Where DOOH still doesn't pencil for local service
DOOH is still a screen-impression-priced channel. The unit is a thousand estimated views of a screen — not one verified delivery to a specific buyer. Three structural limits matter for a local service operator:
First, impression measurement is still Geopath-modeled, with the same audience-estimation methodology static boards use. The screen counts the play; the audience figure for that play is a model of how many people were likely in front of it. That's defensible for national brand reach. It is not defensible as last-mile attribution for a $12,000 roof job.
Second, the cross-screen retargeting layer carries the same accuracy problems as billboard mobile retargeting. The mobile-device-ID passback model depends on bid-stream proximity signals — and peer-reviewed GPS research shows 7–13 meters of horizontal error in urban environments where most DOOH lives, plus mobile-ad-ID match-rate fallout (typically 60–80%), bid-stream latency, and the fact that many networks don't publish exact screen coordinates. The geofence radius gets inflated to compensate, which dilutes who actually saw the screen. We covered this in detail at how accurate is geofence billboard retargeting — the analysis applies cleanly to DOOH-attached retargeting because it is the same proximity-signal pipeline, just sourced from a digital screen instead of a vinyl board.
Third, pDOOH inherits the programmatic supply-chain rake. A DSP fee (Trade Desk historically ~20% of gross spend), an SSP fee (Vistar, Hivestack, Adomni, Place Exchange — all charge a take rate), data fees, and verification fees all stack on top of the DOOH-network share. The end CPM the buyer pays is materially higher than the screen owner receives. For a national brand absorbing the rake across millions of impressions, the math still works. For a local service operator running a $5,000 monthly budget against a defined service area, the CPM-times-supply-chain math compounds against an audience-precision problem the channel was never tuned to solve at the corridor level.
Run the math: a local roofing contractor pushes $2,000 into pDOOH targeting their metro. After supply-chain fees and a Geopath-modeled impression count that includes everyone passing the screen — drivers, passengers, out-of-DMA traffic, renters, people who already have a new roof — the share of true buyer-with-need impressions is small, unmeasurable, and not retargetable into a measurable CAC. That's not a DOOH defect; DOOH was not built for that use case.