Video Monetisation: Where Publisher Revenue Actually Comes From
Video revenue is a function of impressions, fill and rate. Most teams push on the one with the least headroom.
Publisher video revenue decomposes into three things: how many video impressions you generate, what share of them get filled with a paying ad, and what those ads pay.
Almost every team pushes hardest on the third — negotiating rates, adding demand partners, chasing header bidding gains. That is the term with the least headroom, because it is set largely by the market rather than by you.
Where the leverage actually is
The first term is usually the one with room in it. Most sites have far more video than they surface, and a large share of visitors never start a video at all. Moving the share of visitors who begin watching moves impressions, and impressions multiply through the whole calculation.
This is unglamorous work — better previews, better placement, surfacing relevant assets instead of default ones — but it compounds in a way rate negotiation does not.
The measurement trap
Counting video starts rewards behaviour that damages the thing you are actually selling. Autoplay with sound off, interstitials and forced pre-roll all inflate starts while making the experience worse.
The result is a metric that goes up while the underlying asset — a reader willing to watch video on your site — degrades. Watch time per visitor and repeat video sessions are harder numbers to move and much harder to fake.
Ad load and the long run
Ad load has a short-run and a long-run answer that point in opposite directions. Raising it lifts revenue this quarter and reduces the number of people willing to start a video next quarter.
There is no formula for the right level, but there is a useful test: if your video revenue per visitor is rising while video starts per visitor are falling, you are extracting rather than growing. For streaming specifically, the format question is different again — see rethinking ad formats for OTT and news.
The prerequisite
Most of the impression-side gains depend on knowing what is in your library well enough to surface the right thing. That is the case for analysis over metadata, and for making the archive usable.