How to Diagnose a Ranking Drop Without Guessing

Narrow the blast radius before you form a theory. A drop on one keyword, a drop across a page’s whole keyword set, a drop across a section and a drop across the site have almost disjoint sets of causes. Establishing which one you have takes ten minutes and eliminates most of the possibilities for free.

The reason to be systematic is that ranking drops are the single easiest place in SEO to construct a confident wrong story, and then act on it.

Step one: is it real?

Before anything else, check whether the movement exceeds your normal range for that keyword. Most keywords have a band they wander inside, and a drop within the band is not an event. If you don’t know the band, you don’t yet know whether anything happened — baselines before you read a position.

Two specific false alarms to rule out:

A tool artefact. A failed fetch reported as a drop, a location setting someone changed, a tracked URL that’s now redirecting. Check the tracker’s own history for a suspicious flat spot or a single-day spike.
A layout change. If the result page gained a feature block, your clicks can fall with your position intact, and someone may have described that to you as a “drop” — what position one means on a crowded result page.

Step two: how wide is it?

This is the decisive question. Count what moved.

One keyword, one URL. Almost always a query-level or competitive change: someone published something better, the result page’s intent shifted, or two of your own pages are competing for it. Cheap to investigate, rarely urgent.

Every keyword for one URL. Something about that page or that URL. Check the URL first — is it still the URL, is it indexed, did it change, is it canonicalised somewhere else. Page-level causes are the most fixable category.

A section or template. Something shared: a template change, a directive applied too broadly, an internal linking change, a batch of pages published or removed.

The whole site. Technical or sitewide, and the only category that deserves same-day escalation. Server errors, an accidental site-level directive, a certificate problem, a DNS change, a migration that went partly wrong.

Everyone, including competitors. Not about you at all. If your tracker shows the whole result set reshuffling, or a volatility index is elevated, wait — what volatility trackers measure.

Step three: what changed, and when

Now, and only now, look for causes — and start with your own changelog rather than with the outside world.

Order the candidates by how likely they are and how cheaply you can check them:

Your own releases. Deploys, template changes, content edits, redirect changes, robots or meta directive changes. This is the most common answer and the one teams check last, because it’s the least interesting.
Index state. Is the URL still indexed, still the canonical, still returning 200 to a crawler rather than to you.
The result page itself. Look at it. Who’s above you now that wasn’t? Is the page answering a different intent than it was?
Broad volatility. Elevated across the board, and across other people’s keywords too.
A manual or systemic action. Rare, and it announces itself in Search Console. Don’t reach for it as an explanation for a five-place slip.

Timing evidence is the strongest evidence

The shape of the drop tells you a lot about its cause, and it’s free to look at.

A cliff on a single day points at a discrete change: a deploy, a directive, a redirect. Find what shipped that day.

A slide over two or three weeks points at something gradual — a broad reassessment, a competitor gaining, or an accumulating technical issue.

A drop that recovers within days was noise or a transient fetch problem, and the correct response is to have not done anything.

A step change with a flat line after it is often a tool or configuration change rather than a ranking change.

Say what you don’t know

The honest report frequently ends “one keyword moved four places, within its normal range, no action.” That’s a good outcome and it’s most of them.

When the cause genuinely isn’t identifiable, write that down instead of promoting the best-sounding candidate to a conclusion. An unexplained drop that later recovers is common; an unexplained drop that you attributed to a guess is how a team acquires a false belief and then optimises against it. The frequency argument for this restraint is in how often should you check keyword rankings.

What to actually do

Compare against the keyword’s normal range before calling it a drop.
Establish the blast radius — one keyword, one URL, one template, or the site.
Check your own changelog before anything external.
Read the shape of the curve: cliff, slide, or spike-and-recover.
Write “cause not established” when that’s the truth, and set a date to look again rather than acting on a guess.