per-creator health check and never as a campaign scorecard. Decide the floor before you launch.
Read next: how to write a brief that gets usable clips, or how campaigns work here.