Geofencing Marketing: What It Actually Is, and Whether It Is Worth It

Geofencing is the most oversold tactic in local marketing. It is also, in a narrow set of situations, genuinely useful. The problem is that almost everything written about it is published by companies selling it.

We get asked about geofencing constantly, usually after a cold call promising to put ads on the phone of anyone who walks into a competitor's shop. So we went and read the independent research rather than the sales decks, and the picture is a lot less flattering than the pitch.

This guide covers what geofencing actually is, why Google and Meta cannot really do it, what the peer-reviewed evidence says, what it costs, and the privacy law that has quietly made large parts of it unusable. Then an honest verdict on when it is worth your money.

What geofencing actually is, and the three things it gets confused with

There is no standards body for these terms, so vendors define them to suit whatever they are selling. Here is the most defensible reading of how they are actually used.

This matters commercially, because most products sold to small businesses as "geofencing" are actually geotargeting or geoframing. Platforms that genuinely sell it, like Choozle and Feathr, describe historical geofencing as the main deliverable, not live boundary crossing. If a salesperson is describing real-time alerts as someone walks past your shop, ask which of the four they are actually selling.

How it technically works

Geofencing rides on a chain of signals, and the chain is only as good as its weakest link.

• GPS is the only signal precise enough for a genuine store-level fence. The US government puts a GPS-enabled smartphone at roughly a 4.9 metre radius under open sky, and warns that accuracy worsens near buildings, bridges and trees.
• Wi-Fi lookups give tens of metres, and are the main source indoors.
• Cell tower triangulation gives hundreds of metres to kilometres. Any vendor whose accuracy relies on cell towers is not doing store-level anything.
• Bluetooth beacons get under ten metres, but need hardware in the venue and an app on the customer's phone that is listening for it. Effectively unavailable to a small business.
• IP address is the fallback, and it is city-level at best.

The ad itself is served when an app with a location-permissioned SDK sends a latitude, longitude and mobile advertising ID into a real-time bidding request. Which means geofencing only ever reaches people who are inside an app that collects location, who granted that permission, at the moment the app is open. That is a much smaller and stranger slice of the public than "everyone who walks past".

The accuracy problem nobody selling it mentions

This is the part that changed our view of the channel entirely.

The location data flowing through ad exchanges is, by the industry's own admission, largely unreliable. Foursquare has said close to 80% of the location data in the bid stream is inaccurate. In a widely read Digiday interview, a location-data executive described an agency network leader believing up to 80% or more of available latitude and longitude data is fake, and said vendors pad limited legitimate data with datasets from competitors or unknown sources.

There is a straightforward financial reason for it. App publishers can earn a large multiple on their ad yield by passing location data along with a bid request, which is a direct incentive to append coordinates whether or not they are real.

Why Google and Meta cannot actually geofence

The two platforms most local businesses already use are structurally incapable of the thing being sold to them.

Google Ads offers radius targeting, which is proximity, not a fence. There is no device-level entry trigger, no dwell time and no custom polygon. More importantly there is a hard floor: the minimum radius is 1 km, and Google's Display and Video 360 documentation states the constraint explicitly as a minimum one kilometre radius that must contain at least 1,000 people. That is a deliberate privacy floor, and it makes fencing a single building impossible inside Google's stack.

Meta has a minimum radius of one mile, or one kilometre outside the US. You can fence a neighbourhood. You cannot fence a competitor's front door.

Genuine polygon and device-ID work lives in specialist demand-side platforms. Which means when someone offers you geofencing, they are either reselling a specialist platform, or quietly selling you radius targeting you could set up yourself in Google Ads in two minutes.

What it costs, and what you actually get

The money is not usually the blocker. The reachable audience is.

• Display geofencing typically runs at a $6 to $15 CPM, with connected TV considerably higher at $20 to $50. These are vendor-published ranges, so treat them as indicative.
• Some self-serve platforms genuinely have no minimum spend. Others want ten to twenty thousand dollars a month.
• The real constraint is audience size. Feathr, a platform that actually sells historical geofencing, advises that campaigns need to identify at least 50,000 unique mobile devices to work. A single competitor's premises in a small town will never produce that.

Do the arithmetic for a typical local business. At a ten dollar CPM, a thousand pounds a month buys you roughly a hundred thousand display impressions spread across a small geography, aimed at people who are not searching for you and have shown no intent. At realistic display click rates that is a few hundred clicks. Next to search demand from people actively looking for a plumber right now, it is a rounding error.

Does it actually work? What the independent research says

There are exactly two rigorous studies worth knowing about, and they point in opposite directions.

The case for it

A randomised field experiment published in the Journal of Marketing Research in 2015 found that targeting promotions at a competitor's location outperformed targeting your own, because promoting to people already at your own location largely discounts sales you would have made anyway. That is a real mechanism, and it is the strongest evidence geoconquesting has.

Two caveats. It is from 2015, before the mobile privacy changes described below, and it was run by sending promotions to an opt-in customer base, not by buying programmatic bid-stream inventory. It validates the idea, not the product being sold today.

The case against it

A 2024 study in Information Systems Research, summarised by UT Dallas, modelled what happens when everyone does it. The conclusion is blunt. Co-author Dr Amit Mehra states that "businesses would be better off by not engaging in geoconquesting campaigns", but that each firm is worse off if it abstains while rivals participate. It is a prisoner's dilemma that escalates into discount wars, and the study concludes that advertising agencies capture most of the surplus, not the advertisers.

And the vendor case studies

We collected several deliberately. One vendor reports a click-through rate of 0.42%, another 0.27%, another claims 4.2%, all describing the same channel. That is a fifteen-fold spread. Not one of the case studies we found disclosed a control group, which means none of them demonstrate that the ads caused anything.

The law has closed in, and it is still moving

If the evidence does not put you off, the legal exposure might. This has changed enormously in the last two years.

In the United States, the Federal Trade Commission has taken action against a series of location data brokers, banning several from selling precise location data. In its case against Gravy Analytics and Venntel, the FTC alleged the company used geofencing to build lists of people who attended locations related to medical conditions and places of worship. In its parallel case against Mobilewalla, the FTC alleged the company harvested location data out of ad auctions for purposes other than bidding, the first time it treated that practice as unfair in itself.

Several states have gone further and banned geofencing outright in sensitive contexts. Washington, New York and Connecticut all prohibit geofences around health care facilities. California's AB 45 came into force on 1 January 2026, prohibiting geofencing around in-person health care facilities to track, identify or advertise to people seeking care, with a private right of action and civil penalties of up to twenty five thousand dollars per violation.

In the UK and EU, location data is personal data, and where collection involves storing or accessing information on a device the rules require GDPR-standard consent rather than legitimate interests. The Information Commissioner's Office has been clear that it is difficult to justify location tracking on legitimate interests, because it is unlikely to be within people's reasonable expectations. In practice you are relying on a chain of app publishers having obtained valid consent, and you cannot audit that chain.

Finally, the raw material has been shrinking. Apple's App Tracking Transparency has substantially reduced the availability of the iPhone advertising identifier since 2021. On the Android side, Google retired most of the Privacy Sandbox initiative in October 2025, which means the Android advertising ID survives for now. That is a reprieve rather than a resolution, and it leaves geofencing skewed heavily towards Android and towards app inventory.

Our honest verdict

Geofencing is not a scam. It is a real technique with a narrow set of legitimate uses, sold far outside them.

When it is genuinely worth considering

• High-value, infrequent purchases where visiting a competitor is a strong intent signal. Car dealerships, furniture retail, gyms in January. This is the case the 2015 field experiment actually supports.
• Events with dense, time-bounded footfall. Trade shows, stadiums, conferences. These are the only situations that reliably produce enough unique devices for the maths to work.
• Multi-location businesses with the budget to run a proper holdout area and measure lift.
• Address-based household targeting, which is a different and more deterministic product that happens to be sold under the same label.

When it is a bad buy

• Any single-location service business on a modest budget. The same money in search, against people actively looking for you, will beat it comfortably.
• Where the fence is bigger than the intent. If your platform's minimum is a kilometre or a mile, you are buying a neighbourhood and calling it a fence.
• Where the vendor will not run a holdout. No control group means no evidence, only a bill.
• Geoconquesting as a permanent strategy, given what the 2024 modelling says about who actually profits.
• Anything health-adjacent in the United States, where the legal exposure is now severe.

What we would do instead

Before spending anything on geofencing, there is a free change that almost always matters more.

In most local Google Ads accounts we audit, the location setting is wrong, and the account is paying to show ads to people who merely expressed interest in the area rather than being in it. Fixing that takes about ninety seconds and costs nothing. We walk through it in our guide to Google Ads radius and location targeting.

After that, the honest hierarchy for a local business is search demand first, then your local search and map pack visibility, then reviews, then paid social for demand creation. Geofencing sits well below all of those, and only earns a place when one of the narrow cases above genuinely applies.

If you want a second opinion on where your budget is actually going, that is what our Google Ads management audit is for, and it is free.