How the Google Dance worked
In Google’s early years the index was rebuilt in batches rather than updated continuously. During a rebuild Google’s data centres did not all hold the same version of the index, so the same search could return different results depending on which centre answered it. Rankings appeared to swing for several days, then settled until the next rebuild came round.
Site owners named that period the Google Dance, watched for it and planned around it. It was a genuine phenomenon with a mechanical cause: not a mystery about quality, but a consequence of how an index was built and distributed at the time.
Why the Google Dance matters
Mainly as a caution about explanations that outlive their cause. The batch cycle ended when Google moved to continuous indexing with Caffeine, and the dance in its original sense went with it. There is no monthly rebuild to sit out and no scheduled window of chaos to blame.
Rankings still move, of course, and often more than people expect. But the movement now comes from different places: core updates and constant smaller changes, results that differ by location and device, competitors publishing and improving, and Google testing layouts. Calling all that a dance dresses ordinary competition up as a mysterious cycle, and stops anybody investigating it.
Where the Google Dance goes wrong
The first problem is that it becomes an excuse. A drop gets waved away as the dance, nobody looks at the pattern, and by the time someone does the cause is weeks old and harder to trace.
The second is bad measurement mistaken for volatility. A tracker checking from a different location than your customers, a phone result compared against a desktop one, or a single keyword watched every morning will all produce swings that are artefacts rather than changes. Personalised and local results make one reported position far less meaningful than it looks on a chart.
What to do about it
Watch groups, not positions. A set of keywords for one service, tracked from the place your customers actually search, tells you something. One keyword refreshed daily tells you almost nothing. Search Console’s own comparison of clicks and impressions across matched periods is more reliable than any external tool, because it is your data rather than a simulation of a searcher.
Set a review rhythm you can defend — weekly for a look, monthly for decisions — and only investigate a change that holds across a group of pages or queries. When the movement is real, the useful questions are what changed on your site, what changed in the results themselves, and whether an update was announced in the same window. Answering those, rather than naming the wobble, is what honest rank tracking is for.
