Hypothesis
If the same video runs as a short edit and a longer edit with an identical opening, the short edit will deliver the campaign's outcome at a lower cost on feed and short-form placements, because most viewers leave before the longer edit's later scenes. The decision behind it is concrete: run a short edit, a long edit, or both of the same video ad.
The hypothesis can fail in an instructive way. A longer edit has room for a demonstration, a story or proof that a short cut drops, and the viewers who stay may be the ones most likely to act. So the long edit can lose on views and win on outcome, especially where attention is bought deliberately, such as in-stream placements where viewers chose to watch. The YouTube ad formats guide and the Meta ad formats and placements guide describe how each format and placement is bought and counted; this test asks which length earns the outcome more cheaply in the placements the campaign actually uses.
Metrics and guardrails
The primary metric is cost per campaign outcome. For a conversion campaign, that is cost per conversion, which the CPA calculator computes on the same basis for both edits. For an awareness campaign measured by a lift study, it is cost per lifted respondent: spend divided by the number of people the edit moved on recall or consideration. Both measures count what the budget is for, not what the viewer did on the way.
Three guardrails explain the result. View RateShare of video ad impressions that became counted views. shows how many people watched as far as the platform counts a view; it will usually favour the short edit and is read for context, not as the verdict. Average View DurationTotal watch time of a video divided by its number of views. shows how far into each edit viewers stay, and comparing it with the long edit's length shows how many people ever saw its later scenes. Click-through rate shows whether either edit sends more people onward. Where the campaign buys views, CPVCost of each video view. belongs alongside, with the same warning: a cheaper view is not a cheaper customer.
Design
Use the platform's experiment or A/B test tool, with edit length as the single variable. Cut both edits from the same footage with the same opening seconds, the same brand moment and the same call to action; the long edit adds scenes after the opening, the short edit leaves them out. Run both with the same audience, placements, budget, bid strategy and flight dates.
| Design element | Choice | Reason |
|---|---|---|
| Unit of randomisation | User, assigned by the platform's experiment or A/B test tool | Each edit reaches its own randomised audience |
| Arms | Short edit versus long edit of the same footage | One change: length |
| Opening | Identical opening seconds in both | Keeps the hook out of the comparison |
| Placements | The same in both arms, recorded | Placements cut and skip differently |
| Primary metric | Cost per campaign outcome | Judges on what the budget is for |
| Fatigue | Stop before either edit wears out | Fatigue in one arm decides the test |
Placements need care. Feeds, short-form placements and in-stream inventory treat length differently: some are skippable and some are not, and some suit vertical video, as the platforms' format and placement documentation describes. If the campaign runs across several, either test within one placement group or read the results by placement and treat each as its own question.
Sample size reasoning
Size the test on the outcome that decides the budget, not on views, which are plentiful and misleading. Views reach a large sample quickly and will show a confident difference that may have nothing to do with outcome. The sample size and significance guide shows why a rarer outcome needs a longer run, and a lift study needs enough respondents per edit to separate a difference between two versions of the same ad, which is usually smaller than the lift of the ad itself.
If the outcome is too rare to separate the edits in a tolerable period, read a nearer outcome such as clicks to a key page, test the edits on a larger campaign, or accept that only a large difference can be seen.
Duration
Run both edits through the learning phase and then at least two full weekly cycles of stable delivery. Stop before Creative FatigueThe decline in an ad's response as the same people see it again and again. sets in on either edit: if frequency climbs and response falls in one arm, the test starts measuring wear-out rather than length. A short edit may tire faster because it is seen in full more often, which is a finding to record for planning, not a reason to keep the test running.
Reading the results and the decision rule
Read the test once, at the planned end. Compare cost per outcome between the edits with an interval, then read view rate, average view duration and click-through rate to explain the difference. Write the decision rule into the plan before launch:
| Result | Decision |
|---|---|
| Short edit has the lower cost per outcome | Keep the short edit; use the long one only where a test shows it earns its place |
| Long edit has the lower cost per outcome | Keep the long edit; its later scenes are doing work |
| Long edit wins on lift but loses on cost per outcome | Use the long edit where attention is bought deliberately and the short edit in feeds |
| No reliable difference | Keep the edit that is cheaper to produce and refresh, usually the short one |
The retention curves guide shows how to read where viewers leave each edit, which turns the result into editing decisions for the next video. If both edits struggle to earn views at all, the troubleshooting guide for low YouTube view rates covers causes that sit outside the edit.
A running shoe brand tests two edits of a launch video
Brennock Running, an invented running shoe brand, cuts a launch video into a short edit and a longer edit that adds a fit demonstration, both with the same opening. It runs an A/B test on Meta feeds and short-form placements at £100 a day per edit for four weeks, optimising for purchase. The short edit records a view rate well above the long edit's and 96 purchases at £29 each; the long edit records 88 purchases at £32 each. The interval on the difference crosses zero, and click-through rates are close. Brennock keeps the short edit in feeds, where it is cheaper to refresh, and plans an in-stream test for the long edit.
Pitfalls
Different openings are the pitfall that most often turns a length test into a hook test. Editors naturally tighten the opening of a short cut; if they do, the comparison is no longer about length. Lock the opening seconds before editing the rest. Placements that cut or skip differently are the second: if the long edit is mostly served where viewers skip, and the short edit where they do not, the placement decides the test. Run both in the same placements and read the breakdown.
Judging on views is the third. A short edit will nearly always win on view rate and completion, and a team that reads those first may never look at the outcome. Set the report's columns to cost per outcome before launch.
Common mistakes
Length tests fail on their edit and their metric more often than on their statistics. The ad creative hook test changes only the opening and is the natural partner to this one, and the experiment library lists every design.
- Changing the opening with the length. The test then measures the hook.
- Running both edits in one ad set. Delivery picks a winner before the test can.
- Judging on view rate or completion. Short edits win those by construction.
- Mixing placements between arms. Skippable and non-skippable inventory behave differently.
- Letting one edit fatigue. Wear-out then decides the result.
- Applying a feed result to in-stream. Read each placement group as its own question.
