Everything Mira measures starts from the same question: which people had the opportunity to see this campaign?

Digital advertising does not have to ask it. An ad server knows exactly who it served an impression to, because it did the serving. Out of home has no such record - a billboard is visible to whoever passes it, and nobody is keeping a list. Exposure is the process of reconstructing that list from two things that were recorded independently: where the media was, and where people were.

An exposure is the output: a record that a particular device was in a position to see a particular piece of inventory at a particular time. Each carries an advertising ID, a timestamp, and whatever metadata described the media, which is what later makes exposure-side cuts possible.

The Two Inputs

Mobile location data is, in essence, a table of three columns: an advertising ID, a timestamp, and a latitude/longitude pair. Each row asserts that a device was at a point at a time. Mira observes a sample of the population rather than all of it, which is the root of most of what makes OOH measurement statistically distinctive - see OOH Differences from Digital.

A description of the media says where each piece of inventory was and, depending on the type, when it was showing the campaign. This arrives as a Source, and what it needs to contain differs by media type. For creating and managing Sources in the platform, see Sources.

Building the Viewshed

Location data is points. Media, as delivered, is usually also points - a coordinate pair per unit. A point cannot be "visited", so the media has to be turned into an area before the two can be compared.

That area is the unit's viewshed: the region from which the unit was plausibly visible. Its size and orientation are derived from:

• latitude and longitude - where the unit stands
• format - a bulletin beside a highway is legible from far further away than a shelter poster on a footpath
• size - larger copy is readable from further back
• facing, or bearing - which way the unit points. A billboard facing north is not visible to traffic approaching from behind it

The result is not a circle. A viewshed is directional, because inventory is directional, and the asymmetry matters: drawing a radius around a unit would credit exposure to people who were behind it.

📘

Supplying the Shape Directly

Some inventory is not a unit on a pole. Arena sponsorships, venue takeovers, and similar placements are described by handing Mira the geofence itself as a polygon, which skips the derivation entirely. See Static.

The Join

With a viewshed per unit, exposure is the intersection of place and time:

1. Join location points to viewsheds, keeping the points that fall inside one.
2. Filter on time, keeping the points that fall within a window during which the unit was actually showing the campaign.

A point that survives both is an exposure. A device inside a viewshed at a moment the campaign was not on screen is not exposed, and neither is a device that passed the same spot a week before the flight began.

Time Is Where the Media Types Diverge

The spatial half of that join is the same everywhere. The time half is not, and it is the reason exposure has three variants rather than one.

Static is the simple case. The unit does not move and the copy does not change, so any device inside the viewshed during the flight had the opportunity to see the campaign. Time enters only as the campaign's start and end dates.

Digital inventory rotates. The screen showing this campaign at 2:14pm is showing a different advertiser at 2:15pm, so being inside the viewshed is not sufficient - the device has to have been there while this creative was on screen. That requires knowing the exact times the ad played, which is why digital measurement depends on ad play logs rather than a schedule or a play count.

Moving inventory is the general case rather than a third special one. Its position always varies, so there is no fixed area to join against and the asset's path has to be reconstructed from timestamped positions. Its content may or may not vary on top of that: a wrapped vehicle carries one creative for the whole flight, while a digital screen on a bus rotates between advertisers exactly as a roadside digital unit does.

Where both vary, the two constraints compose. A device counts as exposed only if it was near the asset and the campaign's creative was on screen at that moment - and because the asset was somewhere different a minute later, each qualifying window has its own location. That is why moving media is described as intervals rather than points: each record covers a stretch of travel, and for rotating content, only the stretches during which this campaign was playing.

📘

Where Precise Location Data Is Restricted

Some US states restrict the sale of precise location data, and there the join above runs against a much larger Exposure Zone rather than a viewshed. It keeps campaigns measurable at a cost in precision, and it is opt-in. See Exposure Zones for Coarse Regimes.

🚧

Coverage Determines What Can Be Measured

Exposure can only be computed for inventory Mira has data for. If a vendor supplies only part of a buy, the study measures only that part. The Attribution Checklist covers what to confirm with each vendor before a campaign runs.

Why Exposures Are Not Impressions

An exposure is not an impression, and the two numbers should not be expected to agree.

An impression estimate is a projection of how many viewing opportunities a buy delivered in total, produced by the industry's planning methodology from traffic counts and travel surveys. An exposure is an individual observation: this device, this unit, this moment.

Mira observes a sample of the population, so exposures will always be a fraction of impressions. And because measurement counts unique people rather than opportunities, a person who passed the same billboard on forty commutes contributes forty exposures but one user to a conversion rate. Both properties are expected, and both are why a study's user counts look small beside a media plan.

What Happens Next

Exposure defines the group whose behaviour is measured. From there:

• the exposed group is matched against a comparable unexposed group - see Control
• both groups are cross-referenced against conversion events - see Conversion Properties