Explaining a Traffic Drop: A Diagnostic Checklist

Full transcript

1,285 words · script draft, under review.

Intro

Say traffic is down thirty percent in two weeks. Within an hour, you'll have five theories. A Google update. A penalty. A competitor. Something the developers changed. Bad luck.

The problem isn't a lack of ideas. It's that people check them in the wrong order, starting with the most dramatic one. They spend a day reading about algorithm updates, and the real cause was a tracking tag that disappeared. Or there was no cause at all.

In this episode I'll give you the checklist I use. Six steps, always in the same order, and we'll walk through them on the worst week one of my own sites has had.

Why the order matters

The order follows one principle: rule out the cheap and common causes before the rare and expensive ones. And each step removes a whole branch of possibilities from the next.

There's no point analyzing rankings if the drop only exists in Analytics. There's no point studying the results page if half your pages fell out of the index.

Here's the case, from my site, example-city-stats.com. The data is real, but anonymized: the domain, the URLs and the search terms are replaced, so we stay on the principles and not on the keywords. The numbers are untouched.

The site is young. Its Search Console history starts in early March, about seven months ago, and traffic has grown for most of that time. But not in a straight line. The worst week so far was in early April. Clicks fell from about two hundred and thirty to about a hundred and sixty in one week. That's almost exactly the thirty percent from the start. On a weekly chart, it looks like a cliff.

Before step one, define the drop properly. Compare equal periods, with the same weekdays. Look at the drop in Search Console clicks, not only in your dashboard. And check that it's bigger than the normal week-to-week noise.

Step 1: Is it measurement?

Step one: is the drop real, or is it the measurement?

Compare Search Console clicks with Analytics sessions from Google organic. If Search Console is flat and only Analytics fell, Google is still sending visitors. You've lost the ability to count them. Look for a missing tag after a deploy, a new cookie banner, or a change in filters. And check the hostname filter from episode fifteen.

On my site, this check hit a wall right away. Analytics recorded almost no organic sessions before the middle of May. So for a week in April, Search Console is the only witness. Better to learn that about your own data before you need it.

So I checked the other half of step one: is it bigger than the noise? On this site, the week-to-week change in clicks has a standard deviation of about twenty-four percent. Swings that size are normal. And look at the week before. Clicks had jumped by more than half. The drop gave back part of a spike, and the bad week was still above the week before that spike. Across the whole history, not one weekly drop is bigger than the normal noise.

So the diagnosis ends here, at step one. And that's the most common outcome. Most drops die at step one.

Also check the dates. The most recent days in Search Console are often incomplete. A drop in the last two days of a chart is often just data that hasn't arrived yet.

Only when both sources fell together, and the fall is bigger than the noise, is the drop real. Then you move on. For the remaining steps, I'll show you what I'd pull next if this one had passed.

Step 2: Is it demand?

Step two: did fewer people search?

Use the tools from episode thirteen. Normally you'd compare with the same weeks last year. On my site that's impossible. The history is only about seven months long. When that happens, say so in your notes, and lean on the demand index instead: monthly search volumes for your main queries. If demand fell as much as your clicks, and your clicks per unit of demand held steady, it's seasonality or a market shift. Nothing on your site broke.

And split brand from non-brand, as in episode fourteen. If only brand searches fell, look at your marketing, not your SEO.

Step 3: Is it indexing?

Step three: are the pages still in Google's index?

Switch from queries to pages. Find the pages that lost the most clicks, and group them by section or template. A drop that's spread evenly across the site points one way. A drop concentrated in one section points to something technical in that section.

For those pages, check index status with the URL Inspection tool, on a sample, not just one page. And look for the usual suspects after a release: a noindex tag, a robots.txt rule, a changed canonical, a redirect that points to the wrong place.

One warning: don't use the "indexed" count from the sitemaps report as your evidence. It's unreliable. Episode seventeen covers how to read indexing properly.

Step 4: Is it rankings?

Step four: did the pages lose positions?

Compare the same queries in both periods. Position worse and impressions down: that's a ranking loss. Now ask where. A few queries or all of them? One page type or the whole site?

A loss on a few queries often has a local explanation: a competitor published something better, or two of your own pages started competing, as in episode twelve. A loss across the whole site, starting on one date, is when you check whether Google announced an update that week. That's the right moment for that theory. Not step one.

Step 5: Did the results page change?

Step five: did the results page itself change?

Sometimes your position barely moves, but clicks fall anyway. Look at the live results for your biggest queries. Has a new feature appeared above you? An AI-generated summary, a video row, a box of questions, more ads? Each one takes attention, and clicks, from the plain results.

Check the devices too. Search Console lets you split by mobile and desktop. A change on the results page often hits mobile first, because the screen fills up faster.

If you store results-page snapshots, as we'll set up in the SERP module, you can compare before and after directly. If you don't, today's results page is still evidence. Note what's there.

Step 6: Is it click-through rate?

Step six: same position, same results page, fewer clicks. Now it's your snippet.

Check whether your title changed, either by you or by Google rewriting it. Run the comparison against your own CTR curve from episode eleven. If a group of pages dropped below the line at the same time, look for what they have in common: a template change, a new title format, a missing date.

Write the diagnosis

When you find the cause, stop and write it down in one line, with the evidence. Which step found it, what the evidence was, the action, and a date to re-measure. On screen is the line for the week we just walked through. Found at step one. Evidence: inside the normal noise, right after a spike. Diagnosis: no diagnosis needed. Action: none.

If you reach step six without an answer, go back to step one with a narrower question. Not "why did the site drop", but "why did this section drop on this date". Most hard cases become easy once the scope is small enough.

That one line is worth more than a long report, because next time, you'll have a history of what went wrong, what didn't, and how you found out.

Wrap-up

So: measurement, demand, indexing, rankings, results page, and click-through rate, in that order. The full diagnostic checklist is linked in the description. Next episode: crawl, index and rank, and how to read the signals that tell you where a page is stuck.