Video measurement and optimization
What this domain tests
This domain tests whether you can read video metrics correctly, distinguish platform metrics from business outcomes, and optimize without misattributing results. Expect questions on view definitions, view-through conversions, brand lift, asset reporting, and disciplined testing.
Metrics that matter
- Impressions count ad displays; views count 30 seconds of watch or a full view for shorter ads, plus interactions in some formats.
- View rate = views ÷ impressions; CPV = cost ÷ views; CPM = cost ÷ impressions × 1,000.
- Conversions and conversion value are business outcomes; view-through conversions are reported separately because the user saw the ad but did not click.
- Brand lift and search lift measure upper-funnel effects that clicks and conversions miss.
- YouTube Analytics adds watch time, audience retention, and traffic-source context.
Optimization discipline
Optimize at the level where the problem lives: creative asset, audience segment, placement, or bid. Refresh fatigued creative, exclude irrelevant placements, shift budget toward converting audiences, and use experiments or ad sequences for structural questions. Change one major variable at a time so the readout is interpretable.
Worked example
A Video action campaign delivers 2,000,000 impressions, 600,000 views, and $12,000 spend. View rate is 600,000 ÷ 2,000,000 = 30%; CPV is $12,000 ÷ 600,000 = $0.02. It reports 240 conversions, so CPA is $50, above a $40 target. It also reports 120 view-through conversions separately. Do not add the 120 to the 240 to declare a $33 CPA; that changes the attribution definition. Instead, inspect asset and audience splits, then fix the weak creative or audience rather than the bid first.
Traps that cost marks
- Treating every view as a 30-second view regardless of ad length.
- Ignoring view-through conversions or brand lift when judging upper-funnel video.
- Optimizing only on last-click conversions and missing assisted impact.
- Changing creative, audience, and bid simultaneously and calling it a test.
- Reporting view rate as views ÷ spend.