From the brief to deciding what to reshoot: how a video neurotest works and how many people it needs.
A video can please the whole team and still lose viewers at the third second. A survey after the screening won’t show this: people barely remember where exactly they got bored and often answer the way they think is right. A neurotest records the reaction while people watch, second by second and scene by scene. So the weak spot is visible before the ad airs, while the scene can still be moved, cut or reshot.
Below are the five steps of an ad test in our lab in Samarkand.
1. Decide what you are testing and what to compare it with
Start not with the video but with the decision you will make based on the test. For example:
- which of two versions to launch;
- whether the ending needs a reshoot;
- whether viewers notice the product and the price;
- whether the brand is remembered, not just the story.
For this decision you pick control material: the brand’s previous video or a strong competitor video. A score of “70 out of 100” means nothing on its own; what matters is whether it is better or worse than what you compare it with.
2. Recruit participants
Participants are your target audience: gender, age, city, whether they buy your category. To compare two videos, 30–40 people are usually enough. If you need to compare segments, for example younger and older buyers, the sample grows, usually from 60 people. We calculate the exact number from the brief: there is a sample size calculator on the Lab page.
3. Record the reaction second by second
During the screening several signals are recorded at once, so the reaction is visible for every scene:
- EEG: whether the scene holds attention, interests people and triggers memory;
- eye tracker: where the viewer looks, whether they noticed the product, logo and price;
- skin response and heart rate: how strong the emotion is;
- facial expressions: whether the scene is liked or annoying.
A single signal is easy to misread: a strong emotion can be interest or irritation. That is why conclusions are drawn only from a combination of signals.
4. Check what people remember
After 2–3 days participants answer questions from memory: which brands and shots they remember. Buyers recall an ad at the shelf or in their feed, not right after watching it, so a delayed survey is more honest than asking “did you like it?” right after the screening.
5. Compare and decide what to fix
The video’s scores are compared with the control material. The report has five clear business scores (attention, emotion, attitude, memory and purchase intent) and a list of fixes by scene. For example: move the logo closer to the emotional peak, show the product earlier, cut the scene where attention drops.
A sample of such a report on demo data is on the Lab page.
If there is no time for the lab
For quick decisions there is AI creative analysis: in 1–2 days a model predicts where the eye will land, what emotion a shot will evoke and how memorable it is, with no participants. It is a good filter for weeding out weak versions. For expensive launches, confirm the forecast with a video neurotest.
In brief
A test before launch costs less than a reshoot after the ad airs. The earlier you test the video (at the storyboard, animatic or rough-cut stage), the cheaper the fixes.
FAQ
How many people does an ad test need?
To compare two videos, usually 30–40 participants from the target audience; to compare segments, from 60. We calculate the exact number from the brief.
Can a video be tested before it is shot?
Yes: a storyboard, an animatic or a rough cut. Fixing a scene at this stage costs less than reshooting a finished video.
How long does the test take?
An express test takes about a week, a study of several videos 2–3 weeks. AI creative analysis without participants takes 1–2 days.