Google's local search uses a click radius. Here's what the leaked code tells us about how it works.

NavBoost has been the most discussed part of the leaked documents. The general consensus is that Google tracks Chrome browser interactions and uses them as a ranking signal. Rand Fishkin and others have written about this extensively.

What hasn't gotten much attention is how NavBoost works for local search specifically. We've been going through the proto definitions (two of us now - I roped a friend into this project after the review velocity post) and found a subset that references geographic coordinates alongside click metrics.

// navboost/local_click_signals.proto
message LocalClickRadius {
  double center_lat = 1;
  double center_lng = 2;
  float radius_meters = 3;
  float in_radius_weight = 4;
  float out_radius_weight = 5;
  enum DensityClass {
    URBAN = 0;
    SUBURBAN = 1;
    RURAL = 2;
  }
  DensityClass density_class = 6;
}

There's a radius - measured in meters - around each business entity's listed location. Clicks from devices within this radius get one weight (in_radius_weight), clicks from outside get a different one (out_radius_weight). Two separate fields, which means they're almost certainly weighted differently.

The radius isn't fixed. There's a DensityClass enum right there in the same message: urban, suburban, rural. Which means a dentist in suburban Phoenix probably has a different effective click radius than a restaurant in downtown Chicago. This actually makes sense - you wouldn't want the same radius for a Manhattan coffee shop and a rural auto mechanic serving a 40-mile area.

Why this matters

If the weights are significantly different - and the fact that they're stored separately suggests they are - then where your searchers are physically located when they click on your listing matters. A lot of local search clicks come from people at home or at work, not standing outside the business. If someone is searching from 15 miles away, their click might count for less than someone searching from 2 miles away.

This has implications for how you think about your service area vs. your physical location. If you're a plumber and all your customers are in the suburbs but your office address is across town in an industrial park, you might be accumulating clicks from outside your effective radius.

We don't know the actual radius values or the weight ratios. The proto definitions give us the shape of the data but not the tuning. Could be a small difference, could be a large one.

Related: dwell signals inside the radius

There's a related structure that seems to track what happens after the click, but only for interactions originating inside the radius. Fields that look like dwell time and direction-request tracking. We're still working through those and don't want to speculate too much before we understand the full picture.

Separate post on that when we've got more.