When a Keyword Falls Out of the Top 100

Disappearing is a different kind of event from dropping, and most tools store it badly. A keyword that was at 12 and is now beyond your tracker’s search depth hasn’t moved 88 places — you simply stopped being able to see it. How your tool records that absence determines whether every aggregate you compute is now wrong.

Both halves matter: what it means, and what it does to your arithmetic.

Three ways tools record absence

As a null or blank. The most honest option. The keyword has no position, and any average computed over the set has to decide whether to skip it. Skipping means your average silently improves when a keyword disappears, which is the wrong direction.

As the depth limit plus one — typically 101. Convenient and misleading: it’s a made-up number that then participates in averages as though it were an observation. A keyword sitting at “101” for six months is anchoring every average you compute.

As the last known value. The worst option, and some tools do it on a failed fetch. Your series shows a flat line where the truth is “we don’t know”, and flat looks like stability.

Find out which your tool does, because you can’t interpret aggregates without knowing. A flat stretch in a series is always worth a suspicious look — it’s more often a fetch failure than genuine stability.

Why this breaks aggregates

If absence is a null and your average skips nulls, losing your worst keywords improves your average. That’s a metric that rewards failure.

If absence is 101, then one keyword disappearing moves an average across thirty keywords by roughly three places, which will read as a broad decline when it’s one event. And the 101 is not a position anyone observed.

The way out is the same as elsewhere in this topic: count keywords in bands rather than averaging positions, and give absence its own band. “Twenty-two of thirty in the top fifty, one not found” is unambiguous and can’t be broken by a null. The general argument is in position is an ordinal, not a quantity.

What disappearance usually means

Ranked by how often it turns out to be the answer:

A tool problem. A failed fetch, a location change, a tracked URL that now redirects so the domain-matching rule no longer fires. Check this first; it’s the most common and the cheapest to rule out.

A URL change. The page moved, the slug changed, the canonical points elsewhere. The site is fine; the tracker is watching an address that no longer holds the content.

An index problem. The page is genuinely no longer in the index as the version you expect. Directive changes, accidental or otherwise, land here.

A real competitive collapse. Rare in one step. Falling from page one to beyond a hundred in a single day is almost never competitive, because competitive change is gradual.

Depth, not disappearance. If your tool only checks two pages of results, “not found” means “not in the top twenty”. Check your tool’s configured depth before concluding anything at all.

Treat it as its own alert

Because the causes skew technical, disappearance deserves separate, higher-urgency handling than a drop. A keyword crossing from 8 to 15 can wait for the weekly. A keyword going from 8 to nothing wants a look at the URL’s index state today.

It’s also breadth-sensitive: one keyword disappearing is a page-level or tool-level issue, and twenty disappearing at once is a sitewide one. Count first — sitewide drops versus page-level drops — and set the alert up so it can distinguish the two, per rank alerts that don’t cry wolf.

Keywords that were never there

A separate case with the same symptom: keywords you added as targets and have never ranked for. These sit permanently at not-found and do exactly the same damage to aggregates as a genuine disappearance, but no investigation will ever help.

Keep them in a distinct list, excluded from the headline numbers, and review them for one thing only: did any of them start ranking. That’s a genuinely useful signal and it’s invisible when they’re mixed in with everything else — choosing which keywords to track.

What to actually do

1. Find out how your tool stores absence, and whether it ever carries a value forward.
2. Report band counts with a not-found band instead of averaging positions.
3. Rule out tool and URL causes first — they’re most of them.
4. Alert on disappearance separately, with breadth distinguishing page-level from sitewide.
5. Keep never-ranked targets out of your aggregates entirely.