Working with a UGC agency without losing the testing signal

A UGC agency solves a supply problem, and solves it well. The failure is downstream of the supplier. Thirty assets arrive that differ in hook, format, length, creator and offer all at once, so whichever one wins teaches you nothing you can brief against next month. Structuring the batch is the buyer's job.

What does a UGC agency actually solve?

A real shortage. Creator sourcing, contracting, briefing, usage rights and turnaround are operational work, and most ecommerce teams are not staffed for it at the volume paid social consumes. A supplier that can put thirty pieces of native-looking footage in your hands inside a few weeks is doing something genuinely difficult, faster and more cheaply than you could hire for it. That is the service, and it is worth paying for. What the supplier is not usually asked to do, and not usually paid to do, is decide what any of those thirty pieces are meant to prove. The scope is quantity, format and turnaround, because that is what the buyer specified. So when somebody says a UGC agency did not work, what generally happened is that a supply contract was expected to produce a learning programme, and nobody wrote one. That is a briefing failure rather than a supplier failure, and it is fixable on the buyer's side inside one cycle.

How do you brief for signal instead of volume?

By writing the hypothesis before the quantity. A brief asking for thirty assets across mixed creators and mixed formats gets exactly that. A brief that says we believe first-time buyers hesitate on fit, so we want six variants that open with a fit demonstration in the first two seconds and are otherwise identical in length, format and offer, gets you something readable. Ask for variants that isolate one variable, and be explicit about what is held constant, because a supplier cannot hold constant what nobody named. Agree in advance what counts as a win and at what spend, so the review meeting is arithmetic rather than taste. Quantity still matters, since a test needs enough variants to separate a real difference from noise, but quantity chosen after the question is settled is a different thing from quantity as the brief. None of this is extra work for the supplier, and it is usually less, since ambiguity that would otherwise be resolved with guesses on the shoot day has been resolved in the brief. That part of creative testing and performance creative has to sit on the buyer's side, because only the buyer sees the account the results land in.

What has to exist before you order the next batch?

A documented library of what has already been tested. Not a folder of files, a record: what each asset was trying to prove, which variable it isolated, what it spent, what it returned and what the team concluded. Without one, batch four repeats hooks that batch two already ruled out, usually because whoever ran batch two has moved on or forgotten. With one, the next brief starts from a shorter list of open questions, which is how creative testing gets cheaper over time instead of steadily more expensive. Keep it somewhere the supplier can read as well. A creator briefed with the three angles that have already failed and the one currently working will not spend a shoot day rediscovering them, and suppliers are generally glad to have it, because thin briefs are a problem on their side too.

With a womens fashion brand, sales rose 99% while efficiency improved rather than degraded: new customer cost per acquisition down 21%, new-customer ROAS up 58%, and net profit up 136%. That sits on a net margin still close to 3%, which we would rather state than leave out.

Before you scale UGC volume, a free thirty minute audit from Plaid Testing will show whether the testing structure can read the results. Tracking, account structure, three fixes.