Everyone in local marketing repeats the same claim: a radius wastes budget, fitting the real zone saves it. We wanted the number, so we measured it. 970 delivery zones built from official Australian suburb boundaries, every state and territory, four targeting methods each, measured with the same engine that plans production campaigns.
On the median zone, a circle drawn to cover the whole delivery area puts 65.4% of itself outside it. A default 3 km radius puts 67% outside. A set of circles fitted to the zone puts 12.5% outside on Meta and 11.6% on Google. Put the other way round: fitted targeting is 87.5% precise on Meta and 88.4% on Google, against 34.6% for one covering circle and 33% for a default 3 km radius. That is roughly two and a half times more of the targeted area landing inside the zone, and five times less of it landing where the store cannot sell. The pattern holds in every state we measured.
The radius that covers everything is the radius that buys the most ground the store does not serve. Median precision 34.6%, and half the zones sit between 27% and 43%. Full coverage, mostly waste.
The common default does two things wrong at once. Its median precision is 33%, and on 223 of the 970 zones it also failed to cover at least 13% of the area, because real zones are not centred on the store.
A median of 4 include circles, with cut-outs where they earn their keep, covering 92.5% of the zone with 87.5% of the targeted area inside it. Half the zones fit between 82% and 92% precision.
Google cannot cut circles back out, so its plans use a median of 5 include circles, cover 90.9% of the zone and keep 88.4% of the targeted area inside it. Less coverage, less spill, a different trade.
| Method | Covers the zone (median) | Outside the zone (median) | Circles used (median) |
|---|---|---|---|
| One covering circle | 100% by construction | 65.4% of the targeted area | 1 |
| Fixed 3 km radius | 100% on small zones, gaps on large ones | 67% of the targeted area | 1 |
| Fitted, Meta | 92.5% | 12.5% of the targeted area | 4 include, exclusions where useful |
| Fitted, Google | 90.9% | 11.6% of the targeted area | 5 include |
Swipe the table sideways to see the rest.
The two fitted rows are not at 100% coverage, and that is the honest part of the result: coverage and precision pull against each other, and covering the last awkward corner means buying ground outside the zone. The engine exposes that trade as a setting, so a network can lean towards protecting reach or protecting budget. What no setting can fix is the first two rows.
A national median can hide a result that only holds in one place, so here is every state and territory separately. The covering circle wastes between 63% and 68% of its area everywhere, and the fitted plans land between 10% and 16%. The gap never closes, which is the point: this is geometry, not a quirk of one city.
| State | Zones | Covering circle, outside | Meta fit, outside | Google fit, outside |
|---|---|---|---|---|
| New South Wales | 150 | 67.9% | 14.2% | 13.8% |
| Victoria | 150 | 66.0% | 11.1% | 9.9% |
| Queensland | 150 | 63.9% | 11.5% | 10.4% |
| Western Australia | 150 | 67.0% | 12.4% | 11.5% |
| South Australia | 150 | 66.2% | 12.0% | 11.7% |
| Tasmania | 100 | 65.2% | 12.7% | 12.5% |
| ACT | 60 | 63.4% | 15.4% | 15.1% |
| Northern Territory | 60 | 63.6% | 10.2% | 11.5% |
Swipe the table sideways to see the rest.
Two caveats worth stating rather than burying. The ACT and the Northern Territory carry 60 zones each and Tasmania 100, not 150, because those places simply do not have 150 metro-scale suburbs to draw from; treat those three rows as indicative rather than settled. And the ACT fit is the loosest in the country at 15.4% outside, which fits its geography: districts wrapped in reserves are harder to cover with circles than a continuous suburban grid. See franchise marketing in Canberra for what that means in practice.
Each zone is a union of 1 to 4 adjacent real suburbs from official Australian boundary data, every state and territory, sized 3 to 30 km², which is how real networks draw delivery and service areas. No client files are used anywhere.
The fits come from the same solver that plans live campaigns in Amplaro: greedy max-coverage circle fitting, separate plans for Meta (include plus exclude) and Google (include only), platform minimum radius of 1 km respected.
Seeded zone construction, sorted inputs, 150 zones per state. Every per-zone result is kept, so any row can be checked on request. The fits are consistent with what the engine reports on real territory files in production use.
One honest caveat: precision here is a share of targeted area, and money follows area only approximately, because people are not spread evenly. The exact spend share differs per zone. The direction does not: for a delivery business, the area outside the zone converts at zero, whatever it cost to reach. The reasoning is on radius vs delivery zone, and the translation method on turning a zone into an ad audience.
Two thirds of the targeted area, measured. Spend follows area approximately, not exactly. For a delivery business the direction is not in doubt: the outside share cannot produce an order at any conversion rate.
Because the last corner of an awkward zone costs more outside-spend than it is worth. The balance is a setting per network, not a fixed answer.
Different rules. Meta can cut circles back out, Google cannot, so the plans are solved separately. Google kept slightly more spend inside the zone while covering slightly less of it.
Yes. Seeded construction, public boundary data, method above. The per-zone results are kept; ask us and we will share the table and look at any difference together.
The reasoning behind these numbers: what a circle cannot know about a real trade area.
The product page: the same engine, fitting every store's campaigns to its own zone automatically.
Where targeting shows up in the return, and how to measure the rest of it properly.
The benchmark says what happens on typical zones. Your network's number depends on your shapes, and measuring them takes one demo.