Geofencing Is Not a Strategy. It Is a Radius.

Somewhere right now a media plan is being approved because it contains the word geofencing, and what the word means in practice is a five-mile circle drawn around a store. Circles are not strategies. A radius around your address captures whoever happens to stand inside it: commuters passing through, residents who will never be customers, and the parking lot of the office park next door. It is targeting geometry and calling it targeting people.

Run the arithmetic on the circle and the problem stops being abstract. A five-mile radius around a suburban store covers roughly seventy-eight square miles. Depending on the market, that is anywhere from fifty thousand to half a million devices, and the fraction of them with any plausible relationship to your business is a rounding error. You are paying to reach the other ninety-something percent because they happen to share a zip code with your parking lot. No one would approve an audience defined as "people who live sort of near us" if it were written that way on the plan. Draw it as a circle on a map and it sails through.

I have bought location-based media through GroundTruth, Simpli.fi, and their peers for years, across retail, dealerships, and DC advocacy work, and the useful version of this channel looks nothing like the circle. Location is behavioral data. Where a device goes, repeatedly, over time, says more about the person carrying it than any declared interest category: the gym membership that is real, the dealership lot visited twice in a week, the competitor's store they shop every Saturday. Facebook can tell you what someone claims to like. A location graph tells you where they actually stood, and for how long, and whether they came back. The craft is in which locations you fence and what a visit means, not the size of the circle.

The three questions that separate targeting from geometry

Before I approve any location buy, the plan has to answer three questions, and most cannot survive the first one.

First: what does presence at this location tell us about intent? A device at a Ford dealership on a Tuesday afternoon is a car shopper or a service customer, both valuable, both nameable. A device within five miles of a Ford dealership is a suburb. Fence behaviors, not just your own address: competitor conquesting, category venues, the places your customer's day actually includes. A residential real estate client of mine got more from fencing open houses in target neighborhoods than from any radius around their own office, because attendance at an open house is a declaration and proximity to an office is a coincidence.

Second: how fresh is the signal? Recency is most of the value. Someone at a car lot yesterday is a prospect. Someone there in October was killing time between errands, and the vendor charging you to reach them in March is selling you a memory. Set lookback windows by purchase cycle, days for retail and dining, weeks for automotive, and refuse the default ninety-day audience that exists because it makes the vendor's reach numbers look healthy.

Third: what happens if it works? The entire premise of the channel is that the conversion happens on foot, so the measurement has to happen on foot too. Visit lift against a holdout, not clicks. Nobody clicks a mobile banner and drives to a store because of it in a way last-click reporting can see. The platforms will sell you store-visit studies; make them show the control group, the confidence interval, and the raw visit counts behind the percentage. A vendor who will not show the control group is telling you what the study would say if they did.

The part the case studies skip

Here is what almost never makes it into the vendor deck: the accuracy of the underlying location data varies wildly, and you are allowed to test it. GPS-derived signals from apps with legitimate location permission are decent. Signals inferred from IP addresses and wifi triangulation are guesses wearing coordinates. Ask any location vendor what share of their supply is deterministic GPS versus inferred, and watch the meeting change temperature. I have run the same conquesting concept through two vendors in the same market and seen visit-lift results that disagreed by a factor of four, which is not a difference in campaign quality. It is a difference in whether the devices were ever really there.

One more thing the radius crowd skips: sensitive categories have rules. Health-adjacent locations carry legal restrictions on this exact technique in several states, Washington's My Health My Data Act being the loudest example, and location data is precisely the kind of signal regulators now read as identity. A fence around a clinic, a place of worship, or a union hall is not a targeting decision anymore. It is a legal exposure with a media plan attached. Fence like a lawyer is watching. Increasingly, one is.

The channel is real. Physical behavior is one of the few signals left that cannot be faked by a bot or flattered by an attribution model, which is precisely why it deserves better than a circle. If your location vendor's proposal fits on a map instead of in an argument, send it back.