Summary
A Demand Gen Experiment is what the marketer should set up to compare which of the two video lengths drives a stronger CPA. It splits the audience randomly between two arms so each video gets a fair share of traffic, then reports cost per conversion for each arm with a confidence signal. A new ad group would not control for delivery bias, and Search Lift or brand surveys measure awareness rather than acquisition cost.
Nitin Batra is a Google Video Ads certified professional who plans and runs YouTube and Demand Gen campaigns, including creative A/B experiments measured on cost per acquisition.
Last updated: August 23, 2026
To compare which of two video lengths drives a stronger CPA, the marketer should set up a Demand Gen Experiment, which splits traffic between two arms and reports conversion cost differences with statistical confidence.
- A new ad group
- A Search Lift study
- A brand survey
- A Demand Gen Experiment
The correct answer is: A Demand Gen Experiment
Key Takeaways
- – A Demand Gen Experiment is the correct setup for testing which of two video lengths drives a stronger CPA.
- – The experiment splits users randomly between two arms, so the CPA difference is attributable to the creative variable and not to unequal audiences.
- – Experiments exist to remove the delivery bias you get when two creatives compete inside the same ad group, where the algorithm favours one early and starves the other.
- – Run the experiment long enough to gather meaningful conversion volume. Calling a winner in the first few days on a handful of conversions is the most common mistake, and Search Lift or brand surveys measure awareness, not CPA.
Why not just add both video lengths to the same ad group and compare?
Inside one ad group, Google Ads optimises delivery toward the ad it expects to perform better, so impressions are never split evenly. One video ends up with most of the spend within days. The CPA you see then reflects the algorithm’s choice, not a fair comparison between the two lengths.
Where do I find Demand Gen Experiments in the Google Ads interface?
Go to the Campaigns icon in the left navigation, open Experiments, then Demand Gen Experiments, and click the plus button to create one. You select the base campaign, name the experiment, choose the variable such as creative or audience, set the traffic split and the date range. Results appear in the same Experiments view with an arm by arm comparison.
Would a Search Lift study or a brand survey answer this question instead?
No. Search Lift measures the increase in organic search activity for your brand or product after ad exposure, and a brand survey measures ad recall, awareness or consideration. Both are upper funnel measurement tools. Neither reports cost per acquisition, which is what the marketer is comparing here.
How long should a Demand Gen Experiment run before I act on the result?
Run it until each arm has collected enough conversions for the confidence indicator to stabilise, which usually means at least two to four weeks depending on your conversion volume. Avoid editing budgets, bids or creatives mid flight, because that breaks the comparison. If the experiment closes without a clear winner, the honest read is that video length is not the lever moving your CPA.