A higher RPM can arrive alongside lower revenue. A full ad slot can earn very little. A growing impression count can reflect more ads on the same pages rather than a larger audience. None of these situations is inherently contradictory.

Publisher reporting becomes more useful when you connect each rate to the volume beneath it. Start with the money, identify the exact population used to calculate each metric, and examine the reader experience alongside the commercial result.

Establish what the report counts

Before comparing two partners or placements, record the revenue basis: currency, reporting period, time zone, included inventory and whether the amount is estimated or finalized. Also establish whether the figure is before or after the relevant fees and revenue share. A number labelled “revenue” is not enough to settle those questions.

Next, name the denominator. A page view, an ad request and an ad impression describe different events. For example, AdSense counts a page view for a page displaying its ads, regardless of how many ads that page contains. Its page-view total should not automatically be substituted with a differently defined analytics total.

Create a short metric dictionary for your own dashboard. Include the provider’s field name and definition, not just the friendly label your team prefers. That small step makes later comparisons and handovers considerably easier.

Read the denominator in every RPM

Google’s RPM explanation describes revenue normalized to a thousand units. It is a rate calculated from earnings and volume, rather than the amount paid to you. The denominator determines the question it answers.

Three useful rates for the same reporting scope
Metric Calculation Useful question
Page RPM Revenue ÷ counted page views × 1,000 How much did this set of pages earn relative to its page-view volume?
Ad impression RPM Revenue ÷ counted ad impressions × 1,000 How much revenue came from each thousand counted ad impressions?
Ad request RPM Revenue ÷ counted ad requests × 1,000 How much revenue came from each thousand requests, including those that did not produce an impression?

Ad platforms may use eCPM for related effective rates. Check the actual formula instead of assuming that every RPM or eCPM column shares a denominator. Compare a placement’s rate with its own history on the same basis before comparing it with another format or partner.

Separate a response from an impression

An ad opportunity can pass through several reporting stages. A request asks for an ad; a response may supply one; impression counting follows the platform’s delivery rules. Do not assume these stages produce identical totals.

Google Ad Manager’s reporting definitions illustrate the distinction: Ad Exchange match rate compares responses served with requests, while its delivery rate compares ad impressions with requests. That documentation also notes that the programmatic eligible request metric counts each matched query when an optimized ad pod fills, rather than always counting one opportunity for the pod. Metric scope therefore matters as much as the formula.

“Fill rate” is used differently across tools. Ask whether a supplier means returned responses, delivered impressions or another event, and which requests qualify for the denominator. For a diagnostic dashboard, retain the underlying counts so the rate can be reconstructed.

A gap between responses and impressions deserves investigation, but it is not automatically lost cash. First check whether the counts describe the same inventory and counting sequence. Then inspect delivery behaviour and implementation rather than assigning a price to every missing event.

Follow one fictional set of numbers

Consider a fictional publisher reporting $120 of revenue from 10,000 counted page views. The same inventory generates 20,000 ad requests, 18,000 responses and 16,000 counted ad impressions. For this simplified example, every request represents one ad opportunity and can produce at most one response and one impression.

  • Page RPM: $120 ÷ 10,000 × 1,000 = $12.
  • Ad request RPM: $120 ÷ 20,000 × 1,000 = $6.
  • Ad impression RPM: $120 ÷ 16,000 × 1,000 = $7.50.
  • Response rate: 18,000 ÷ 20,000 = 90%.
  • Impression-to-request rate: 16,000 ÷ 20,000 = 80%.

These are different descriptions of the same $120, not separate revenue streams. Adding the three RPM values would create a meaningless total. The higher page RPM does not indicate a better advertiser price; a page can contain more than one ad impression.

Now imagine a change reduces impressions to 12,000 while revenue falls to $108 and page views remain unchanged. Impression RPM rises to $9, but page RPM falls to $10.80. The placement became more valuable per counted impression while the publisher earned less overall. Those fictional results show why rate and volume belong together.

Recalculate totals instead of averaging rates

Suppose two non-overlapping inventory groups earn $40 from 10,000 impressions and $60 from 30,000 impressions. Their impression RPMs are $4 and $2. The combined rate is $100 divided by 40,000, multiplied by 1,000: $2.50. A simple average of $3 would give the smaller group too much influence.

Apply the same discipline to devices, countries, placements and partners. Recalculate from compatible revenue and volume totals. Before combining rows, check for overlapping populations and duplicated reporting. A subtotal and its component rows cannot both be added to the grand total.

For page RPM across partners, agree on one page-view population and avoid counting the same pages repeatedly. For impression RPM, use compatible impression definitions. If the definitions cannot be reconciled, keep the figures separate and explain the limitation instead of presenting a precise-looking combined rate.

Evaluate yield alongside delivery and experience

A pricing change can alter both the value of sold inventory and how much sells. Ad Manager’s pricing-rule documentation warns that incorrect configuration can cause delivery problems such as low fill. Raising a floor should therefore be treated as a test, not as a guaranteed revenue increase.

Viewability is another separate dimension. In Active View reporting, the viewable rate uses measurable impressions, not all ad requests. A high response rate cannot establish that ads were viewable; a high viewability rate cannot establish the revenue earned.

When testing a new placement, monitor revenue, volume, delivery and reader behaviour together. Record any changes to ad density or loading behaviour. Use the publisher experience checklist to keep the site usable, and the traffic quality guide to interpret measurement coverage.

Close the loop from estimate to decision

Revenue reports can change after the initial observation. AdSense explains adjustments during filtering and finalization, including processing corrections and invalid activity checks. Use the status shown by your own provider rather than treating a provisional dashboard as the final payment record.

  1. Save a baseline. Capture the same inventory, definitions and revenue basis before the change.
  2. Change one main variable. Keep placement, pricing and traffic acquisition changes distinguishable.
  3. Compare raw totals and rates. Recalculate combined metrics from compatible counts.
  4. Inspect the audience mix. Separate a placement effect from a shift in devices, geography or content.
  5. Revisit finalized figures. Record adjustments and decide whether the result still supports the change.

The purpose of the report is a defensible decision: keep, adjust or reverse a specific change. A smaller set of well-defined metrics makes that decision clearer than a dashboard full of incomparable averages.

Sources & further reading

Explore advertising with Promoweb.
Meet our self-serve platform