How Brazilian brands test UGC creatives: hooks, variants, what to measure and how to scale the winners.
Isolating one variable at a time
Seven rules that make a test readable.
Change the hook, and nothing else, if the hook is what you are testing.
Keep the same creator from one variant to the next, unless the creator is precisely the variable.
Keep the same duration, because a shorter video is favoured on indicators that have nothing to do with it.
Use the same call to action, worded in exactly the same way.
Send the variants to the same landing page.
Name the files after the variable tested, which avoids confusion at reading time.
And write the hypothesis before launch, in one sentence, failing which any result will seem to confirm something.
Building and ordering your UGC variations
Seven variables to test, in the order in which they pay off most.
The hook of the first three seconds, which decides everything else.
The objection handled, because two audiences do not have the same barrier to purchase.
The opening format: face, product in hand, or result shown first.
The setting, an ordinary kitchen not speaking to the same audience as a polished set.
The creator's accent and region, a specifically Brazilian variable and often a decisive one.
The presence or absence of burnt-in subtitles, which changes everything on public transport.
And length, tested last, once the content works.
How many variants to launch, and with what budget
Seven markers for sizing a test that will actually answer.
Three to four variants per test, never ten, otherwise none receives enough budget.
An identical daily budget for each variant, decided before launch and not adjusted along the way.
A minimum duration of seven days, for the reason explained above.
A single conversion goal, the same for every variant.
An audience wide enough that delivery does not run into too narrow a public.
One authorised change during the test: stopping a clearly failing variant.
And a production budget planned for what follows, because a successful test immediately creates the need for three new videos.
The metrics that really count
Seven indicators, from the most reliable to the most misleading.
Retention at three seconds, which judges the hook alone.
Retention at fifteen seconds, which judges the body of the video.
Cost per purchase, the only figure that speaks to the accountant.
Click rate, useful but easy to inflate with an exaggerated promise.
The number of purchases per variant, which has to reach a minimum volume before being read.
Frequency, which explains the fatigue of a creative that was working.
And comments, to be read but never counted as a result.
Reading results without fooling yourself
Six safeguards against false conclusions.
Wait for a sufficient number of purchases, not an impressive number of impressions.
Do not declare a variant the winner on a two per cent gap.
Check that both variants received a comparable budget.
Be wary of an isolated spike, which often comes from an unrepresentative audience.
Rerun the test if one variant suffered an ad delivery problem.
And accept the result that contradicts your intuition, otherwise the test served no purpose.
When a test is pointless
Six situations where a test costs more than it returns.
When the daily budget is too small for each variant to reach a readable volume.
When the landing page is the real problem, and no video will fix it.
When the product's price has just changed, which disturbs the comparison.
When a commercial promotion is running at the same time and skews every reference point.
When the two variants in fact say the same thing in different words.
And when nobody on the team has set aside time to read the results the following week.
What is tested, one variable at a time
| What you change | What you learn | Test duration |
|---|---|---|
| The first three seconds | Whether the hook holds | 48 h |
| The face on screen | Whether the person fits the audience | 7 days |
| The close | Whether the call to action works | 7 days |
What invalidates a test
Changing two things at once, stopping before the learning phase ends, or comparing two different periods. Each of these produces a result you cannot interpret.
What is kept when one version wins
The opening, almost always. The rest can be recut, shortened or lengthened, but the hook that worked becomes the base of the next three tests.
Scaling winners into more UGC
Seven steps for turning a winner into a series.
Identify what won: the hook, the creator, the objection, rarely all three.
Order three new videos that keep that element and change the rest.
Buy the advertising rights of the winner for a longer period.
Adapt the same hook with creators from other regions, to widen without starting over.
Watch frequency, and prepare the successor before fatigue sets in.
Reinvest the budget in steps, not all at once, which destabilises delivery.
And keep a log of tests, because the same idea will be proposed again in six months by somebody who does not know it has already lost.
One last marker: in Brazil the value of a UGC programme does not come from the perfect video, it comes from the rhythm of production. A brand testing four creatives a month learns four times faster than a brand producing one very polished one, and it ends up finding the hook nobody had imagined at briefing time.
