Publisher Revenue Changed: A Calm Anomaly-Checking Sequence

Before you start: two rules

Rule one: measure before you feel. "Revenue is down" is not a fact until you can write "down how much, over what window, compared with what".

Rule two: one hypothesis at a time. The moment you change your ad layout, your theme and your posting schedule in the same panicked afternoon, you destroy your ability to know what happened.

The sequence, in order

Step	Question	Where to look	Evidence that satisfies	Common mistake
1	Is the data even final?	Your report's date range, especially the last 1 to 3 days	Recent days excluded or re-checked later	Panicking over provisional figures
2	How big is the change?	Earnings and Page RPM versus the same-length previous window	A percentage inside or outside your usual range	Comparing unlike windows
3	Did traffic move?	Visits or page views for both windows	Traffic explains the direction of the move	Ignoring traffic entirely
4	Did the traffic mix move?	Sources: search, social, referral, direct	A source grew or shrank noticeably	Averaging away the shift
5	Did geography move?	Top countries or regions for both windows	A shift in where readers come from	Assuming home-market behaviour everywhere
6	Any policy alerts?	Account notifications and email	Empty inbox, or a read and handled alert	Treating alerts like spam
7	Did the site change?	Your deployment log, theme updates, plugin or speed changes	A dated change near the shift	Forgetting Friday's deploy
8	Is it seasonal?	Last year's same weeks, or your notes	A recurring pattern	Rediscovering the same dip each year

Step 1: freshness first

The last few days in any report are the least trustworthy, and figures are described by Google as estimates that can be revised. Before you investigate anything, re-run the comparison with the freshest three days removed. If the anomaly lives only in those days, wait 48 hours and look again. You will save yourself several false alarms a year with this habit alone.

Steps 2 and 3: size, then traffic

Put a number on the change: "Page RPM is 18 per cent below the previous four-week average". Your normal range is whatever your own logged history says it is; no other publisher's numbers apply.

Then check traffic before anything else, because it is the most common explanation. The three-way read from our companion piece applies: earnings moving with traffic is normal site behaviour; flat traffic with changed earnings points toward click value or ad serving; falling traffic with rising RPM is a mix shift. If discovery is the patient, our guides on local SEO and growing from 10K to 100K users treat it.

Steps 4 and 5: mix and geography

Two sites with identical traffic totals can earn very differently if one is mostly search visits from readers who arrive with intent and the other is mostly passing social traffic. Check whether one source replaced another, then check whether the geography of your readers shifted. Neither is a problem by itself; both are explanations. Note them and stop.

Step 6: policy, always

Policy alerts are the one thing in this sequence that cannot wait. Google's program policies set out what publishers must not do, including encouraging clicks or using layouts that invite accidental clicks, and its invalid traffic guidance prohibits any artificial generation of clicks or impressions. If your anomaly is a sudden spike in clicks you cannot explain, the correct response is to document it, check the official guidance, and use the account's official channels. The incorrect responses include clicking ads to "see if they work", asking anyone to click, and quietly hoping.

Step 7: your own changes

Pull up your own change log: deploys, theme updates, new plugins, removed sections, hosting moves. Anything dated within a few days before the shift is a suspect, and hosting-level changes belong on the same list, which is why we keep notes alongside guides like choosing hosting in India. If a change and the shift line up, you have a hypothesis, not yet a verdict: revert or observe, and give it a full cycle.

Step 8: seasonality, the quiet explainer

Festivals, exam seasons, monsoon disruption of reading habits, fiscal-year advertiser swings: all of these move Indian publishing revenue in patterns that repeat. Your notes from previous months and years are the only authority here. This is the strongest argument for keeping a three-line log with every review.

When to stop and when to escalate

Stop when: the change is explained by traffic, season or a known site change, or when it is inside your normal range. Log it and move on.

Escalate through official channels when: alerts are present, the shift is large and unexplained after the full sequence, or you see click patterns you cannot account for. Use only your advertising account's official support route.

Never: respond to a dip by testing ad behaviour yourself, buy traffic to "fix" a fall, or take a stranger's diagnosis from a screenshot.

FAQs

How long should I wait before calling a change real?

Give any unexplained movement at least seven full days, and prefer fourteen. Short windows amplify noise, and estimated figures settle.

Revenue fell but traffic is flat. Is that bad?

It means value per visit moved, so check clicks and CPC next. It can be advertiser demand or geography, neither of which you control, so verify before acting.

Should I check my dashboard daily during an anomaly?

Check on a fixed cadence, every two or three days, with the same windows. Daily checks during an anomaly invite rushed changes that corrupt the evidence.

What if I find nothing in all eight steps?

Write "unexplained" in your log with the numbers, keep the change list clean, and re-check in a week. Unexplained is an honest result; a manufactured cause is not.

Can a tool check all this for me?

Tools can collect the numbers, but the sequence's judgement, comparing windows and refusing invented causes, is yours. If you run many properties and this sequence repeats weekly, that is the signal to consider purpose-built software rather than more tabs.

Your next step

Give the sequence a home. Create a dated checklist in Taskss, from the Big Helpers apps page, with one task per step and a due date spread across the week, so an anomaly gets investigated once, calmly, and then either closed or escalated. Taskss is free while in beta, available on iOS, with Android in controlled testing.