Choosing a grid size and reading the heatmap
You now know what a grid is and what its two parameters do. The next two questions are which grid to run, and how to read what comes back without flattering yourself.
Resolution is a statistical choice, not a budget one
Grid size is usually taught as a spending decision: small grid cheap, big grid expensive, choose by the client's budget. That framing is mostly wrong, and it is worth understanding why before you form the habit.
A rank measurement needs exactly one thing from Google: which places came back, in what order. Not their ratings, not their phone numbers, not their hours — you are looking for your own entry in a list and reading off its index.
Google prices description, not results. It charges a place search by how much it tells you about each result, not by how many results come back. Ask for the rich profile of every result and you are in an expensive tier; ask for less and you are in a cheaper one.
What does not change is the result set or its order — the field mask governs how much each result is described, not which results Google returns. The position is the index either way.
So the search half of a geo-grid scan has a wholesale cost of zero, at any grid size, at any volume. What remains genuinely chargeable is drawing the result on a real Google map, which happens once per scan and does not depend on n. The derivation — which requested fields push a call into which billed tier, with the tier prices — is What the Places API will and will not give you.
The consequence is not "grids are free" — every tool prices its own work, this one included, and you should read the price on the button before pressing it. The consequence is about reasoning:
Choose the grid for the question, not for the invoice. If the question is "do I hold my immediate neighbourhood", 3×3 answers it and 7×7 adds noise. If the question is "where does my visibility stop", 3×3 cannot answer it at any price, because the answer is outside the grid.
Reading your first heatmap
Read the shape first. Read the numbers second, and read them suspiciously.
The legend
Pins are coloured by position. The legend under the map carries five swatches:
- Green — top 3
- Yellow — 4–10
- Orange — 11–20
- Red — 20+
- Grey — not found
The red one is vestigial. A grid reads exactly twenty deep, so a point either comes back with a position of 20 or better or comes back with nothing, and in a real scan you will only ever see the other four. The distinction that matters is coloured versus grey.
Five shapes
Five shapes cover most real grids. The causes below are inference from repeated reading, not documented behaviour.
| Shape | What it looks like | Usual cause |
|---|---|---|
| Smooth decay | Green centre fading outward in rings | Healthy and proximity-dominated. The normal case. |
| Plateau | The same colour nearly everywhere | Strong prominence across the area, or a thin market. Check which. |
| Cliff | Good on one side, grey on the other, sharp edge | A boundary: competitor cluster, city line, river or motorway. |
| Ring | Weak at the centre, better further out | A dominant near neighbour sitting on top of you, or a centre point that is not really your address. |
| Scatter | No spatial structure at all | Low prominence. Position is decided by tie-breaks, not geography. |
The two readings around the map
Above the map: a written second opinion. The app prints a plain-language reading of the same thing — coverage, the average, the found rate ("you appear in the top 20 at 9 of 9 points"), and the compass direction you are strongest and weakest in ("strongest to the west and weakest to the south-east").
Treat that as a second opinion to check your own reading against. It only claims a direction when the gap between the best and worst side is large enough to mean something, so a genuine plateau produces no direction claim at all.
Beneath the map sit the two summary figures, which have different denominators and are routinely misread because of it:
- Avg rank — your average position across the points where you appear. Grey pins are not in it.
- Top-3 coverage — the share of all points where you rank in the top 3.
A business found at 2 of 25 points, at #1 in both, shows an average rank of 1.0 and top-3 coverage of 8%. The first number is true and nearly useless.
Never quote the average without the found rate beside it — how many of the points you appeared at, out of how many were checked.
Two properties to carry forward
Part III makes both of these rigorous. Learn them now as habits.
Rank is ordinal
Position is a label of order, not a measured quantity. The step from #1 to #2 is enormous — the map pack shows three results, so it moves you down inside the only box anyone sees. The step from #11 to #12 is nothing; both are invisible. Arithmetic treats those two steps as identical, which is why "improved by 4.2 positions on average" sounds like a result and means very little.
You are allowed to do arithmetic on ordinals. You are not allowed to believe the answer without saying what you did.
Not found is censored, not bad
A grey pin does not mean rank 21. It does not mean 25, or 50, or 100. It means the measurement ran out of depth before it found you: your true position is somewhere worse than 20, and the instrument cannot say where.
Statisticians call this right-censoring. It matters because every average over a grid must do something with those pins, and whatever it does is a choice that moves the result — Reading a geo-grid without fooling yourself has you score one grid four ways and watch the number travel. For now, one rule: when you report a grid, give the found rate alongside anything else, and never let a grey pin quietly become a number.