Start with the decision the test should unlock
Imagine a footwear brand commissioning a 15-second vertical film. The creative team likes two openings: a runner tying a lace and a close-up of the shoe flexing. The buyer needs to decide which opening to carry into the next edit. Write that decision at the top of the brief. “Which opening earns a longer view?” gives the team a different job from “Which edit sells more shoes?” Ask the buyer which available measure answers their question before approving the deliverables.
Mark the exact frames that can change
For this example, give each opening the first two seconds. In A, the fingers pull the lace tight. In B, the sole bends under load. At two seconds, cut both versions to the same locked sequence. Use the same caption, voice track, audio level, grade, end card, duration and export settings. With AI-generated inserts, reuse the approved master for everything after the cut. Generating the whole sequence twice can introduce a second difference as small as a changed shoe colour.
Sometimes a fair comparison needs a different brief. If the flex shot needs a sound effect to make sense, either use the same sound bed in both openings or label the question “Which opening sequence works better?” The second option tests picture and sound together. It cannot tell you which element made the difference.
Give production and media the same handover sheet
- Question: Does opening on the product demonstration improve the agreed viewing metric compared with opening on the runner?
- Control A: Lace pull, 0–2 seconds. Variant B: Sole flex, 0–2 seconds.
- Shared material: Approved master from 2–15 seconds, audio bed, captions, end card and landing page.
- Delivery: Two vertical masters, named Runner-A and Flex-B. Attach a thumbnail of each first frame so the buyer can match uploads to the brief.
- Media owner: Record the success metric, intended audience, placement and decision rule before launch. Keep the export names beside the reporting labels.
These are example instructions for a fictional brief. For a real buy, check the platform's experiment setup with the media team. Google's video experiment guide explains how to choose a success metric, set a control and isolate a variable. It also distinguishes early directional results from results at the end of the experiment.
Review the pair before it goes to trafficking
Stack the two exports in an edit timeline, align their first frames and toggle the upper track at the handoff. Check that the contact frame, captions and end card line up. Listen through the cut on headphones. If one version lands a beat later, record whether that timing is intentional. Ask someone who did not make the exports to identify A and B from the handover sheet alone.
If the buyer needs a square cut as well, deliver a square A/B pair and label the placement separately. Keep the vertical result attached to the vertical pair. That small piece of bookkeeping helps the next editor see exactly which opening, in which frame, earned another round.
For the next production step, see how to adapt a campaign film across placements or use the production brief template to agree the deliverables.
Start a project 