← Back to all lessons Lesson 03 of 09

Create & test · Lesson 03

Build a test you can actually learn from.

You made a batch of creator videos and one of them won. Now try to say why. If the creator, the angle, the hook, the format, and the CTA all changed at once, you can’t.

Change fewer things. The winner will tell you what to repeat.

~5 min · leave with a test matrix and a copyable test plan

TL;DR

Don’t test five things at once.

A useful UGC test changes one meaningful variable and keeps the rest reasonably steady. Not laboratory steady. Steady enough that the result means something.

01

Pick the variable

What are you actually trying to learn?

02

Keep the rest stable

Same creator, angle, and CTA if you are testing hooks.

03

Build the matrix

Decide exactly which variations run, on one screen.

04

Name the variation

So future you can find the result.

What to check in Airbridge

  • Installs by variation: did one version bring more users into the app?
  • Trial starts by variation: did those users show real intent?
  • Paid subscribers by variation: which version actually converted?

Hook-level numbers (watch rate, click-through) live in the ad platform. Airbridge picks up from the install.

01 Pick the variable

Start with one thing you want to learn.

“Which video wins?” finds a winner. “Which hook wins?” finds a pattern.

A lot of “UGC testing” is publishing several different videos and seeing which one gets the biggest number. That can find a winner. It does not tell you why it won.

Imagine you test:

  • Creator A + problem-first hook + talking head + free-trial CTA
  • Creator B + testimonial hook + screen recording + “download now”
  • Creator C + humor angle + voiceover + discount CTA

Creator B wins. What did you learn? Was it the creator, the angle, the hook, the format, or the CTA? You changed almost everything, so the answer is “one of those.”

Ask one question first

Hook test

Which opening gets more people to keep watching?

Angle test

Which customer problem creates more trials?

Creator test

Which creator brings better users?

Format test

Does talking-head beat screen recording?

Offer test

Does the free-trial message produce more trial starts?

That question becomes the test. Everything else should support it.

02 Keep the rest stable

If you are testing the hook, don’t secretly test five other things too.

Remove the differences that would hide the answer, not every difference.

UGC is messy by nature. A creator will phrase things differently take to take, and that is fine. The variable you care about should still be the obvious difference between the videos.

Testing hooksTesting creators
Keepsame angle, same creator, same product proof, same CTA, roughly the same formatsame angle, same basic brief, same CTA, similar distribution
ChangeHook A / Hook B / Hook CMaya / Jake / Sarah
You learnwhich opening works for this angle with this creatorwhich creator brings the better subscriber, not the prettier video

Good test: same creator + same angle + 3 hooks.

Hard to learn from: 3 creators + 3 angles + 3 formats + 3 CTAs.

The second setup might produce a winner. It just won’t tell you what to repeat.

Keep the parts that are not the question as consistent as reasonably possible.

03 Build the matrix

Put the test on one screen before you make anything.

If it does not fit on one screen, the test has too many variables.

ABC
CreatorMayaMayaMaya
AngleGym confusionGym confusionGym confusion
FormatTalking headTalking headTalking head
Hook“I finally got to the gym…”“The hardest part isn’t working out…”“POV: you made it to the gym and…”
CTAStart free trialStart free trialStart free trial

What this test can tell you: which hook performs best for this angle with this creator.

What it cannot tell you: whether Maya is your best creator overall.

That is a different test. Lesson 04 gives each creator their own row, and Lesson 06 shows how to read it.

Pick the metric that decides it

One number gets the final vote (Lesson 01). Which number depends on where the variable shows up.

In the ad platform · Watch → Click

Hook and format tests usually decide here: watch rate, completion, click-through. Airbridge does not measure these, and this playbook won’t pretend it does.

In Airbridge · Install → Trial → Paid → Revenue

Angle, creator, and offer tests decide here: installs, trial starts, paid subscribers, revenue per variation.

Use the ad platform to judge the creative signal. Use Airbridge to check whether that signal turned into valuable users.

04 Name the variation

Future you should know what “video_final_7” was.

The angle and the creator are already named. This names the variation.

Do not wait for results to organize the test. Each variation gets a short, readable name that goes into the link (Lesson 04 shows exactly where).

The angle was named in Lesson 02 (angle-gym-confusion). The creator is named when you build the link (creator-maya). So the only thing this lesson names is the variation itself:

hook-a hook-b hook-c

Or, for a format test:

hook-b_talking-head hook-b_screen-rec

Keep it to the variation. Not gym-confusion_maya_hook-b. The angle and the creator are separate labels with their own columns. If you put them in the video name too, you get one long string that cannot be compared on any axis.

Build your test

Structure the test, not the creative. Pick what you are testing and the plan fills in.

01What are you testing?

Your test plan

Pick what you’re testing to start the plan.

Before launch

  • What are we testing?
  • What stays the same?
  • What changes?
  • What metric decides it, and where does that metric live?
  • What is each variation called?

If one of those is unclear, fix it before anything is filmed.

Turn a winning video into a repeatable learning.

Airbridge compares the variations you tested by what happened after the click, not just by the metric that looked good in the ad platform.

Hook A / B / C→Install→Trial→Paid→Revenue

One variable, one metric, one name per variation. Then the winner teaches you something.

Start measuring in Airbridge →