Start by verifying the comparison
Before looking for a cause, make sure the drop is real and the two periods are comparable. Confirm the same property, currency, time zone, date range length, reporting filters and earnings definition. An incomplete current day, a late report update or a filter left on in one export can look like a performance drop.
Step 1: compare earnings, pageviews and page RPM
Use three values together: estimated earnings, pageviews and page RPM. Page RPM is earnings divided by pageviews multiplied by 1,000. If earnings fell while pageviews were stable, the ratio likely changed. If pageviews fell and page RPM stayed close, traffic volume may account for much of the revenue decline. If both fell, investigate both rather than choosing one cause too soon.
Step 2: decompose the arithmetic
Consider this illustrative example. Period A has 100,000 pageviews and $420 in earnings: page RPM is $4.20. Period B has 90,000 pageviews and $324 in earnings: page RPM is $3.60. Revenue fell by $96. Holding the old RPM constant, the 10,000-view decline corresponds to a $42 arithmetic traffic effect. Applying the RPM decline to the latest 90,000 pageviews corresponds to a further $54 arithmetic RPM effect. Together they account for the $96 difference.
This split is a math identity based on the order used. It does not prove that “traffic” or “RPM” had one specific cause. RPM itself can reflect the mix of countries, devices, pages, formats, demand, viewability and reporting definitions.
Step 3: segment by page or country
Compare the largest pages and regions across both periods. A site-wide average can hide the fact that one high-traffic page lost visits or that traffic shifted toward pages with a lower historical RPM. Use comparable segments, and avoid making decisions from tiny samples. If a report does not include page or country dimensions, do not invent them or estimate missing data.
Step 4: check acquisition and technical changes
- In Search Console, inspect clicks, impressions, average position and affected landing pages for the same dates.
- Check analytics traffic by channel and landing page, without treating correlation as proof.
- Review recent content edits, redirects, canonical tags, robots directives, consent behavior and ad-layout changes.
- Confirm that reporting filters, time zones and page definitions did not change.
- Look for invalid-traffic notifications or reporting adjustments in the ad platform, if present.
Step 5: decide what evidence would confirm the hypothesis
Write the observation separately from the hypothesis. “Page RPM fell from $4.20 to $3.60” is an observation. “A template change caused it” is a hypothesis until timing and segmented data support it. Change one relevant factor at a time where possible, document the date and compare a suitable post-change period.
Use the report analyzer
Publisher Pulse can compare adjacent date ranges, show a page-level earnings table and export daily summaries from a CSV you provide. It processes the selected file in the browser and cannot verify that the source export is authentic or establish causation. Open Publisher Pulse.
Use a decision tree instead of a guess
If pageviews and RPM both remain stable but earnings change, first inspect the earnings definition, report completeness and filters. If pageviews fall while RPM is stable, inspect traffic channels, search visibility and the landing pages contributing the largest decline. If pageviews are stable but RPM falls, compare page, country and device groups and review whether the report or ad configuration changed. If both fall, investigate each branch separately.
This is a triage sequence, not a causal model. Several effects may happen at once, and small samples can move sharply. Prefer evidence that repeats across comparable periods over a single day or isolated page.
Sources and definitions
- Google AdSense Help: Page RPM
- Google Search Central: Search Console guidance
- Google Search Central: People-first content