A volatility tracker turns red, someone posts a screenshot, and by lunchtime three people have asked whether the site was hit by an update. Most of the time the honest answer is: the tracker measured exactly what it measures every day, and nothing about your site can be inferred from it.
The trackers are useful. But they are useful the way a seismograph is useful — for detecting that the ground moved, not for telling you whether your house cracked. Using them well means knowing how the number is built and what a normal day looks like.
What a volatility index actually measures
Every major tracker — Semrush Sensor, MozCast, Accuranker's Grump, Advanced Web Ranking's index, and the rest — is built on the same skeleton. The vendor tracks a fixed corpus of keywords daily, records the top results for each, compares today's rankings to yesterday's, and aggregates the amount of position change into a single score.
The details differ by vendor: the size and composition of the keyword set, how deep into the results they look, whether a URL entering or leaving the tracked range counts more than a small position shift, and how the raw churn is normalized onto the published scale. MozCast famously expressed its score as a weather temperature to make the point that some heat is normal. Most vendors publish their methodology only at a high level, so treat the score as an index, not a measurement with units.
Three properties follow directly from the construction:
- It is a sample. The corpus is thousands or tens of thousands of keywords, skewed toward whatever the vendor chose — often competitive, mostly English, mostly one country. Your niche may be barely represented.
- It measures churn, not direction. A score of 8 says many tracked results moved. It does not say who won, in which verticals, or why.
- It is relative to that tracker's own baseline. Scores are not comparable across tools, and a "high" reading only means high relative to that corpus's recent history.
Rankings move every day, even when nothing happened
The base rate is the part most people underestimate. On a completely uneventful day, a meaningful fraction of tracked top-10 results still shuffle. Several mechanisms guarantee this:
- Continuous indexing. Google's index updates constantly. New pages enter, stale ones are re-scored, freshness signals decay. There is no static resting state for rankings to return to.
- Live experiments. Google has stated that it runs large numbers of ranking experiments continuously. Some of what a tracker records as volatility is test traffic.
- Query-result instability. Some queries are inherently unstable — news-adjacent, seasonal, or sitting between intents — and flip between result sets without any algorithm change.
- Measurement noise. Trackers query from specific locations and configurations. Data-center variance and result personalization defenses add jitter that has nothing to do with rankings changing for users.
So the meaningful question is never "did the index move today?" It is "did it move materially more than its own recent baseline, for a sustained period, across multiple trackers?" One tool spiking for one day clears none of those bars.
Why the trackers disagree with each other
It is common for one tracker to report turbulence while another reads calm. This is expected, not a flaw. Different corpora sample different verticals; a shake-up concentrated in, say, health queries will register strongly on a corpus rich in health keywords and weakly elsewhere. Different tracking depths matter too: churn in positions 11–20 is chronically higher than in the top 3, so a tool watching deeper results runs hotter. Disagreement between trackers is itself information — it usually means the movement is localized, not systemic.
What a spike does and does not tell you
A spike tells you that many results in that vendor's sample changed position — plausibly an algorithm update, a large-scale test, or turbulence in one heavily sampled vertical. When Google confirms an update on its Search Status Dashboard, the trackers usually corroborate the timing, which is the main honest use: dating an update window after the fact.
A spike does not tell you anything about your site. Your rankings are not in the sample, or are a rounding error within it. Sites gain during volatile periods as often as they lose. Reacting to the index without looking at your own data is responding to weather in another city.
| Signal combination | Reasonable reading |
|---|---|
| Trackers spike, your metrics flat | Update happened elsewhere; do nothing |
| Trackers spike, your traffic drops same window | Plausibly affected; investigate which queries and templates |
| Trackers calm, your traffic drops | Look for a site-side cause first: releases, tracking, seasonality |
| One tracker spikes, others calm | Localized or corpus-specific churn; wait |
The overreaction cost is real
The damage from volatility-watching is rarely the watching. It is the changes shipped in response: rolled-back content, panicked deindexing, rewritten titles — made during a window when rankings were unstable, which then makes the changes impossible to evaluate. If an update is genuinely rolling out, the worst time to change variables is mid-rollout. Google's own public guidance around core updates has consistently pointed the same direction: broad improvements over reactive tweaks.
What to actually do
- Set your baseline first. Track your own priority queries daily, so you know your site's normal churn before any red-dashboard day arrives.
- Use trackers only for dating. When your metrics move, check whether the window matches tracker spikes and the Search Status Dashboard. That is corroboration, not diagnosis.
- Require three signals before acting: sustained multi-day tracker elevation, a confirmed or widely corroborated update, and movement in your own query-level data. Anything less is noise.
- Diagnose from your data, not the index. Segment the loss by query class, template, and country. The volatility score contains none of that resolution.
- Freeze during rollouts. If an update is confirmed and rolling out, note it, hold changes, and evaluate after it settles.
The trackers are a shared clock, not a diagnosis. Read them the way you read weather at the airport: occasionally decisive, usually irrelevant, and never a reason to rebuild the house.