OOH campaign verification is the process of proving that each poster, billboard or citylight in a campaign went up on the right site, with the right creative, at the right time – usually from installation photos and the metadata stored with them. Photos remain the main proof: the billboard categories in the OAAA’s year-end 2025 count add up to 420,500 US faces, and only 25,000 of them are digital (OAAA display counts).
This article is for out-of-home operators, the operations and IT teams behind them, and the integrators who build software for both. It covers classic printed formats – posters, billboards, citylights and banners – and the process it describes works the same in any market; the figures quoted come from the US and Italy.
Outdoor operators close thousands of installation work orders a month, each with a photo from the field. Nobody reviews that many photos carefully, so a missing or wrong posting surfaces only when the advertiser complains.
Software can read every installation photo and its metadata, approve the clear cases on its own and send only the doubtful ones to a person, so OOH campaign verification stops depending on how many photos your team has time to open.
Questions this article answers:
- What is OOH campaign verification?
- How do you check thousands of installation photos without manual review?
- What can AI read from a billboard photo?
- Can you test it on one month of your own data?
Proof of posting today: photos, spreadsheets and spot checks
Proof of posting, part of what US contracts call proof of performance, is the photo record an operator collects to show that a campaign went up as booked. The way it is produced has barely changed in years, and that is where the review problem starts.
Proof of posting photos from the field
An installation crew gets a list of work orders for the day: site, face, creative, posting date. After each job, the installer photographs the face with a phone in the field app and closes the order. The photo goes to storage, usually under an anonymised file name, and the system links it to the work order ID. EXIF data – the metadata a phone stores inside the photo file: capture time, GPS position and device model – is often extracted into text fields at the same time.
What the photo should prove is well defined. The OAAA’s photography best practices list it plainly: that installation occurred, that the correct copy was posted, on the correct location, and that the material went up properly, “i.e. no wrinkles”. The same document describes close-up and approach shots and asks for photos that stay unaltered apart from nominal cropping. Billboard installation verification, in other words, has a checklist; what it lacks is someone to apply it to every photo.
From there, the data usually lives in two places: the work orders in the operator’s ERP or campaign system, grouped by internal campaign IDs, and the photo results linked to those orders. Both can usually be exported to a spreadsheet. The raw material for automatic checks already exists in most operators’ systems; nobody reads it at scale.
Limits of manual review at thousands of work orders a month
Do the arithmetic. At 30 seconds per photo, every 1,000 photos take more than eight hours, so a few thousand work orders add up to days of one person's time every month, and that is before anyone checks a timestamp against the posting date or a GPS position against the site. In practice teams sample: they open photos for key clients, for complaints and for crews with a history of problems.
Public data on how often postings go wrong is scarce. The clearest figure we found comes from a case study published by the field-audit company BeMyEye for Kinetic in Italy, which runs more than 100,000 billboards there: its checks found nearly 10% non-compliance nationwide, with local peaks of up to 23%. Treat it as one vendor's number, but it shows the order of magnitude.
Sampling misses exactly what costs money. A poster pasted on the wrong face, an old creative left up for another week, a photo taken from so far away that the panel is a smudge, or the same photo reused for two orders all look fine in a list of green ticks. The operator finds out when the advertiser's agency runs its own check, or when a campaign report for the client needs evidence nobody looked at.
Photo and metadata checks in OOH campaign verification
A photo carries two kinds of evidence: what the camera saw and what the phone recorded about the moment. Automatic checks work on both, and the table below shows which problem each one catches.
| Check | Data it uses | What it catches |
|---|---|---|
| Face visible and photo usable | Image | Blurred, dark, obstructed or distant shots |
| Correct creative | Image vs. campaign artwork | Wrong or outdated poster on the face |
| Posting quality | Image | Torn, wrinkled or partly pasted posters, unlit citylights |
| Location | EXIF GPS vs. site coordinates | Photo taken somewhere else, wrong site in the order |
| Time | EXIF capture time vs. posting window | Photo taken before the job or long after it |
| Duplicates | Image fingerprint across all orders | One photo used to close several orders |
| Missing evidence | Work orders vs. photo results | Orders closed with no photo at all |
Image checks: site, creative and photo quality
The first question is whether the photo can prove anything. An image model can tell whether there is an advertising face in the frame, how much of the frame it fills and whether the shot is too dark, blurred or blocked by a tree or a truck. A photo that fails here goes back to the crew before anyone argues about its content.
The second question is whether the right creative is on the face. The model compares the photo with the campaign artwork attached to the work order: colours, layout, logo and readable text such as a brand name or a slogan. A loose comparison like this works better than a pixel match, because outdoor photos are taken at an angle, in rain, in low sun and through reflections on the citylight glass. The same model can flag torn or wrinkled posters and citylights that are not lit after dark, although these checks need the most tuning on your photos before you trust them.
EXIF checks: location, time and duplicates
EXIF metadata answers where and when. If the GPS position in the file is 2 km away from the site in the work order, either the photo is wrong or the site data is, and both are worth knowing. If the capture time is three days before the posting window, the crew may have photographed the old campaign.
Metadata has limits. The Exif standard treats both the capture date and the GPS position as optional fields, so a photo may simply lack them. Free tools such as ExifTool can rewrite or delete every tag, shift dates and even add a GPS position after the fact, and phones let users share a photo without its location. Some systems may also allow office staff to change a photo's date before it reaches the client report. On top of that, phone GPS can drift by tens of metres between tall buildings. A missing or inconsistent EXIF block is a reason for review; on its own it proves nothing.
Duplicate detection does not depend on EXIF at all: an image fingerprint finds the same or nearly the same photo used for two orders, even after light cropping or resizing. A match is not always a problem, though. When the same creative stays on the same face for a follow-up booking, two orders can legitimately end up with near-identical photos, so duplicates go to the review queue instead of being rejected automatically.
Limits of AI in OOH campaign verification
AI reads what the photo shows; it cannot know what the photo leaves out. It will not confirm that the poster on the far side of the board is the right one if the crew photographed only one side, and it will not settle a contract dispute. Its job is to sort thousands of photos into those that clearly pass, those that clearly fail and those a person needs to see.
That sorting has to be tuned on real data. Thresholds that work for a city with good light may flag half of the photos from a motorway route shot at dusk. Any vendor who quotes an accuracy figure before seeing a month of your photos is guessing, us included.
Tagging and rejecting work orders in OOH campaign verification
Checks are useful only if their results land where your team works: on the work order. In OOH campaign verification each order then carries a status, a reason and, when needed, a task for the crew.
Three statuses instead of thousands of thumbnails
Each work order gets one of three statuses, with the reason attached:
- Approved – the photo is usable, the creative matches, location and time fit the order. Most orders in a clean month should land here and need no human attention.
- To review – something is uncertain: the creative match is weak, EXIF is missing, the photo is dark. A person looks at the photo next to the artwork and decides in seconds.
- Rejected – the evidence clearly fails: wrong creative, a photo from another site, no photo at all.
Tags add detail for reporting and for the next campaign: "citylight unlit", "obstructed by vegetation", "duplicate of order 4471". Operators usually ask to try the tool on a real month of data before discussing tags. A pilot is built around that order.
Rejections, re-shoots and extra photo orders
A rejected order should go back to the installation crew as a task to re-shoot or fix, with the reason and the photo attached, so the crew does not have to guess what was wrong. When the new photo arrives, it goes through the same checks.
The same mechanism handles photo requests that are not tied to an installation: a client wants a fresh photo of a site, or the sales team needs current images of a location. Someone creates the request, a crew member logs in, takes the photo, and it goes through the checks like any other. An AI assistant in the CMS can take this further and answer questions like "which orders in the June campaign are still waiting for a re-shoot?" without anyone building a report.
ERP connection without replacing the system
Operators do not want another system that requires a migration. Their work orders, campaigns and invoices already live in an ERP or a campaign management system, and that is where they should stay. Integration with the systems already in place is one of the critical challenges facing the DOOH industry in 2026, and printed OOH has the same problem.
Spreadsheet exports as the first interface
The quickest connection is two exports from the existing system. The first holds the month's work orders with internal campaign IDs, which are safer for grouping than campaign names that repeat. The second holds the photo results, each linked to its work order by ID, with the EXIF fields as text and the photo's file name. Add the web address where photos can be fetched, and the checks can run without touching the ERP.
For a pilot, two files are enough. Your IT team prepares an export, nobody gets access to production systems, and you see results on real campaigns before anyone writes an integration.
Live OOH campaign verification through the API
Once the checks prove useful, the connection moves to an API integration with the ERP. New photos are checked as they arrive instead of once a month, and statuses and tags are written back to the work orders in the ERP, so planners and account managers see them where they already look. Rejections can open re-shoot tasks in the field app automatically.
The ERP stays the source of truth for orders and clients, and the new layer reads from it and writes back only statuses and tags. A verification layer you can switch off without breaking anything is much easier to approve than a new system.
Before the pilot, it helps to agree on who owns the result: who reviews the "to review" queue, who handles rejections with crews, and who reports to clients. If you want to go through that with someone who has built the software side, book a call below.
A pilot of OOH campaign verification on one month of data
The fastest way to find out whether automatic checks work for your network is to run them on a month you already know. You know which campaigns had problems; the pilot shows whether the software finds them too.
Data we need from you
- Work orders for one month – an export with order IDs, sites, faces, campaign IDs and posting dates. Pick a busy month.
- Photo results – an export linked to the work orders by ID, with file names and the EXIF fields you already extract.
- Access to the photos – a web address or a storage location from which photos can be fetched by file name.
- Campaign artwork – the creatives for the campaigns in that month, so the software has something to compare against.
- A few known problems – orders you already know were wrong, as a test of whether the checks catch them.
Your team's view after about a week
We load the month into a working application, run the checks and show every work order with its status, reasons and photos side by side with the artwork. Your team clicks through it: filters by campaign, crew or status, opens the doubtful ones, says where the software was too strict or too lenient. We estimate about a week from receiving the files.
The point of the pilot is a decision based on your own photos: whether the share of automatically approved orders is high enough to change how the team works, which checks need tuning, and what the screens should show differently.
How we start: a one-month pilot
Yes, we can automate the review of installation photos for you, and we start with a pilot because accuracy depends on your photos. The one-month pilot described above is our offer for outdoor operators: we run the checks on your own data first, building on what we already do – AI applications that read images and documents, integrations with existing systems and reporting. You see the results on real campaigns before either side commits to more.
Share your data and we will set up a pilot: one month of your work orders and photos, results in about a week.
FAQ about OOH campaign verification
What is OOH campaign verification?
OOH campaign verification is the process of confirming that every face in an out-of-home campaign was posted on the right site, with the right creative, during the booked period. The evidence is usually installation photos and their metadata. The result goes into the campaign report for the advertiser.
What is proof of posting?
Proof of posting is the evidence an outdoor operator collects to show that an ad was installed as booked, typically a geotagged, time-stamped photo of each face taken by the installer. The OAAA files these images under proof of performance photographs: installation, close-up, approach and archival shots.
How do you check thousands of installation photos without manual review?
By letting software do the first pass:
- Every photo is checked – usability, creative match, location, time and duplicates.
- Clear cases are decided automatically – approved or rejected, with a reason.
- Only doubtful orders reach a person – shown next to the campaign artwork.
The team reviews a short queue instead of the whole month. The pilot shows the size of that queue for your network.
What can AI read from a billboard photo?
AI can recognise whether an advertising face is in the frame, whether the photo is sharp and bright enough, whether the creative matches the campaign artwork, and whether a poster looks torn or a citylight unlit. From the file's metadata it reads the GPS position, the capture time and the device. It cannot see what is outside the frame, so one photo of a two-sided board proves only one side. It works best with the two shots most guidelines already ask for: a close-up for posting quality and an approach shot that shows the site.
Can you test OOH campaign verification on one month of your data?
Yes. We need exports of one month's work orders and photo results linked by order ID, access to the photos and the campaign artwork. Our estimate is about a week from receiving the files to a working version your team can click through.
Is EXIF metadata enough to prove a posting?
No. EXIF shows where and when a phone says a photo was taken, but it can be edited or lost when a photo is re-saved. It works well as one signal next to the image itself and the work order data. Missing EXIF should send an order to the review queue. Stronger evidence comes from the field app's own camera, which records the time and GPS position at the moment the photo is taken and blocks uploads from the gallery. GPS can still be spoofed, so even then location is one signal among several.
Do we have to replace our ERP or field app?
No. The checks read exports or an API from the systems you already use and write statuses back to the work orders. Your ERP stays the record of orders and clients, and the installation crews keep their app.
