What is CPV?

A video campaign spends A$3,000 and the platform reports 150,000 views under a two-second definition. CPV is 3,000 ÷ 150,000 = A$0.02. The same campaign reports 12,000 views under a fifteen-second definition, which is a cost per view of A$0.25 — more than twelve times higher, from the same spend and the same video. Neither number is wrong. Only one of them is about somebody who watched.

Cost per view divides what a video campaign spent by the number of views the platform counted. The division is trivial. The word "view" is not, and it is not standardised anywhere.

Across the major video surfaces a view has been defined as two seconds, three seconds, six seconds, fifteen seconds, thirty seconds, and the whole video, sometimes with a condition about how much of the player is on screen and sometimes without. Several platforms report two or three of these definitions at once, in the same table, under labels that differ by a word. A campaign can therefore report a cost per view of two cents and twenty-five cents simultaneously, both correct.

This is not a reporting quirk to be worked around. It is the first thing to establish before the figure is used for anything, because the ratio between the loosest and the strictest definition in a single campaign is routinely several times over, and it varies by creative. A video with a slow opening loses most of its audience between the two-second count and the fifteen-second one; a video that opens on the point loses far fewer. The gap between the two figures is itself the most useful signal the metric produces.

Like every cost-per-event figure, this one decomposes:

Cost per view = (CPM ÷ 1,000) ÷ view rate

The view rate is the share of impressions that met the threshold. An A$10 CPM at a 20% view rate produces a one-cent view. The same A$10 CPM at a 5% view rate produces a four-cent view. The auction did not move; the video did.

That identity settles where to look when the number changes. A cost per view that rose while CPM held flat is a creative or audience-fit problem — fewer of the people reached are choosing to watch. A cost per view that rose while the view rate held is a delivery problem: competition, a more expensive audience, a shift toward pricier placements.

It also explains why bidding harder is a weak response. Auctions rank on expected value, and a video the platform expects to be watched wins impressions more often and at a lower clearing price. Better creative lowers the cost of a view twice over.

Most in-feed video autoplays silently. A person scrolling past at speed crosses the two-second threshold without ever deciding to watch anything, and the platform counts it. This is the mechanism behind almost every implausibly cheap cost per view, and it is why in-feed, sound-off inventory reliably produces the lowest figure in an account and the weakest evidence of attention.

Placements where the person pressed play, or where the ad interrupts content they chose, produce far more expensive views and far better evidence. Skippable pre-roll sits in between and has a useful property: the person who did not skip made a decision, which is why a view counted after the skip point is worth more than several counted before it.

An account that judges video buying on cost per view alone will migrate toward the cheapest definition on the cheapest inventory, because that is what the metric rewards. Nothing about that migration is visible in the number itself.

The average price of a view compresses a distribution into a scalar and throws away the shape, and the shape is where the instructions are.

A retention curve that falls off a cliff in the first two seconds is a hook problem: the opening frame, the first spoken words, the absence of motion, a logo where a face should be. A curve that holds for five seconds and then collapses is a promise problem: the opening earned attention the rest of the video did not repay. A curve that decays gently and evenly usually means the video is simply longer than the idea in it.

Two campaigns with identical cost per view can have completely different curves and completely different instructions. Reading the scalar alone produces bid changes; reading the curve produces edits, and the edits are worth more.

Whether anybody remembered it. A view is exposure, not impact. The link between watching and behaviour is real, but it is loose enough that it cannot be inferred from a price per view.

Whether the view was chosen. An autoplayed, muted, thumb-scrolled two seconds and a deliberate fifteen seconds with sound are the same row in most reports.

Whether the audience was right. Cheap views from people who will never buy are cheaper than expensive views from people who might, and the metric prefers the former by construction.

How it compares to anything else. Across platforms the definitions differ; across campaigns inside one platform the definitions can differ too, if the objectives differ. A cross-platform cost-per-view table is one of the least defensible reports in advertising.

Start with the first three seconds, because that is where most of the loss is and it is entirely under your control. Open on motion, on a face, or on the problem. Put the point before the branding rather than after it. Assume no sound.

Then match the length to the idea. A cut-down edit of a long video usually beats the long one on cost per view, and sometimes beats it on outcomes too.

Then check the placement mix, and split the report by it. The blended figure across feed, stories, in-stream and audience networks describes none of them.

Then decide, deliberately, which definition the account is going to manage against, write it next to the target, and stop comparing figures that were counted differently.

• State the view definition next to the figure, always. A two-second view and a fifteen-second view are different events, and a report that mixes them is a report that cannot be acted on.
• Read cost per view beside the retention curve rather than on its own. Where the curve falls is a creative instruction; the average price of a view is not.
• Work the first three seconds before the bid. Cheap views come from videos people keep watching, because the auction serves a video it expects to hold attention more often and for less.
• Choose the view length that matches the message. A campaign whose point lands in two seconds should not be bought on a fifteen-second view, and a campaign that needs a demonstration should not be bought on a two-second one.
• Check what the cheap views are made of. Autoplaying, sound-off, in-feed inventory produces the lowest cost per view in almost every account and the least evidence that anybody chose to watch.