Most SaaS founders test creative the same way they test everything else in a startup: ship one thing, watch the number, guess why it moved. That works fine for a pricing page headline. It works badly for ad hooks, because a single hook tells you almost nothing — you have no baseline to compare it against, and no way to separate "this angle doesn't work" from "this angle works but the creator was wrong for it."
Here's a structure that actually produces an answer instead of a feeling.
Step 1: Pick the metric before you write a single hook
Decide, in writing, what "winning" means before anything is filmed — demo bookings, trial signups, cost per acquisition, whatever number actually matters to your business. This sounds obvious and gets skipped constantly. Without it, "which hook won" becomes a subjective argument after the fact instead of a number everyone already agreed to before they had a favorite.
Step 2: Test the Angle, Then the Creator
An angle is the actual argument the ad makes — problem-agitation, founder story, feature walkthrough, customer voice, before/after. Swapping creators while keeping the same angle only tells you about the creator. Testing multiple genuinely different angles, ideally with more than one creator each, is what actually isolates what's working: the story, or the person telling it.
A reasonable starting spread
- Problem-led: opens on the pain point your buyer already recognizes, before your product ever appears.
- Founder story: works especially well for early-stage SaaS — buyers trust a person more than a brand at this stage.
- Feature walkthrough: shows the product doing the specific thing your buyer is trying to solve for.
- Customer voice: a real (or realistic) user explaining the outcome in their own words, unscripted.
Step 3: Run Them in Parallel
Testing one hook per week for a month feels methodical but isn't — a month of shifting seasonality, ad platform changes, and audience fatigue makes week-to-week comparisons unreliable. Running several variants at once, against the same audience and budget conditions, is the only way to compare them fairly against each other rather than against a moving target.
Step 4: Decide what "beat the current best" actually means, in advance
Before launch, agree on the threshold: does a new hook need to beat your current best by 10%? 20%? Any amount? Deciding this after you've already seen the results is how motivated reasoning creeps in — suddenly a 3% lift feels like a win because everyone's tired of testing.
Step 5: Scale the winner, cut the rest — and say so plainly
The point of testing isn't to have a portfolio of nice videos. It's to find the one or two that actually move your metric, put budget behind those, and stop pretending the others are "still worth watching." A test that doesn't end in a clear scale-or-cut decision wasn't really a test.
This is close to the exact structure Demofy runs for every Litmus Test — 5-8 hooks, one metric agreed in writing, ten days. See the offer, or submit a brief to run one on your own product.