CPM, eCPM, RPM and fill rate: the metrics that actually matter

A monetization dashboard shows half a dozen acronyms that look like they say the same thing, and they do not. Each answers a different question, and using the wrong one leads to expensive decisions. Here is what each measures, why two tools never agree, and which to check when.

A dashboard full of acronyms will not tell you if you are winning

Anyone opening a monetization platform for the first time finds a row of numbers that seem to say the same thing under different names: CPM, eCPM, RPM, fill rate, viewability, impressions. They rise and fall together, and none of them alone explains whether the month went well.

They are not synonyms: each answers one specific question, and trouble starts when you use one to answer another one’s question.

CPM: the price the advertiser pays

CPM stands for cost per mille. It is what an advertiser pays for every thousand impressions of their ad: a buying price, describing the side putting up the money, not the side receiving it.

As an invented arithmetic example: a campaign bought at a 3 dollar CPM that ends up serving 50,000 impressions on your site has spent 150 dollars. What reaches the publisher is less, because intermediary platforms take fees along the way. Hence the first question to ask of any dashboard: is the CPM it shows gross or net?

eCPM: the effective price, the one you can compare

eCPM (effective CPM) is not an agreed price but a calculation made backwards: revenue divided by impressions, multiplied by a thousand. With invented numbers: 200 dollars of revenue on 100,000 impressions gives an eCPM of 2.

A cost-per-click campaign, a fixed-price deal and an open auction are not comparable to each other, but their eCPMs are: that is why it is the metric a mediation system uses to decide who gets each impression.

There is a catch: eCPM only describes the impressions that source actually served. Very expensive demand that responds once in ten requests can leave you less money than cheap demand that responds every time.

Fill rate: how many requests end up as an ad

Fill rate is the percentage of ad requests that return an ad. If your pages fire 100,000 requests and 80,000 ads are rendered, the fill rate is 80%. The rest are slots that stayed empty or were filled with something that pays nothing.

A slot going unfilled is rarely random. The usual causes: a price floor above what demand is willing to pay, countries with no buyers for that inventory, formats with little demand, users without consent — which sharply reduces the bids received — and slots that request the ad so late the user has already left.

A high eCPM with a low fill rate is an excellent price that almost nobody ends up paying.

RPM: the one that actually answers “how much am I making”

RPM stands for revenue per mille. The key question is: per thousand what. There is RPM per thousand impressions, per thousand sessions and per thousand page views: different numbers on different bases, and comparing them to each other is one of the easiest ways to fool yourself.

What a media owner cares about is almost always page view RPM: revenue divided by page views, times a thousand. Invented example: 300 dollars on 150,000 page views gives an RPM of 2. That number summarises three things — how many slots there are per page, what share of them fill, and at what price — and it answers whether publishing a page pays more today than last month.

Viewability: serving an ad is not the same as it being seen

Viewability measures what percentage of served ads had a real chance of being seen. The most widespread standard counts a display ad as viewable when at least half of its pixels have been on screen for one second; for video the usual threshold is two seconds.

It matters because many buyers only bid on inventory that meets that threshold: low viewability means less demand and worse prices. And it explains what looks contradictory: a slot at the end of a long article piles up impressions and contributes little, because almost nobody scrolls that far. Removing slots can raise revenue if the remaining ones are seen more.

Served versus measured impressions: why two dashboards never match

One dashboard says one million impressions and another nine hundred and forty thousand for the same day and the same site. There is almost never an error: they count different things at different moments.

• Moment of counting: some record the impression when the ad is requested, others when it is returned, others only once it has been rendered on screen.
• Ad blockers and tabs closed early: the ad left the server but never got to render.
• Time zone: one dashboard closes the day on the publisher’s clock and another on the buyer’s, so the days do not line up.
• Invalid traffic filtering: each platform discounts what its own system detects, and they do not all detect the same things.

The practical rule: a small, stable discrepancy is normal. Pick one dashboard as the source of truth for money — usually the one belonging to whoever pays — and use the others for diagnosis. The alarm signal is that gap changing suddenly: it usually means something broke in the implementation.

The classic mistakes when reading these metrics

• Looking at CPM and ignoring fill rate. Raising the price floor pushes the reported eCPM up and revenue down: fewer impressions sold, even if each at a better price.
• Comparing RPM across sections with no adjustment. The homepage, a long article and a gallery have neither the same slots nor the same kind of visit.
• Drawing conclusions from a single day. Weekends, holidays and quarter changes move demand all by themselves.
• Mixing session RPM with page view RPM when comparing periods, and believing the site improved when what changed was the denominator.
• Celebrating a record number of impressions without checking viewability: more ads served to fewer people can leave the same money with a worse experience.

Comparing RPM across sections deserves a note, because it is the most tempting of all. It only makes sense with the same format, device and traffic source: a section that lives on search and one that lives on social will have different RPMs even if the content is identical. If you cannot level the context, compare each section against itself over time.

Which metric to look at for each decision

The quick way to organise this is to start from the question rather than from the dashboard.

• Is the site making more than before? RPM per thousand page views, comparing equivalent periods.
• Is this demand source worth it? eCPM and fill rate together, never one alone.
• Is this slot well placed? Viewability first, then the eCPM of that position.
• Is something broken? A falling fill rate, or a dashboard discrepancy that suddenly spikes.
• What is my audience worth to an advertiser? CPM, the price seen from the buying side.

None of them is better than the others; what does not work is reading them in isolation. CPM without fill rate is a theoretical price, fill rate without price is cheap filler, and RPM without knowing its base means nothing. Together they tell you where money is being left on the table.