Reach and engagement tell you whether an audience is real. Neither tells you whether the creator will deliver. The single most predictive signal for a second campaign is what happened on the first one — did they file on time, did the asset pass review, how many revision rounds did it take — and most brands never record it. Creator vetting has spent a decade getting better at measuring audiences and no better at measuring counterparties.
Why did follower count stop working as a vetting benchmark?
Because the number moved underneath everyone at once. In May 2026 Instagram removed a large volume of bot and inactive accounts, and creator follower counts fell overnight — in some cases by millions. We covered what that did to vetting in the Great Purge and what brands should do next, and the practical conclusion has not changed: a metric that can be revised downward by a platform cleanup was never measuring what brands thought it was measuring.
Most vetting frameworks adapted by moving one step sideways — from follower count to engagement rate, audience overlap, authenticity scoring. That was the right move and an incomplete one. All three still describe the audience. None of them describes the creator.
What can’t audience-quality vetting tell you?
Whether the person will do the work.
An engagement rate is silent on whether a creator answers email, hits a posting window, reads the brief, or hands over raw files without three reminders. Those are the variables that decide whether a campaign lands on the date you promised your CMO — and they are invisible in every audience metric on the market. A creator can have an impeccable audience profile and still be the reason your launch slipped a fortnight.
This is a measurement gap, not a people problem. Brands ask for audience data because audience data is what platforms expose. Delivery behaviour only exists in the record of campaigns you have already run, which is exactly the data most programmes never capture in a form anyone can query later.
What does creator reliability actually mean?
It resolves into four things you can observe, all of which come out of your own campaign history rather than a platform API:
| Signal | What it tells you | Where it comes from |
|---|---|---|
| On-time delivery | Whether your campaign calendar will hold | Past posting windows vs. actual post dates |
| First-pass approval rate | The review cost you will actually absorb | Content review history |
| Revision rounds per asset | The hidden cost of a “cheap” creator | Content review history |
| Repeat participation | Whether they want the relationship or the fee | Roster history across campaigns |
None of these require new instrumentation. They require that someone kept the record — and that the record is attached to the creator rather than buried in a closed campaign folder.
What does an unreliable creator actually cost you?
Not their fee. The fee is the smallest number in the equation, and it is the only one most programmes track.
The real costs are structural. A missed posting window inside a launch window means paid support running against content that is not live yet. Four rounds of revisions on one asset consumes reviewer attention that was budgeted across a roster. A creator who goes quiet mid-campaign forces a replacement at short notice, at whatever rate is available rather than the rate you negotiated. And every one of those absorbs the scarcest resource on a creator team, which is attention — the hours spent chasing are hours not spent briefing the next campaign.
We have deliberately not put a figure on this, because we have not found a defensible one. Anyone quoting you a precise cost-of-delay number for creator programmes is estimating. The argument does not need the number: if your team can name the creators who cost them a week last quarter, you already have the evidence, and it is already unrecorded.
How do you build a reliability signal you don’t have yet?
Start logging three fields at the close of every campaign, per creator. Committed date versus actual post date. Number of review rounds before approval. Whether they completed every contracted deliverable. That is enough to make your next selection round meaningfully better than this one, and it costs one row in a spreadsheet per creator per campaign.
Two rules make the data usable rather than merely present. Record it at campaign close, while it is fresh and uncontested — retrospective judgement drifts toward whoever was most recently annoying. And attach it to the creator, not the campaign, because the entire value of the signal is that it is there the next time that name comes up.
If you already have a content library and no such record, an audit of the assets you own will recover part of it: approval dates, version counts, and who delivered what are usually reconstructible from the files themselves.
Why does your own history beat any market-wide score?
Because reliability is partly a property of the brief. A creator who missed two deadlines on a complex multi-deliverable campaign may be entirely dependable on a single-asset one. One who needed four revision rounds under a rigid brand brief may pass first time given creative latitude. A market-wide average flattens all of that into a number that is true on average and wrong for you specifically.
This is the same argument that governs fit over reach in creator selection: the useful question is never “is this creator good?” but “is this creator good for this brief, for this brand?” First-party history answers that. An aggregate score cannot.
It is also why reliability belongs next to performance rather than instead of it. A creator who delivers flawlessly and converts nobody is not a good partner, and the reverse is equally true. Read delivery history alongside the performance metrics that actually matter, and against a clear-eyed view of why influencer ROI so often looks worse than it is.
In Social Native, creator profiles now show whether a creator delivered on their past campaigns alongside their audience metrics, so selection can weigh both at the point of decision rather than after the fact.
What to do before your next campaign
- Ask your team who they would re-hire. They know. That knowledge is currently held in individual heads and leaves when they do — write it down this week.
- Add three fields to your campaign close. Committed vs. actual date, review rounds, deliverables completed. Per creator, every time.
- Put delivery history next to audience data in selection. Not instead of it. A roster chosen on both is the only one whose timeline you can actually commit to.
Follower counts got revised overnight and engagement rates can be bought. What cannot be faked is a record of having done the work, on time, twice. That signal is sitting in your own campaign history right now, and the only reason it is not helping you is that nobody wrote it down.
