When Google sends fewer people, every number in Google Ad Manager goes down together, and that is the trap. A 30 percent fall in sessions drags requests, impressions and revenue down by roughly 30 percent, and a dashboard full of red arrows makes the ad stack look broken when it is doing exactly what it did last month with fewer visitors. The job in a traffic decline is to separate the two problems, measure the one you can affect, and stop touching the one you cannot.
Here is how to read the reports when the top of the funnel narrows.
Confirm it is traffic, in one report
Put two series next to each other for the same hours: sessions (or pageviews) from your analytics tool, and ad requests from Ad Manager. On a healthy site the ratio between them is close to constant. That ratio is the first thing to check.
- Sessions down 30 percent, ad requests down about 30 percent: the stack is fine. The loss is traffic.
- Sessions down 30 percent, ad requests down 50 percent: something on the page also changed. A template that lost a slot, a consent flow that now blocks the first ad call, a lazy-load rule that no longer fires on the shorter sessions you are left with.
- Sessions flat, revenue down: not a traffic story at all. Go to the yield section and skip the rest.
People Inc. reported in August 2026 that sessions were down 22 percent year on year and Google search traffic down 40 percent, while its open-marketplace programmatic revenue grew. Those two facts in the same quarter are the whole point: traffic and yield are separate problems, and they can move in opposite directions.
Switch the dashboard to rates, not totals
Totals follow traffic. Rates do not. For the duration of a decline, the numbers to watch are:
| Instead of | Watch | Why |
|---|---|---|
| Ad server revenue | Revenue per 1,000 ad requests | Immune to volume; moves only with fill and price |
| Impressions | Fill rate (impressions divided by ad requests) | Shows whether demand is keeping up with what you still have |
| Ad requests | Ad requests per pageview | Catches slots that stopped loading, which traffic alone cannot explain |
| Total unfilled | Unfilled share by ad unit | Points at the units where demand left, not just the ones with less traffic |
Build this as a saved report with Ad unit, Device category and Country as dimensions, daily, with a 28-day window. A rate that holds while totals fall means you have nothing to fix in the stack this week. A rate that falls means you do.
What falling traffic does to the auction
Traffic rarely falls evenly, and the shape matters more than the size.
It falls on specific pages. Search sends people to evergreen articles and guides. When those referrals shrink, the sessions you keep are more homepage, more section fronts, more returning readers from newsletters and direct. Those pages have different ad units, different viewability and different buyers. Revenue per request can rise in total while individual article units get worse, simply because the mix changed.
It falls on specific devices. Search referrals skew mobile on most news sites. Less mobile traffic shifts the mix toward desktop, where CPMs are usually higher. A rising blended CPM during a traffic decline is often mix, not improvement. Always split the rate reports by device before celebrating or panicking.
It changes who buys. Fewer sessions means fewer impressions for the audience segments that retargeters and PMP buyers chase. If a deal or a line item has a pacing goal, it will now under-deliver, and the ad server will push it harder into inventory it did not used to win. Check the Delivery report for line items that moved from "on schedule" to "behind".
It changes refresh and viewability math. Readers who come through search tend to stay for one page. Readers who come direct stay for several. More pages per session means more in-view time, more refreshes if you run them, and a higher share of impressions that count as viewable. Again, a rate can improve while the business shrinks.
The yield checklist, for when the rate is falling too
If revenue per request is falling along with traffic, run this order:
- Ad requests per pageview, by template. A template that dropped from four requests per view to three has lost a slot, whatever the traffic did.
- Unfilled share by ad unit and by country. Demand pulls back from specific markets first. A country whose unfilled share doubled is a partner or a consent problem.
- Pricing rules report. Floors set during a stronger quarter may now sit above what the remaining buyers bid. The report shows impressions blocked by rule; a rule that blocks more than it used to while fill falls is the one to loosen. One rule at a time.
- Line items that ended or were paused. Sponsorships that lapsed this month explain a surprising amount of "programmatic weakness".
- Prebid bidders that went quiet. With fewer auctions, a bidder that timed out on a wrapper update is easier to miss in the totals. Look at bid rate per bidder, not just revenue per bidder.
Most of these are comparisons across time, which is tedious to do by hand every morning and easy to do by machine. Optimon reads your Google Ad Manager and Prebid data every day and tells you which rule, which unit and which bidder moved, with the numbers next to it, so a traffic decline reads as what it is and the real yield changes do not hide inside it.
What to say to the people who ask
The question from the top of the building will be "why is revenue down" and the honest answer has two parts. The first is a traffic number that belongs to search, social and product, not to ad ops. The second is a rate number that belongs to you, and the rate report is how you show that it held. Bring both to the meeting. The ad stack cannot bring back the visitors Google kept, but it can be shown to have made the most of the ones who arrived, and that is the number that decides whether the next conversation is about yield or about headcount.
FAQ
How do I know whether a revenue drop is traffic or yield?
Put ad requests next to sessions for the same period. If both fell by roughly the same share, the ad stack is doing its job and the loss is traffic. If sessions held and revenue fell, or revenue fell much more than sessions, you have a yield problem worth a morning of work.
Which GAM metric is the best single health number when traffic is falling?
Ad server revenue per thousand ad requests (revenue divided by ad requests, times 1,000). It is immune to the volume change and moves only when fill or price moves. Track it per ad unit and per device, not just in total.
Should I change floors when traffic drops?
Not as a reflex. Fewer impressions do not make each one cheaper; the buyers on your remaining traffic are the same buyers. Change a floor only when the pricing report shows a rule blocking bids that used to clear, and change one rule at a time so the report can tell you what happened.
