Twelve creator videos went live for one serum launch in Southeast Asia. Together they collected just over 200,000 plays in three weeks. The brand shipped eleven orders. When the team pulled the numbers to decide which creator to rebook, they picked the video with 84,000 plays, the obvious winner by a factor of three.
That video had sold nothing. The eleven orders came from a 6,400-play clip posted by a mid-tier creator whose audience was almost entirely women aged 25 to 34 in the same metro area as the brand’s warehouse. The 84,000-play video had been watched by an audience the brand could not ship to cheaply, and its product link had been tapped 31 times.
This is not an unusual story, and it is not a case of bad luck. It is what happens when a team has one number on the dashboard and no habit of reading the second layer. For a beauty brand running 10 to 50 SKUs with forty-plus active creators, the difference between reading plays and reading revenue is the difference between rebooking the right person and burning next month’s sample budget on the wrong one.
Why views and GMV diverge
Plays measure whether the opening frame worked. GMV measures whether four more things worked after that: the product was legible, the price was acceptable, the viewer could actually buy from where they were, and the link was placed at a moment when intent existed.
Any one of those can break while plays keep climbing. A clip that opens with an unboxing sound and a hand on camera can hold attention for eight seconds without ever making the product’s function clear. A clip with a great hook can push traffic to a listing that is out of stock in that region. A clip can go wide in a country you do not serve.
The divergence is strongest in beauty for a reason worth naming. Beauty purchases are frequently the second or third touch, not the first. One clip builds recognition, another one converts, and the two are rarely the same clip. If you only measure first touches, you will conclude that your best-performing video sold nothing while a quieter one did all the work, and you will draw the opposite of the correct conclusion about which creative direction to repeat.
The four numbers that identify the seller
Four fields, read together, will identify the video that moved inventory in almost every case. None of them is plays.
| Number | What it means | How teams misread it |
|---|---|---|
| GMV per video | Attributed revenue for that clip | Treated as a ranking of creators, when it mostly ranks product-audience fit |
| GPM | Revenue per thousand plays | Ignored because it looks like a vanity ratio; it is the only field that normalizes for reach |
| Click-through rate | Plays that became product taps | Read as creative quality when it often reflects link placement and timing |
| Attached product | Which SKU the clip actually links to | Skipped entirely, which is how a video gets credited for selling the wrong item |
GPM is the one that changes decisions. A 6,400-play video generating eleven orders is running a GPM far above a 84,000-play video generating none, and once you sort by GPM the ranking inverts. That inversion is the entire point of the exercise.
The caveat: GPM is unstable at low volume. A clip with 900 plays and two orders produces a GPM number that looks spectacular and means nothing, because two orders is not a sample. Set a floor before you rank anything, somewhere around 3,000 plays, and treat everything below it as “not enough data” rather than “high potential”.
Read the overview first, then the detail
There are two layers of reporting and they answer different questions. The overview answers “what is happening to the account”; the detail answers “which asset caused it”. Confusing the two is why weekly reviews turn into arguments.
The overview layer covers GMV, average daily visitors, SKU orders, product impressions, clicks, and click-through, shown as a trend line with filters by time period and account type. Use it to spot shape, not to make decisions: a GMV line that rose while impressions fell tells you conversion improved, which is a different story from a GMV line that rose on the back of impressions.
The detail layer is per video: poster, post time, length, creator type, hashtags, plus GMV, GPM, plays, click-through, visitors, orders, and the attached product, sortable by transaction value. This is the layer where you decide who to rebook and what to repeat.
The practical sequence for a weekly review is fixed: look at the overview to see whether the week was traffic-driven or conversion-driven, then open the detail sorted by GPM to find the two or three videos worth studying, then read those specific clips frame by frame against the winners from the previous week. In DAMI, both layers are in the short video module, which is the reason to run the review there rather than exporting to a spreadsheet: the detail view keeps the attached product and the creator record on the same screen, so a video that sold can be traced to the person who made it and added to the creator list in the same pass.
Why beauty breaks the pattern, and what to look at instead
Beauty has three properties that make the plays-to-GMV gap wider than in most categories, and a team that ignores them will keep rebooking the wrong creators.
First, the product is not visually self-explanatory. A charger shows what it does in one second. A serum does not. Beauty clips therefore depend on verbal claims and before-and-after framing, which means the highest-converting clip is often the one with the least dramatic visuals and the clearest claim.
Second, repeat purchase distorts attribution. Existing customers buy again after seeing a new clip that had nothing to do with their decision. That order gets attributed to the clip, the clip looks like a winner, and you scale a creative that did not cause the sale. There is no clean fix for this in platform reporting; the mitigation is to compare new-customer share alongside GMV, not instead of it.
Third, the audience geography matters more than the audience size. A clip that travels to a market you do not serve generates plays, clicks, and refunds, in that order. Filter by where your warehouse ships before you rank anything, and tag the creators who cleared that filter in DAMI’s creator library so the next campaign starts with the geography already applied.
For a brand in the 10 to 50 SKU range, the practical implication is to track per-creator GPM across at least two clips before rebooking. One clip is a coin flip. Two clips by the same person with GPM above your account median is a pattern worth paying for.
The three clip types that show up in a beauty program do not deserve the same yardstick:
| Clip type | Judge it by | Ignore |
|---|---|---|
| Recognition (routine, texture, shelf context) | New-customer share and saves | Direct GMV, which will look near zero by design |
| Conversion (problem, claim, before-after) | GPM and click-through | Plays, which reward hooks rather than arguments |
| Retention (how-to, pairing, restock) | Repeat purchase rate on that SKU | Attribution, which will credit the clip for orders that were already coming |
Mixing these three into one ranking is the second most common mistake after ignoring GPM. A recognition clip that produced almost no direct revenue is not a failure; it is doing a job that shows up two weeks later in a different clip’s numbers.
What the numbers cannot tell you
Three things this data will not answer, and pretending otherwise is how teams end up with confident wrong conclusions.
It will not tell you why a video worked. Attribution tells you which clip carried the sale, not which frame caused it. For that you have to watch the clip and compare it against a failure, which is slow, manual, and the only thing that actually improves creative.
It will not separate the clip from the offer. If a creator’s clip sold well because the brand ran a discount that week, you will attribute the lift to the creator and pay for it again next month at full price. Check whether a promotion overlapped before you rebook.
It will not rescue a product nobody wants. A well-executed clip on a poorly priced SKU will still produce a low GPM, and no amount of creator selection fixes that. The attribution method for creator ROI matters here, because double-counting shared sales is the fastest way to overpay a creator who was riding an already-popular product.
One more limit worth stating: platform-side attribution windows are short and vary by market. Orders that land after the window get credited nowhere, which systematically undercounts clips with long consideration cycles. If your category has a long consideration cycle, add your own post-purchase survey question rather than trusting the dashboard alone.
Building the weekly read
A review that takes twenty minutes and produces two decisions is worth more than an hour-long meeting that produces a slide. The shape that works:
Step one, set the floor. Drop every video under roughly 3,000 plays from the ranking. They are noise, and including them guarantees you will chase a fluke.
Step two, sort by GPM, not GMV. GMV rewards reach; GPM rewards efficiency. You are looking for the clips that converted disproportionately, because those are the ones whose structure you can copy. In DAMI the detail view sorts on GPM and transaction value without an export step, which is the difference between a routine you keep and one you abandon in week three.
Step three, take the top two and the bottom two. Read all four. The failures are more informative than the winners, because a failure usually has one identifiable break: no product legibility, wrong market, link buried, price mismatch.
Step four, make exactly two decisions. One creator to rebook, one structural change to test next week. More than two and you will not be able to tell which change moved the number.
Then log the creators who cleared the GPM bar into a tagged list, so the next campaign starts from a shortlist instead of from the full database. In DAMI that means tagging the creator record in the resource library with the same label across campaigns, which is the difference between a list that compounds and a list rebuilt every month.
The cost of this routine is real: about twenty minutes a week plus the time to watch four clips properly. Teams that skip the watching part end up with a dashboard they check and a creative direction they never change.
FAQ
Should I rank by GMV or by GPM?
GPM, with a plays floor. GMV ranks reach, GPM ranks efficiency. Use GMV only when you are allocating a fixed budget across creators who already cleared the GPM bar.
My best video has high plays and zero orders. Is it broken?
Usually not broken, just aimed at the wrong audience or shipping region. Check the attached product, the market breakdown, and whether the link was placed before the moment of intent.
How long before a video’s numbers mean anything?
Give it seven days and at least 3,000 plays. Under that, you are reading noise. Beauty clips with long consideration cycles may need two weeks before attribution settles.
Do I need the detail export or is the overview enough?
The overview is enough for a monthly trend conversation. It is not enough to decide who to rebook. Per-video detail with the attached product is the minimum for that decision.
Can I compare creators across different products?
Only loosely. GPM depends heavily on price point, so a creator on a $12 item and a creator on a $60 item are not directly comparable. Compare within a price band.
The decision this week
If you take one thing from this, take the sorting change. Open your short video detail, set a 3,000-play floor, sort by GPM, and write down the top two and bottom two. Then watch all four clips and name the single thing that differs.
That exercise will tell you more about your creative direction than any benchmark, because it is built on your own buyers rather than on what works on average. For the benchmark ranges that tell you whether a GPM number is good or merely normal, see what counts as a healthy GPM; use it as context, not as a target, since category and price band shift the number substantially.
And before scaling any winning clip into a batch of new placements, run it through the platform’s pre-publish checks rather than assuming a clip that converted organically will clear review everywhere. One rejection can cost you a launch day.
The workspace that keeps the overview, the detail, and the creator record in one place is the DAMI short video module; the reason to use it is not the dashboard itself, but that a decision about which creator to rebook can be logged against that creator in the same session, before the insight evaporates.
