Ratings are only worth reading if the people leaving them had to do something costly first.
That's the whole principle. Almost every problem with online reviews — fake positives, competitor sabotage, review farms, the friend who leaves five stars — comes from the same root: leaving a review is nearly free, so the ratings measure willingness to type rather than experience of the work.
What "booking-gated" means
A rating can only be left by a brand that actually booked the creator through the platform, on a booking that reached the delivered state, and only once per booking.
Four constraints, each closing a specific hole:
You must have booked them. No rating without a real engagement recorded on the platform.
It must have been delivered. A booking that never completed can't be rated. That closes the abandoned-project revenge review.
One rating per booking. No stacking multiple ratings from a single job, in either direction.
Only the brand that owns the booking can rate it. Not a colleague, not a competitor, not an account that happens to know the creator's name.
Together those mean a rating represents a real project between two identified parties that ran to completion. That's a much higher bar than "someone typed five stars," and it's the reason the number carries information.
What it costs to fake
The useful way to evaluate any reputation system is to ask what a fake entry costs.
On an open review site: an email address and two minutes.
Here: you'd have to create a brand account, post a brief, book the creator, run the booking through to delivered, and only then leave one rating. That's not impossible — no system is — but it's expensive and leaves a trail, which is most of what you can ask of a reputation system.
Why the average isn't the score
A second problem sits behind gating: a raw average is a bad ranking, because a single 5.0 from one job outranks a 4.8 from forty.
Acumin ranks on a Bayesian mean rather than a raw average. In plain terms, every creator starts blended towards a global prior, and their own ratings pull them away from it as evidence accumulates. Concretely, the prior is a 4.3 rating carrying the weight of five ratings, and the result is scaled to a 0–100 score.
The consequence is what you'd want: one perfect rating moves you a little, forty consistent ratings move you a lot. Consistency at volume beats a lucky single review, which is exactly the property a raw average lacks.
What gets rated
Ratings carry an overall figure plus three dimensions — quality, communication and timeliness — and a would-rebook flag.
The dimensional split matters when you're hiring. A creator rated highly on quality and moderately on timeliness is telling you something specific and useful: excellent work, plan slack into the schedule. An overall number alone would hide that, and it's frequently the thing that determines whether a project goes well.
Would-rebook is the bluntest and often the most honest signal on the card. It's harder to be politely generous about than a star rating.
Badges are derived, not awarded
A small set of badges come from the rating history rather than from anyone's discretion: a first five-star rating, ten projects delivered, an on-time badge that requires at least five timeliness ratings averaging 4.5 or better, and a repeat-booked badge for being hired twice by the same brand.
That last one is worth weighting heavily when you're reading a profile. A brand rehiring someone is the least gameable signal in the system — it costs real money and it's a decision made with full information.
What booking-gating still can't fix
Being honest about the limits, because a reputation system that oversells itself is worse than one that doesn't exist.
Small samples are still small. A gated 5.0 from two jobs is two people's opinion. The Bayesian score accounts for this in the ranking, but if you're reading an individual profile, check the count.
Ratings skew positive. People who've just paid for something and received it are inclined to be generous, and both parties know they may work together again. That's true here as everywhere.
It only covers work done on the platform. A creator with fifteen years of excellent off-platform work starts with no ratings. Absence of a score is not evidence of anything — treat an unrated profile on its portfolio and its approach, the way you would have anyway.
It measures the engagement, not the footage. A rating tells you how the job went. Whether the work is right for you is a separate judgement, and the reel and the approach are still where you make it.
Speciality matters. Someone excellent at documentary may be unremarkable at product. Scores can be read per speciality, and a bucket needs at least three ratings before it's ranked on its own — below that you're reading their general standing.
Where Acumin fits
The mechanics above are how the Creator Network works, not a description of an ideal. Ratings are gated on delivered bookings you own, one per booking; the score is the Bayesian mean described above; badges derive from the history.
There's a second half worth knowing about: when a brand leaves a rating, it mints a shareable referral card — a brand-branded credential the creator can send to anyone, showing the stars, the dimensions, the note and the would-rebook flag. Where the brand linked the live post, the card can also carry the content's real-world performance, labelled as measured rather than claimed.
That's the point of the whole system from the creator's side: proof that travels off-platform without them having to ask you for a testimonial.
How to use this today
Next time you're reading any creative-industry rating — here or anywhere — check two numbers before the average: how many ratings, and what the reviewer had to do to leave one.
If a profile has a high average from three ratings, you're reading three people. That's worth something. It isn't worth what the number implies.
Related: Understanding your Acumin Score is the creator-side view of the same mechanism. How to read a pitch is the judgement a score doesn't replace.