Start a creator performance review with follower count and you rank the wrong people first. It happens in most small TikTok Shop teams. Follower count is the only number that arrives on every profile before you work with anyone.
The number that actually predicts whether a creator sells is GPM, gross merchandise value per thousand video views. It is available after one video, it is comparable across account sizes, and it reorders the list almost every time.
Written for sellers running ten to fifty SKUs who work with twenty to two hundred creators and review them monthly. Not for teams running paid amplification, where the economics are different.
Why the first sort decides the whole review
A creator performance review is a sorting exercise. Whichever column you sort by becomes the definition of a good creator for the next quarter. Every decision downstream inherits it.
Follower count is the default because it is visible before any work happens. That is its only advantage. It measures accumulated attention, not current purchase intent, and it says nothing about whether that attention converts for your category.
GPM corrects for the obvious distortion. A creator with 400,000 followers and 3,000 views per video is not reaching 400,000 people. A creator with 6,000 followers and 40,000 views per video reaches far more people than her follower count suggests. Sorting by the first number hides both facts.
What GPM measures, and the three cases that fool it
GPM is total product sales driven by a video, divided by that video’s views, multiplied by a thousand. It answers one question: how much revenue does this creator generate per thousand people who watched.
The bands we use, for mid-priced categories between fifteen and sixty dollars, look like this. Under 100 is weak and usually a fit problem. Between 200 and 500 is solid. Above 800 is worth protecting and paying more for. These bands are a starting point, not a universal scale.
Three cases still fool the number, and a review that ignores them will drop good creators. That is a limit of the metric, not of the people it measures.
Low view counts are the first. A video with 900 views can post a spectacular GPM on two orders. That is arithmetic, not performance. Below roughly 5,000 views, treat GPM as noise and look at the creator again next month. DAMI shows views beside GPM at the video level. A small sample then looks small, instead of looking like a winner.
High average order value is the second. A 300 dollar item shows a high GPM on the same conversion rate as a 25 dollar one. Comparing them on GPM alone penalises the cheaper catalogue. Compare within a price band, not across one.
Attribution lag is the third. Some sales land days after the video, especially for considered purchases. A review run seven days after posting will undercount those creators and overcount the ones with instant-purchase products.
One more limit, and it matters most for fashion and footwear. GPM is built on gross sales, before returns. A creator driving a 25 percent return rate can look strong on GPM and weak on net contribution. In those categories the review sheet below is not optional.
A review sheet that survives a monthly pass
The benefit of a fixed sheet is that it stops the review from becoming an argument about who feels promising. Six columns are enough, and the last two do the real work.
| Column | Source | What it decides |
|---|---|---|
| Views per video | Short video detail | Whether GPM is even readable this month |
| GPM | Short video detail | The sort order for the whole review |
| Videos posted | Content log | Whether the sample paid for itself |
| Commission paid | Affiliate report | Cost side of the same creator |
| Net contribution | GPM minus commission and sample cost | Whether to rebook, hold or drop |
| Reply latency | Chat history | Whether a rebook is worth the wait |
The last two columns are where most reviews stop too early. A creator with a strong GPM and a 30 percent commission can still lose money per unit. A creator who takes eleven days to answer a message costs a slot in a launch week, whatever her numbers say.
One honest limit. Net contribution needs the sample cost allocated per creator, and most teams allocate it by guesswork. If the sample cost is not tracked against a specific creator, the fifth column is fiction. DAMI keeps that log against the creator record, which is the only reason the column can exist at all.
The review itself takes about two hours for a hundred creators, provided the data is already in one place. The problem is that it usually is not. Views live in one screen, commission in another, and the sample log in a spreadsheet nobody updated after week two.
DAMI puts the video-level view and revenue figures beside the commission paid. The sort then happens on one screen, not across three exports. That is the difference between a review that gets run monthly and one that gets run twice and then abandoned.
For the benchmark side, what counts as a good GPM on TikTok Shop videos covers category variation in detail. If every review concludes that revenue per creator is flat, measuring something other than GMV per creator is the next step. A total hides the variance that matters.
The DAMI analytics workspace is where those columns sit together. The analytics pillar explains which report to read first.
FAQ
Should a creator with a high GPM but only one video be rebooked?
Once more, yes, and with a specific ask. One video is a sample of one. The second either confirms the number or reveals it as a lucky format. Do not raise the commission after a single result. Raise it after the second, and only if the first video also produced repeat buyers.
What is a reasonable GPM to expect from a micro creator?
Often higher than from a large account, because the audience is narrower and the recommendation reads as personal. It is common for accounts under 20,000 followers to beat larger ones on GPM while losing on total volume. The decision then depends on whether you need reach or revenue this quarter.
How often should the review run?
Monthly for active creators, and never inside a campaign week. A creator performance review run mid-launch changes decisions while the data is least reliable. A review that lands mid-launch changes decisions while they are in flight. That is the one week when the data is least reliable and the interruption costs most.
