Panel Data vs Ground-Truth Counting: Where OOH Impression Estimates Go Blind

Two very different things get called "audience data" in out-of-home, and confusing them costs money. This post compares panel data vs ground-truth counting — where each comes from, what each can and can't see, and when a small operator genuinely needs a per-face count rather than an area estimate. It also covers, honestly, when a measurement bureau is still the right call.

It builds on the OOH audience measurement guide and complements the five-question audit, where "by what method?" is one of the questions this post helps you answer.

Key takeaways

• Panel data is modelled from a device sample and reported at area or road level; ground-truth counting is measured at a single face.
• Panels go blind exactly where you sell: the specific face, the specific side, the specific visibility zone.
• Bureau currency is built for large national inventories; a per-face certificate is the ground-truth alternative when that model doesn't fit.
• Use area data for broad planning and ground truth for pricing and defending a specific face.

What "panel data" actually is

Mobile-panel providers estimate footfall from a sample of mobile devices — a panel — and then model that sample up to represent a whole population across an area. It is genuinely useful for regional patterns: which districts are busy, how a corridor trends month over month, where demand is shifting.

But three things follow from how it's built:

1. It's modelled, not observed. The reported number is an inference from a device sample, scaled by assumptions about how representative that sample is.
2. It's area-level. Data is reported for a block, road, or zone — not for one hoarding facing one direction.
3. It has coverage gaps. People who don't carry a tracked device, or who share one, or who are in a group, are under-counted, and different populations are affected differently.

None of this makes panel data bad. It makes it the wrong tool for one specific job.

What "ground-truth counting" actually is

Ground-truth counting observes the real world and counts it. In AdWitness's case, that means video of the face's visibility zone, with the engine detecting and tracking each pedestrian and vehicle and counting the ones that cross a line you draw. There is no population model in between — the count is what passed, on which side, during the observed window.

The trade-off is the mirror image of panels: ground truth is narrow and measured where panels are broad and modelled. It tells you about one face extremely well, and says nothing about the next district over.

Panel data vs ground-truth counting: where OOH estimates go blind

The blind spot matters because it lands exactly where you make money: the individual face.

• The specific face. An area estimate can't separate your hoarding from the one across the street; ground truth counts yours.
• The specific side. A one-way or single-facing panel gets impressions only from the traffic that could actually see it. Area data doesn't split direction; a counting line does. This is why the certificate reports a direction split.
• The visibility zone. Impressions should count only the audience with an opportunity to see the face, not every device on the whole road. Ground truth counts the zone; area data counts the area.

When a buyer's analyst asks "how many people actually passed this face, on the side it faces?", an area-modelled number can't answer and a per-face count can. That is the difference between a price you can defend and one you can't — the theme of never selling on an unverified number.

The bureau question — answered honestly

Measurement bureaus like Geopath provide an audience currency that national agencies plan and trade against. That is real value, and for a large operator selling into national bureau-based buys, membership can be worth it.

The problem for small operators is fit, not quality:

• Pricing and structure assume a large inventory — membership runs into the hundreds per month, with per-unit onboarding fees, and reported rates have risen over time. (Treat specific figures as reported and dated; verify current fees before you rely on them.)
• Coverage and geography are geared to the bureau's footprint, which may not include your faces at all.

So is there a Geopath alternative for small operators? For the job of pricing and proving a specific face, yes: a per-face Audience Certificate gives you video-verified, ground-truth counts for $99 a face with no membership and no minimums. It is not a national currency and doesn't pretend to be — if an agency demands bureau currency for a national buy, you may still need the bureau. For everything at the level of this face, this quarter, ground truth is the honest and affordable answer.

When to use which

The rule of thumb: use modelled area data to decide where to invest, and ground-truth counting to price and defend what you sell.

Get ground truth for your faces

If you are quoting a CPM or building a media kit, you are working at the face level — the exact level where area estimates go blind. Certify one face for $99 to see a ground-truth count for one of your faces, or view pricing to do a network. Then sharpen your scepticism with the five-question impressions audit, or step back to the OOH audience measurement guide.