How to Prove an OOH Campaign Drove Store Visits

A billboard does not get clicked. That has always been the hard part of selling and buying out-of-home: the medium reaches people in the real world, and the result, if there is one, also happens in the real world. Visit attribution tries to close that gap by asking a simple question. Did people who were exposed to the campaign go to the store more often than similar people who were not?

This guide is for Canadian agency planners, OOH and DOOH media owners, and brand marketers. It explains how OOH is measured in Canada today, how a visit-lift study works, what can go wrong, and what to look for in a vendor's report.

How OOH audiences are measured in Canada

In Canada, the industry's measurement body is COMMB, the Canadian Out-of-Home Marketing and Measurement Bureau. It describes itself as the national not-for-profit organization for the Canadian OOH industry, and it develops audience measurement methods for different OOH formats ¹. Its methods go through a Research Committee of buy-side and sell-side experts, separate from its board, whose job is to make sure the approaches are statistically sound ¹.

COMMB's planning platform is called ROADMAP. In May 2025 it added an enhanced reach and frequency tool that reports down to individual billboard assets, after about five years of development funded by its member media owners ².

That measurement answers an important question: how many people, and which people, had the chance to see a campaign. It is the currency for planning and buying. COMMB describes its work in terms of audiences, reach and frequency ¹², not what people did after seeing an ad. That is the job of an outcome study such as visit attribution, which sits on top of audience measurement rather than replacing it.

What visit attribution is

Visit attribution, often called footfall attribution, links exposure to an ad with later visits to a physical location. In OOH, exposure is usually inferred from location data: a mobile device was observed near a screen or board while the campaign was running. Visits are inferred the same way: the device was later observed at the store.

Both steps need clear definitions. The Media Rating Council's location-based measurement guidelines, developed with the IAB and the Mobile Marketing Association, define a visit as a device's location falling inside a place's boundary, and they ask vendors to filter out drive-bys and other brief proximity that does not amount to a visit ³. On exposure, the Out of Home Advertising Association of America's 2021 DOOH guidelines recommend that for outdoor screens, a device should be within viewing distance while an ad plays and travelling towards the screen ⁴. A north-facing screen is seen by people heading south.

If a vendor cannot tell you how it defines exposure and a visit, you cannot judge its results.

Exposed versus control: the core design

A visit-lift study compares two groups:

Exposed group: devices observed in the exposure zone of the campaign's boards or screens during the flight.
Control group: similar devices that were not exposed.

You then compare the share of each group that visited the store during a set window after exposure.

Lift in plain terms

Lift is the difference between the two visit rates, expressed relative to the control.

Here is a made-up example. Suppose 2.4% of exposed devices visited the store during the window, and 2.0% of control devices did. The difference is 0.4 points. Divided by the control rate of 2.0%, that is a 20% lift. In plain words: the exposed group visited about a fifth more often than the comparison group.

The control rate matters as much as the exposed rate. It is your estimate of what would have happened anyway. Many people who walked past your board were going to shop with you regardless. A sound study counts only the extra visits above that baseline, not every visit by an exposed device.

Why the control group is the hard part

Exposed and unexposed people are rarely alike. Boards go on busy roads, near malls and in neighbourhoods the brand already does well in. People who pass them often live or work near the store. Compare them with a random set of devices and you will find "lift" that is really just geography.

This is not a theoretical worry. In a study of 15 large advertising experiments run on Facebook, researchers compared randomized results with common observational methods, such as matching and regression. The observational methods often failed to recover the true effect, tended to overstate it, and in about half the studies were off by roughly a factor of three ⁵. That was online advertising, not OOH, but the lesson carries over: a weak control group can make an ordinary campaign look like a big success.

The strongest designs randomize. Google researchers have described geo experiments, in which separate geographic areas are randomly assigned to see or not see a campaign ⁶. True randomization is hard in OOH, since you cannot randomly decide who drives past a board. Matched markets, or holding back comparable boards or regions, are practical ways to get closer.

Common pitfalls

Selection bias from board placement

Covered above, and the biggest one. If the boards are near the stores, exposed devices will visit more whether or not the ad worked. The control must come from similar places and similar people, ideally people with a similar past visit rate.

Seasonality and outside events

Campaigns often launch with a sale, a holiday or a new product. Visits may rise for everyone. Compare exposed and control groups over the same dates, and check both groups' visit rates before the campaign.

Small samples

If only a small share of exposed devices is observed visiting, a handful of visits can swing the result. Lift should come with a range, not just a single number. The MRC guidelines say standard errors around sample-based projections should be disclosed ³.

Attribution windows

The window is how long after exposure a visit still counts. Too short and you miss real effects. Too long and unrelated visits creep in. Set the window before reading results, match it to the category's purchase cycle, and do not shop around for the window that looks best.

Counting people who would have visited anyway

Some reports present "exposed visitors" or "attributed visits" as if every one was caused by the campaign. Without a control, those numbers measure overlap, not effect.

A step-by-step outline of a sound visit-lift study

Write the question first. For example: did the campaign raise the visit rate to our stores in the markets where it ran?
Define exposure. Board and screen locations, facing direction, viewing distance and flight dates. For DOOH, use ad-play times where you have them.
Define a visit. The store boundaries, and the rules that separate a visit from a drive-by.
Pick the window and the outcome before launch. Write them down.
Build the control. Use unexposed devices matched on where they live, their past visit behaviour and their usual travel patterns, or hold back comparable markets.
Measure a pre-period. Check that exposed and control groups had similar visit rates before the campaign. If they did not, the comparison is suspect.
Calculate lift with a range. Report the exposed rate, the control rate, the lift, and the uncertainty around it.
Check robustness. Does the result hold with a slightly different window, or for different store groups? A real effect usually does.
Report shares and indexes, with the method stated. Readers should be able to see how the number was made.

Red flags in vendor reports

No control group, or a control described only as "the general population."
Large "attributed visits" totals with no baseline.
A single lift figure with no range or significance test.
No definition of exposure or of a visit.
No pre-period comparison between groups.
A window chosen after the fact, or windows that change between reports.
Results that are always positive, for every campaign and every market.

The honest limits

Lift is an estimate. Even a well-designed study gives a range of plausible effects, not a precise figure.
Visits are not sales. Location data shows that a device was at a store. It does not show what anyone bought ⁷. Pair visit lift with your own sales data where you can.
It is a sample. Visit data reflects a large sample of devices, never every customer, so results are shares and indexes, not headcounts ⁷.
Exposure is inferred. Being near a board is a strong proxy for the chance to see it, not proof that someone looked.

How Measured Reporting approaches it

Our measurement work compares exposed audiences against a control group and reports the lift in visits ⁸. Results are built from aggregated, anonymized, privacy-compliant mobile data from our Canadian data partner, with no personal information ⁷. Media owners use the same data to show who passes their screens, and brands use it to see whether a flight moved store traffic.

FAQ

Does COMMB measure store visits?

COMMB measures OOH audiences, including reach and frequency, for planning and buying ¹². Store-visit outcomes come from separate attribution studies.

What is a good lift?

There is no universal benchmark. It depends on the category, the control and the window. A modest lift with a sound control and a clear range is more useful than a large lift without either.

Can you measure static billboards, or only digital screens?

Both can be studied. For a static board, exposure is usually inferred from devices observed near it, travelling towards its face, during the flight. Digital screens can add ad-play timing, which sharpens exposure ⁴. Ask any vendor how it handles each.

How long should the attribution window be?

Long enough to match how often people buy in the category, and fixed before results are read.

Is visit attribution privacy compliant?

Our insights are aggregated and anonymized, with no personal information, following PIPEDA and provincial privacy rules ⁷. Get your own advice for your situation.

Next step

See how measurement and attribution works, and how media owners and DOOH networks use audience proof, or request a complimentary Sample Insight on a location you care about. It takes about a week.

See how this works in practice on our footfall attribution page.