Meta Ads: 53% of ad sets target a whole country. Only the trades band mostly went local
Across 23 Australian Meta Ads accounts, 538 ad sets returned a targeting object. 284 of them, 52.8%, target a whole country and nothing finer. The pooled lifetime spend across the 23 accounts is $783,518 AUD. In the health and care services band, where the customer has to physically turn up at the business, 110 of 187 classified ad sets are set to the entire country.
We pulled the geo_locations object from every ad set. Meta stores location as up to five keys: countries, regions, cities, zips and custom_locations (a latitude, a longitude and a radius). An ad set can carry more than one key, so we counted twice. The first count asks which keys are present at all, and it over-counts because of the overlaps. The second classifies each ad set by its finest key, so an ad set with a country plus three cities counts once, as cities.
The two counts agree on the number that matters. 284 ad sets carry a countries key. 284 ad sets carry a countries key and nothing else. Not one ad set in the set combined a country with a finer level; every overlap is among the finer keys.
Geo key present | Ad sets | Share of 538 |
|---|---|---|
| 284 | 52.8% |
| 108 | 20.1% |
| 103 | 19.1% |
| 31 | 5.8% |
| 28 | 5.2% |
The shares sum to more than 100% because of the overlaps. Classified by finest level and split by industry band:
Band | Country only | Regions | Cities | Custom radius or postcode | Classified |
|---|---|---|---|---|---|
Consumer goods & food | 79 | 2 | 21 | 8 | 110 |
Health & care services | 110 | 3 | 22 | 52 | 187 |
Home improvement & trades | 36 | 22 | 32 | 68 | 158 |
Manufacturing & equipment | 45 | 2 | 3 | 0 | 50 |
Professional & B2B services | 14 | 1 | 5 | 6 | 26 |
Total | 284 | 30 | 83 | 134 | 531 |
The row and column totals are ours, summed from the pull's counts. The classification covers 531 of the 538 ad sets; the pull does not place the remaining seven, so we leave them out rather than guess. Of the 549 ad sets the API returned in total, 11 carried no targeting object at all and sit outside both counts.
Read down the country-only column against what each band sells. Manufacturing and equipment: 45 of 50 ad sets national, and that is probably right, because a machine ships. Consumer goods and food: 79 of 110, defensible for anything sold online. Health and care services: 110 of 187. A clinic is a building. Home improvement and trades, the other band where the customer's address decides whether the job can be done at all, is the best-behaved group in the set: 68 of its 158 ad sets use a custom radius or a postcode list, and only 36 go national.
Why the whole country is the default
When you create an ad set in Ads Manager, the location field is pre-filled with the account's country. Leave it alone and the ad set is national. Nothing in the interface flags this as a decision; it looks like a setting that has already been made. The countries key with nothing beside it is what that untouched field becomes in the API.
The second reason is that location is the one part of the targeting object Meta will not expand or correct. Advantage+ audience, which 145 of the 538 ad sets have switched on (part 13 covers the detail), can reach outside the age, gender and interest settings you give it. It treats location as a hard boundary. That is the right design, but it has a consequence: every other targeting mistake has a mechanism that might quietly fix it, and this one does not. A national ad set with Advantage+ audience on is exactly, and permanently, the whole country.
The third is how delivery works inside the boundary. Meta's auction buys the cheapest impressions or results it can find within the constraint. Where those are cheapest depends on population density, competition from other advertisers and the platform's own supply of inventory, none of which has anything to do with where your customers are. A national ad set optimising for leads for a business in one city will find leads wherever leads are cheapest to find.
Key | What it does | Where it goes wrong |
|---|---|---|
| Whole nation, every state and territory | Pre-filled on creation and left untouched |
| A state or territory | Right for a state-wide service, coarse for a metro one |
| A named city plus a radius (Meta applies one by default) | The radius is easy to miss and can exceed the service area |
| A postcode list | Precise, but a long list is tedious to maintain |
| A pin plus a radius in kilometres | The closest match to a real service area; the least used in the set |
Our pull did not capture location_types (whether the ad set reaches people living in a location, recently in it, or both), so this post says nothing about it. It is worth checking alongside the level.
What a national ad set does to a local account's numbers
The first casualty is the lead count. A national ad set for a local business still generates leads. They are real people who filled in a real form. Some of them live in a different state. The ad set reports them as leads, the cost per lead looks acceptable, and the person who phones them back finds out they are a thousand kilometres away. That lead never gets un-counted. Reporting divides spend by the full count, including the ones the business could not serve.
The second casualty is the optimisation signal. Meta optimises the ad set toward whatever the pixel or the instant form records as the lead event, and it has no idea which of those leads were useful. If out-of-area leads are cheaper to acquire, and they often are because a less competitive market has cheaper impressions, the ad set learns to find more of them. The account gets better at the wrong thing. It is the same failure as the inflated denominator in the Google Ads double-counting post: the algorithm is doing its job against a number that does not mean what the label says.
The third is what happens when the national setting stacks with the open demographic settings from part 13. We cannot cross-tabulate the two from the pull, so the table below is separate facts placed side by side, not an intersection.
Setting, across the same 538 ad sets | Ad sets | Share |
|---|---|---|
Country-level location, nothing finer (this post) | 284 | 52.8% |
18-65, all genders, no interests or behaviours (part 13) | 254 | 47.2% |
18-65, all genders, no interests and no custom audience (part 13) | 214 | 39.8% |
Advantage+ audience on (part 13) | 145 | 27.0% |
Somewhere in the overlap of the first two rows are ad sets whose complete targeting is "every adult in Australia". For a consumer product sold online that is a legitimate audience. For a clinic, a tradesperson or a local professional firm it is a budget with no address on it.
Where it varies
Band | Country only | Classified | Share national |
|---|---|---|---|
Manufacturing & equipment | 45 | 50 | 90.0% |
Consumer goods & food | 79 | 110 | 71.8% |
Health & care services | 110 | 187 | 58.8% |
Professional & B2B services | 14 | 26 | 53.8% |
Home improvement & trades | 36 | 158 | 22.8% |
The shares in the last column are ours, derived from the pull's counts; the pull gives the counts only.
The ordering is the interesting part. The two bands with the strongest case for a national audience are at the top. The band with the weakest case, health and care, sits in the middle at 58.8%, and it is also the band that uses custom radii the second most (52 ad sets). That mix inside one band suggests different accounts, or different eras of the same account, working from different assumptions about who the ad set is for. Professional and B2B services is a small group (26 classified ad sets across five accounts) and its 53.8% should be read as a direction, not a rate.
Home improvement and trades is the counter-example. Six accounts, 158 ad sets, 68 of them on a custom radius or postcode list, 22 on regions, 32 on cities. Tradespeople know what their service area is because they drive it every day, and the targeting reflects that. It is the one band where the tool was used the way the business works.
We should be plain about what the pull cannot tell us. It records the geo level of each ad set and the band of each account. It does not record each business's actual service area, so "serves one city" is an inference from the band, not a field we pulled. A health services account with three locations in three states might reasonably run national ad sets. The 110 is still worth explaining, account by account, and we could not find a reading of the band that explains most of it.
What to check in your account
Open Ads Manager at the ad set level, add the breakdown By Delivery, then Region, and sort by amount spent. If you serve one metro area and money is landing in other states, the location setting is the cause.
Repeat the breakdown with Results as the column. Count how many of your reported leads came from regions you do not serve. That count is the amount by which your cost per lead has been flattered.
For every ad set with spend in the last 90 days, open the edit view and read the location field. Write down the level: country, state, city, postcode list, or pin plus radius. Note the radius on any city or pin; Meta applies one whether or not you set it.
Check the location type dropdown under the map. "People living in or recently in this location" is the default and includes visitors. For a business people have to travel to, "People living in this location" is usually the intent.
Before you rebuild a national ad set as a radius, note that a significant targeting edit re-enters the learning phase. Plan for a week or two of unstable cost per lead, and do not judge the new setting on fewer than about 30 leads. The Google Ads sample-size post covers why that floor exists.
If you want a script to do step 3 across every ad set, the field is
targeting.geo_locationson theadsetsedge. Count which keys are present. That is exactly the pull behind this post, and it is the first thing we look at for any business with a service area. When we set up Facebook ads for Brisbane businesses, the targeting starts from the suburbs they actually serve, not the whole state.
The principle
Location is the only targeting setting that describes the business rather than the customer. Age, gender and interests are guesses about who might buy; the service area is a fact about who can. Meta will expand every guess for you and none of the facts. Set the fact yourself, at the level the business actually operates, and read the region breakdown often enough to notice when it drifts. A lead the business cannot serve is not a cheap lead. It is a cost with a name attached.




