By Chris Gadek, AdQuick.
Structured data is the precondition for programmatic anything, and physical ad inventory is among the least structured data in all of advertising. That single fact explains why out-of-home was late to automation, why the fix is fundamentally a data-engineering problem, and why the interesting work here looks a lot like the work of turning any messy real-world domain into something a system can reason over.
Start with why the web made programmatic buying look easy. Every web ad slot came pre-structured. It had an address, it sat inside a document with a known shape, and it reported back through a pixel whether it had loaded. The environment was machine-readable from birth because machines were already serving it. Automated buying inherited a schema it did not have to build. The hard modeling work had, in effect, already been done by the medium itself, which is why the industry mistook the ease for something intrinsic to programmatic rather than a gift from the environment.
A billboard inherited none of that. Its location is a point on the earth with an orientation. Its audience is a function of the road it faces, the direction of travel, and the speed of the traffic. Its availability is a schedule with holds and expirations. Its price is a function of date and demand dressed up as a laminated number. For most of the medium’s history, all of that lived in a PDF and a salesperson’s memory, not in anything a program could query. The medium was, from a systems perspective, dark: valuable resources with no interface in front of them.
So the real work of bringing programmatic buying to the physical world is not the bidding logic, which is well understood. It is the data engineering that turns a large set of inconsistently described physical objects into one normalized, queryable representation with reliable current state. It is worth naming the parts, because they will be familiar to anyone who has built a serious data pipeline. Entity resolution, because the same board appears in multiple vendor feeds under different names and coordinates, and one physical object has to resolve to one record before anything downstream can trust it. Geospatial modeling, because location is not a label but a point with a bearing, and the useful field, estimated exposure, has to be computed from road geometry and traffic rather than read off directly. Normalization across formats that do not agree on what a unit is. State management, because availability and price change underneath you and stale data is worse than no data.
Once that representation exists, programmatic ooh becomes possible, because the buying system finally has clean, current data to act on. The demand side was always solved. What was missing was a supply side with a machine-readable shape. This is the reframe worth internalizing: the medium was never the bottleneck. The interface to the medium was, and an interface is a thing you build.
There is a broader lesson here that applies far beyond advertising, which is why it is worth spelling out for anyone who works with data infrastructure. A great many domains look resistant to automation when they are really just unstructured. The physical world is full of valuable systems, logistics, real estate, physical retail, that lag their digital counterparts not because they are less amenable to software but because nobody has done the unglamorous work of modeling them into queryable data. When you see a domain that seems stubbornly manual, the useful question is whether the manual quality is fundamental or whether it is simply waiting for someone to build the schema. More often than not, it is the latter, and the opportunity is in the modeling.
The parallel to how we think about any data platform is close. The value is rarely in the flashy layer on top. It is in the pipeline underneath: the ingestion, the normalization, the entity resolution, the current-state store. Those are the parts that determine whether anything built on top can be trusted, and they are the parts that are hard to replicate. In a world where the querying and the algorithms are increasingly commoditized, the durable advantage sits with whoever has done the work of structuring a hard, messy, valuable part of the world into clean data. Physical advertising inventory is about as hard and messy as it gets, which is exactly why structuring it well is worth doing.
The takeaway, whether or not you ever touch advertising, is that structure is the precondition for automation, and structuring the physical world is a genuine and underexplored engineering frontier. Out-of-home is one large, concrete instance of a physical medium getting the data treatment every digital channel received years ago. The interesting problem is not the buying. It is turning the physical world into something a program can query, and that problem recurs everywhere the physical and the digital meet.
