Most local SEOadvice is a survey.Ours is a measurement.
The numbers this industry quotes come from asking practitioners to rank factors by importance — which measures what experts believe, and belief comes from the same folklore the survey was meant to settle. We run the searches instead — tens of thousands of them — and report what came back, including the parts that went against what we expected to find.
Three rules, and the third one costs us.
These are the rules a study has to clear before it gets a URL here. Two of them are ordinary method. The third is the one most companies quietly skip.
Design out the confound
The businesses closest to you are not a random sample — established, well-reviewed ones sit in central locations. No statistic separates two causes that never move independently, so the design has to separate them first.
Validate the instrument first
A data provider can return a plausible, entirely non-local answer and never raise an error. Every study opens by proving the instrument varies with the thing it claims to measure — before a single finding is read off it.
Reviews lift rank across the board
Raw correlation was −0.095. It only appears once you hold distance still.
Publish the disconfirmations
Our first study registered a hypothesis the raw data did not support. It is written up as disconfirmed rather than quietly reframed, because a research page that has never been wrong is not a research page.
What we have measured so far
Each one states its sample, its method and its limits on the page — including the hypotheses the data refused to support, which is the part a survey has no way of telling you.