Influencer & affiliate agencies, reviewed and ranked
Creator partnerships and affiliate programmes bought for reach and sales. Scored on fraud control and incremental revenue, not follower counts.
21 evaluated, of which we would put 20 in front of a client and cut 1.
Audience fraud is the baseline risk
Fake followers and engagement pods are endemic, and the cost lands entirely on the buyer. The joint-heaviest dimension here is whether the agency runs real audience-quality vetting rather than working from a rolodex. Ask what they check, what tooling they use, and how many creators they rejected on the last campaign — a shop that rejects nobody is not screening.
Reach and EMV are the industry’s comfort metrics
Earned media value is a number the industry invented to make impressions sound like money. The measures that matter are incremental sales, CAC and affiliate revenue. Ask how they would know whether a campaign caused purchases rather than coincided with them.
Paid amplification is where this stopped being organic
The performance unlock is running the best creator content as paid ads through Spark Ads, Partnership Ads or whitelisting — so a winner scales past whoever happened to see it. If you are buying creator content primarily to run as ads rather than for the creator's audience, that is a different purchase: see UGC & creator content.
Cut: reach-first influencer shop with severely variable results, one client got 7 total clicks from 8M reach across 5 influencers and was ghosted on a refund request, plus a reported creator non-payment. Vanity-reach model with client- and creator-handling red flags.
Reach-vanity with terrible conversion (7 clicks / 8M reach)
Ghosted a refund request
Reported creator non-payment
How this page is built: every vendor here is scored against the published influencer_affiliate rubric — the one written for what this category actually sells, not a general agency rubric. Several of these records were first filed elsewhere and scored against a rubric they were not competing in; where that happened, the record carries a rescored_from field showing the previous verdict and score, so the change is auditable rather than silent.