What is eCPM?

A campaign bought on clicks spends A$3,200 and delivers 640,000 impressions and 6,400 clicks. The cost per click is A$0.50 and the effective CPM is (3,200 ÷ 640,000) × 1,000 = A$5.00. A second campaign bought on impressions at a A$7.00 CPM delivers the same 640,000 impressions for A$4,480. The first buy was cheaper for the same exposure, which is a comparison neither campaign's own pricing model could have made.

Effective cost per thousand impressions is derived rather than agreed. Take what a campaign actually spent, divide by the impressions it actually delivered, multiply by a thousand.

eCPM = (total spend ÷ impressions) × 1,000

The reason the metric exists is that media is sold in several incompatible units. One campaign is bought per thousand impressions, another per click, another per view, another per install or per conversion. Those prices cannot be compared with each other because they are not denominated in the same thing.

Impressions are the one unit every campaign has, because every campaign has to be shown to somebody before anything else can happen. Converting every buy into a price per thousand impressions puts them all on one axis, which is what makes a mixed portfolio legible.

On the supply side, a publisher has a finite number of impressions and receives bids priced in different ways. A supply platform's job is to decide which bid to accept for a given impression, which requires converting each of them into an expected price per thousand impressions. That is exactly this calculation, run forward on an expectation rather than backward on a result.

On the demand side, an advertiser has the mirror-image problem. A campaign bought on clicks and a campaign bought on impressions both consume inventory, and the advertiser wants to know which one paid more for exposure. Again, the same calculation.

The symmetry is why the term appears in both a publisher's revenue report and a buyer's media report meaning slightly different things: one is the price received, the other is the price paid, and the difference between them is the fee stack.

Comparing suppliers. Two partners quoting in different units become comparable. Including fees, ad serving and data costs in the numerator turns it into a comparison of what the buyer actually paid, which is frequently a different ranking from the one the quoted rates suggested.

Diagnosing delivery shifts. Inside a single performance campaign, a falling effective price with flat or falling outcomes is a reliable signal that delivery has moved toward cheaper inventory — a different placement, a broader audience, a lower-attention format. The campaign's own headline metric may not move at all while this happens.

Sanity-checking an auction. Sharp movement in the effective price with no change in targeting or creative usually means the auction changed: a seasonal demand spike, a competitor arriving, or an inventory supply change.

Valuing inventory. For anyone deciding what a placement is worth, this is the price the market is paying for it.

It is a cost metric and it contains no information about value. Two campaigns at the same effective price can produce completely different results, and a cheaper effective price is frequently the symptom of worse inventory rather than a better deal.

It also averages away everything that matters about where the impressions were. A blended figure across feed, stories, in-stream video and an audience network is a weighted average of four different markets, and the weighting is decided by the delivery algorithm rather than by the buyer. Splitting the figure by placement is the first step in any serious use of it.

And it says nothing about whether the impressions were seen. An impression counted below the fold, for a fraction of a second, at the edge of a screen, costs the same in this calculation as one that occupied the viewport for five seconds. Viewability sits alongside this metric for exactly that reason.

The most useful application of this number is also the least obvious one, and it concerns campaigns that were never bought on impressions at all.

An automated bid strategy optimising toward conversions will buy whatever inventory produces conversions most cheaply, and it is indifferent to what that inventory is. Over time delivery drifts: toward audience networks, toward less premium placements, toward formats with lower attention, toward retargeting inventory where the conversions were coming anyway.

The campaign's own reporting shows a stable cost per conversion and nothing else. Computing the effective impression price and watching it fall — while conversion volume stays flat — is one of the few ways to see the drift from outside the platform's own narrative.

Whether that drift is a problem depends on whether those conversions were incremental, which this metric cannot answer either. It can, though, prompt the question.

On a buy priced per thousand impressions, the agreed rate and the effective rate should nearly match. When they do not, there are two usual explanations.

The first is the fee stack. Platform fees, ad-serving costs, data fees and any managed-service margin all sit between what the buyer paid and what the seller received. Both parties can compute a correct effective price and get different answers, because they are dividing different numerators by the same denominator.

The second is impression discrepancy. The buyer's ad server and the seller's counting system rarely agree exactly, and a difference of a few percent is normal. A large or growing discrepancy is worth chasing, because it moves every derived metric on both sides.

Neither is a reason to distrust the metric. Both are reasons to state which numerator and which impression count produced the figure being discussed.

CPM is a price you agree to: what the buy charges per thousand impressions. Effective CPM is a price you discover: total spend divided by impressions delivered, times a thousand, computed after the fact whatever the campaign was bought on. On a campaign bought on a CPM basis the two are nearly identical, and any gap is discrepancy or fees. On a campaign bought on clicks, views, installs or conversions the agreed price is in different units entirely and the effective figure is the only way to put the buy on the same axis as an impression buy. The practical division of labour is that CPM is a negotiation term and effective CPM is a reporting term: one is what a seller quoted and the other is what the media actually cost.

• Compute it for every buy regardless of how it was priced. A portfolio of campaigns bought on four different models cannot be compared any other way.
• Include fees and ad-serving costs in the numerator when comparing suppliers. A low media price with a high platform fee is a common shape and the effective figure is where it becomes visible.
• Split it by placement before drawing a conclusion. A blended figure across feed, stories, in-stream and audience network describes an average nobody bought.
• Read it beside the outcome rate rather than alone. Cheap exposure that converts at nothing is more expensive than dear exposure that converts, and the effective price cannot tell them apart.
• Use it to test whether a performance buy is quietly buying cheap inventory. A falling effective price with flat outcomes usually means the delivery moved to weaker placements.