TL;DR
Traffic is down eighteen per cent on last month. That number is worth nothing on its own, and it is the number most reports lead with. Month over month is a random number generator that produces a panic about four times a year. Compare the same month last year, over two years if you have them.
Traffic is down eighteen per cent on last month. That number, on its own, is worth nothing, and it is the number most reports lead with.
Almost every business has a shape to its year. Until you know yours, month-over-month comparison is a random number generator that produces a panic roughly four times a year.
The comparison that actually works
Same month, previous year. Not last month. Not the rolling average.
If November is always down on October, November being down on October tells you nothing. November being down on last November tells you something, and it is the only comparison that controls for the shape of your year.
Two years of overlay is better than one, because a single previous year could itself have been unusual.
The trap in that, though
Year-over-year has become less reliable than it used to be, and it is worth knowing why before you lean on it entirely.
The share of searches ending without a click has been rising steadily. SparkToro’s analysis of Similarweb’s US clickstream put it at 68.01% in early 2026, against 60.45% in 2024.
So a site holding all of its positions can be down year over year on clicks, every year, for reasons that have nothing to do with the site. A modest annual decline in clicks against flat impressions is now close to the baseline condition, and treating it as a fault sends people looking for a problem that is not there.
Which is why I check impressions alongside clicks before concluding anything. Those two columns answer different questions and the difference between them is usually the actual finding.
Things that look seasonal and are not
A competitor who launched in your quiet season. You notice in the busy one, and it reads as a bad year rather than a new entrant.
A slow decline with a seasonal shape on top. The seasonality masks the trend. Plot twenty-four months rather than twelve and the underlying line becomes visible.
Something that broke during the quiet period. Nobody noticed because traffic was low anyway, and it surfaces as a disappointing peak.
Things that are seasonal and get treated as problems
School holidays for anything sold to parents or professionals.
The genuine off-season in trades, tourism, and anything weather-driven.
Christmas and the fortnight after it, which distorts a December-to-January comparison in nearly every B2B account.
The reporting month itself. A month with fewer working days, or five weekends rather than four, moves B2B numbers by several per cent for no reason at all.
And be careful about which pages you are averaging across. Ahrefs looked at 14 billion pages and found 96.55% get zero traffic from Google. On most sites a handful of pages carry nearly everything, so a “site-wide dip” is frequently one page moving. Segment before you diagnose, or you will investigate a whole site to find one URL.
How I would set this up once
- Build the seasonal baseline before you need it. Twenty-four months of clicks and impressions, plotted. Half an hour, once. Every future conversation about a dip refers to it.
- Annotate it. Launches, migrations, campaigns, confirmed Google updates. A dated line on a chart resolves most arguments about causation in seconds.
- Report year over year by default, with month over month as secondary. This one change removes most false alarms.
- Set a threshold in advance. Decide what size of year-over-year move is worth investigating before you are in the middle of one. Otherwise every number becomes a discussion.
The honest limit
Sometimes there is not enough history. A site eighteen months old has no reliable baseline and you have to say so rather than manufacture confidence. In that situation I would watch impressions, watch the competitor set, and be explicit that the first full year is the one that produces the baseline rather than the answers.
The short version
- Month over month is noise. Same month, previous year.
- Two years of overlay beats one.
- Zero-click has risen from 60% to 68% in two years, so a modest annual click decline can be the baseline, not a fault.
- Check impressions alongside clicks before concluding.
- Plot twenty-four months to see a trend hiding under seasonality.
- Annotate launches, migrations and confirmed updates.
- Set the investigate-threshold in advance, not mid-panic.
If a site should be ranking and it isn’t, that’s the work I do. SEO recovery covers telling a real decline from a normal one. If you’re not sure which of several plausible problems is costing you, that’s what an SEO audit is for.
