When it breaks
RPM fell, traffic held
Pageviews are flat and the revenue line is not. Before anyone sells you a fix, give the dashboard one honest afternoon: an RPM that falls under steady traffic has four ordinary explanations, two of which are the market's fault and two of which are yours to fix this week. Telling them apart is mostly a matter of asking the reports the right questions in the right order.
First confirm the shape. If pageviews fell too, this is the wrong page; that failure is covered at When it breaks. If impressions collapsed and your Policy center has a notice, you have a serving limit, which is enforcement, not economics. Undiagnosed? Take the triage quiz.
The four causes
The ordinary explanations, in the order to check them
1. The calendar
Advertisers spend against budgets, and budgets live on a calendar. The fourth quarter piles holiday money into the auction and bids up every impression, including yours; in January the budgets reset and the pressure releases. The result is the Q1 cliff: RPM falls off the December peak while your traffic does nothing wrong. This is an evergreen pattern the ad market produces every year, which is exactly why the diagnostic is a date trick: compare this month against the same month last year, never against the quarter you just left. A January that looks bad against December and normal against last January is not a problem. It is a season.
2. Demand moved
Between seasons, demand itself shifts: advertiser categories pull back, verticals cool, and the money bidding on your topic thins out. The tell is that everything on your side of the ledger is flat while the price side sags: viewability unchanged, click-through unchanged, but cost per click or impression RPM drifting down. You cannot fix this from your ad settings, and the attempt usually does damage. What raises the ceiling against soft demand is more bidders per impression, which is a network question, not a layout question; that ladder is on the networks page, and the mechanics of who is actually bidding on you are in the AdSense teardown.
3. Viewability and placement drift
The fixable one, and the one nobody suspects because "we didn't change the ads." You changed the site: a theme update, a new hero component, a plugin that slowed the load, content edits that pushed the first unit below the fold, an anchor that quietly stopped rendering. Ads that are seen less earn less, at the same traffic and the same demand. The tell is per-unit: open the Active View viewability trend for each unit in your reports and look for the slide. Our data on our own network puts a 60-point viewability spread between good and bad positions on the same pages, which is why every unit gets held to the 60% floor. The free scan reads your live layout and Core Web Vitals from the outside in seconds.
4. Consent and privacy
The auction pays for signals, and consent decides how many signals it gets. A consent banner update, a CMP misconfiguration, or a privacy change upstream can shift a slice of your impressions into unpersonalized territory, where advertisers bid less because they know less. The tell is geographic: segment revenue by region, and if the drop concentrates in consent-governed markets like the EEA or UK right after a banner, CMP, or platform change, you have found it. Platform-side changes arrive on their own schedule, which is why we keep a plain-English log of AdSense policy and feature changes worth checking against your drop's start date.
The afternoon
Telling them apart in your dashboard
| What the reports show | Likely cause | Your move |
|---|---|---|
| Down against last month, normal against the same month last year | Seasonality | Nothing. Compare year over year and do not redesign in February. |
| Viewability and CTR flat, price per click or impression down, all regions | Demand shift | Ride it; raise the ceiling with more bidders when eligible, not more units. |
| One or more units' Active View viewability sliding over weeks | Placement drift | Find what changed on the page; restore or move the unit. This is the audit's home turf. |
| Drop concentrated in consent-governed regions after a banner or CMP change | Consent change | Test the consent flow end to end; verify the CMP fires and the choices reach your ad code. |
| Impressions collapsed, Policy center notice present | Serving limit | Different failure entirely: the serving-limit page. |
The same logic applies on a managed network's dashboard; the metric names differ, the questions do not. Your network's reporting also shows fill and bid depth, which sharpens the demand-shift read.
The verdict
Your problem, or the market's
Seasonality and demand are the market's problem. The calm response is the hard one: change nothing rash, compare against the right baseline, and put the energy into content and, when you clear a bar, into a network that adds bidders. The panic response, adding more ad units to make the number back, is the one move that reliably backfires; it trades the rankings that bring the traffic for a temporary bump, which is the whole case against it in the AdSense playbook.
Placement drift and consent breakage are your problem, which is the good news, because yours are the only ones that can be fixed this week. If the afternoon in the dashboard points at placement, that is precisely the case the audit exists for: it scores every unit on your real viewability reporting, finds where the layout leaks, and tells you whether the leak or the market explains your number. No hype in this: if your drop is seasonal, the audit will tell you that too, and the placement work simply makes the recovery land on a better layout.