Geofence Radius and Audience Best Practices

Geofence radius and audience settings decide most of a geofencing campaign's results. Trace the fence to the place that signals intent, check it against location precision and platform minimums, and set two time windows: when a visit counts and how long visitors stay in the audience. Then screen out sensitive places and plan how you'll measure before launch. This guide is part of our paid advertising guide.

You'll need a clear campaign goal, a map tool that can draw shapes, and the current location rules for your ad platform or geofencing partner. Conversion tracking on your website makes step 8 possible.

What makes a geofence work

A geofence works when being inside it means something. A trade show hall during show hours tells you people are interested in that industry. A busy intersection tells you very little. So the first rule is about the place, not the size.

The second rule is about precision. A fence can only match phones that share a precise location, and every reading carries some error. Google's ads help says GPS accuracy varies with signal, and Wi-Fi accuracy is roughly a router's range. The OpenRTB standard has each reading report its own accuracy estimate, in meters. A fence much smaller than that error can't reliably tell inside from outside.

The third rule is about time. A geofence is two windows: when a visit counts, and how long the visitor stays in your audience afterward. We explain the full chain in how geofencing advertising works.

Steps

1. 01
   Write down the goal and the visit that signals it.

   Name the action you want, such as a booked appointment, a showroom visit or a quote request. Then name the visit that suggests someone might take it. "People who attended the home show" is a usable signal. "People near our office" usually isn't. If you can't link the place to the goal in one sentence, use ordinary area targeting instead.

2. 02
   Choose places you can justify, and screen out sensitive ones.

   List each place and why it qualifies. Remove any that reveal sensitive details about visitors. The FTC lists places such as medical facilities, places of worship, shelters and military sites, and says businesses should in most cases not use that data at all. Check category rules too: Google bans ZIP-code targeting for housing, employment and consumer finance ads in the US and Canada, and Meta has its own limits.

3. 03
   Trace the place itself, not a circle around it.

   Draw a polygon that follows the building, venue or grounds, if your platform supports it. A circle around a pin usually spills onto roads, car parks and neighbouring businesses. Leave out shared spaces unless being there means a visit. For a multi-building site, draw one shape per building rather than one big shape over everything.

4. 04
   Check the fence against precision and minimum sizes.

   Ask your partner which location sources they accept and whether they filter out readings with poor accuracy. Then check platform minimums. Google Ads won't accept a radius under 1 km and enforces minimum area and user counts. It also warns that small targets may serve intermittently or not at all. A small fence on a platform built for wide targets may never deliver.

5. 05
   Set the visit window.

   Decide which dates and hours count as a visit. For an event, that's show days and opening hours, so set-up crews and next-door visitors on other days are left out. For your own location, it's opening hours. Write the window down with the fence, so anyone reviewing the campaign can see exactly what counted.

6. 06
   Set how long devices stay in the audience.

   This is the lookback or audience duration. Match it to how long your buyers take to decide. A food or event offer may need days. A boat, a home project or a medical procedure may need weeks or months. Start with your normal sales cycle, then shorten it if late impressions don't produce results.

7. 07
   Check the audience size before launch.

   Ask the platform or partner for an estimate of how many devices match. If it's too small to serve, widen the visit window or add qualifying places before you enlarge the fence. A bigger fence adds people who never visited, while more qualifying visits add the right people.

8. 08
   Set up measurement and a comparison group.

   Make sure conversion tracking works before spend starts. Agree what counts as success, such as calls, forms, site visits or store visits where supported. Where your partner offers it, keep a holdout: a share of the audience that sees no ads. Comparing the two groups shows what the ads changed, not just who converted anyway.

9. 09
   Launch, check delivery and adjust.

   In the first days, check that impressions are serving, that they come from the places you fenced, and that frequency isn't piling onto a few devices. Fix fences that serve nothing or far too much. Then leave settings alone long enough to read results before changing the fence, the windows or the creative.

If you'd rather have the fences planned and the campaign run for you, our in-house ads team handles geofencing alongside the ad creative.

Choosing a fence shape and size

The shape follows the place. These hypothetical examples show how the same logic gives different fences.

Writing a fence plan

Before launch, write each fence down in a short record. It keeps decisions visible and makes results easier to read later. A useful record covers the place and why it qualifies, the shape and how it was drawn, and the visit window. It also notes the audience duration, the sensitive-place check, the platform rules checked and their date, and the success metric.

The record also helps when you add another channel. A fenced audience can follow up on phones and, with some partners, on streaming TV in the same household. If that's your next step, start with OTT advertising.

Troubleshooting

The audience is too small to serve

Cause: the fence is small, the visit window is short, or few visitors share precise location with apps. Platforms also apply minimum area and user-count thresholds. Google, for instance, may show ads intermittently or not at all for very small targets. Partners may filter out readings with poor accuracy, which shrinks the count further.

Fix: widen the visit window first, for example to include every show day, or add other places that qualify for the same reason. Lengthen the audience duration if your sales cycle allows it. Only then enlarge the fence, and only to the edge of the qualifying place. If the audience is still too small, switch to area targeting for this goal.

Ads seem to reach people who never visited

Cause: the fence spills onto roads, shared parking or neighbouring businesses, or the partner accepts imprecise readings, such as IP-based locations. Google notes that its location targeting is a best effort and isn't 100% accurate, and the same is true of any location data. Very old readings can also place a device where it no longer is.

Fix: redraw the fence as a polygon on the place itself and remove shared spaces. Ask your partner which location types they accept, whether they filter by the accuracy estimate and how old a reading may be. Check placement and location reports where available. Accept that some drift will remain, and judge the campaign by results rather than by audience purity.

Results can't be separated from other campaigns

Cause: the fenced audience overlaps with search, social or retargeting audiences, so the same people see several campaigns. Conversions get claimed by more than one platform, and no one can say what the geofence added. Missing conversion tracking makes this worse.

Fix: decide up front which conversion actions you'll count and where. Use a holdout group if your partner supports it, and compare people who saw the fenced ads with similar people who didn't. Track branded searches, direct visits and calls during the campaign as well as last-click results. Read results over the full audience duration, not just the first week.