GSC average position for rank tracking: good enough or misleading?

Plenty of teams use the average position in Search Console as their rank tracker, and I don't think it holds up well. It's weighted by impressions, so a page that shows at position 3 for a query it rarely appears for can look identical to one that sits steadily at 8. Position also shifts by device and location, and a query with two impressions a week tells you almost nothing.

For a stable set of money terms with decent volume, it works fine as a directional trend line. But if you're trying to see whether a specific content change moved a keyword, use a rank tracker with fixed locations and devices, or at least read clicks and impressions alongside position. How are others handling the low-impression queries? That's where I'd expect GSC to fool people most.

Same trap shows up on big terms too, because the number depends on which filter you're looking at. With no query filter on the Performance report, a page's position is averaged across every query it appeared for. So a product page that ranks 2 for its main term and 40 for a dozen long-tail variants can show an average of 9, and it looks like it dropped when it didn't.

Filter by query and page together and you get the number that actually reflects that keyword. Do that before judging any content change. And give it a couple of days for the data to catch up, since a change made on Tuesday won't show cleanly until later in the week.

Agreed, and local service sites make it messier. For a clinic or plumber, "near me" and city queries often trigger a map pack, and as far as I know GSC doesn't separate those placements from regular organic positions. So one average can blend a map listing with a blue link. It also filters by country but not by city, which means you can't isolate what someone in your service area actually sees. A rank tracker set to your town's coordinates will tell you more than GSC does here.

Agreed, the map pack is where this bites hardest for local businesses. A single average from GSC can't show you that a plumber ranks well for "plumber" from the shop's street and falls off a cliff three kilometres away. Local rank trackers often use a geo-grid, which samples a set of points in a ring or square around the business and records the position from each one. That shows you where you appear and where you don't, which is what a local business actually cares about. GSC gives you one number for the whole country, and if most of your searches come from a 5 km radius, that number is mostly noise. So for that kind of business, I'd use GSC to see which non-local queries are bringing in traffic and use the grid for the map pack.

Geo-grids have a catch: each point is a simulated search from those exact coordinates, so you're seeing rank for that spot, not what a real customer with their own search history would get. And set the radius from where your bookings or calls actually come from, not from a guess, or you'll end up sampling empty ground.

Agreed on the simulated search point. Grid density matters about as much as radius. A 7 by 7 grid across a 5 km radius puts points roughly 1.7 km apart, and map pack results can change between two streets that close. You could see a plumber ranking from the shop and still miss the stretch where they drop out completely. For a tight service area, a denser grid at the same radius catches more of that edge.

I'd take Jun's radius advice a step further too. Pull it from where your calls and bookings actually come from, by plotting the postcodes or addresses in your records, not from a circle that just looks right on a map.

The 1.7 km spacing only holds if the grid spans the full diameter. If the tool builds a square around the radius instead, the corners land about 7 km out, so you end up sampling well past the area you meant to cover.

Corner points are the sneaky part, because the map looks just as tidy there as everywhere else, so nobody questions them. A plumber can rank 4 in most squares and still lose a corner to a competitor from the next town. That one dropped point makes the whole grid look like a problem in your own area, when the real issue is a different market.

If the tool lets you pick the shape, use a circle sized to your real service radius. If it only builds squares, check the coordinates it generates before you trust the results, and drop any point beyond the distance you actually serve. You'll end up with fewer samples, but each one comes from somewhere a real customer would actually search.

Dropping the far points is fine, but do it once and then lock that exact point set for every later run. If you trim the grid differently each month, a change in your score can come from the sampling instead of the rankings. Say a plumber's grid goes from 49 points to 38 after a cleanup and the average jumps from 3.1 to 4.2. It looks like a collapse, but it's the same market with the outlying corners gone. Save the point list with each report so the next comparison is like for like.

Locking the point set is the right call, and I'd turn it into a simple rule: don't compare two grid runs unless the center, radius, grid density, device and date window all match. If any of them changed, start a new baseline instead of reading the difference as a trend.

The average is the number most likely to move for the wrong reason. Dropping ten corner points can shift it by a full position without a single ranking changing. The share of points ranking in the top three is usually steadier to track alongside it, since a few outliers can't distort it as easily.