Out-of-home (OOH) and connected TV (CTV) teams share a dependency that rarely appears on a media plan: accurate place boundaries. Billboard geofences, retail proximity segments, and closed-loop attribution from ad exposure to store visit all require POI & Geofencing polygons that do not bleed into the parking lot next door. When geometry is weak, lift studies inflate, lookalikes drift, and procurement discovers the problem only after a six-figure test. This guide walks through the adtech wiring: geofence build, exposure join, visit proof, and governance, with links to CTV/ACR and cross-channel measurement. Procurement and marketing teams should keep public product claims aligned with tested specs. See AI search readiness for B2B data sites for crawl and schema discipline.
POI (point of interest) data in media measurement is the polygon-primary catalog of store and venue boundaries used to define visit outcomes after OOH proof-of-play or CTV household exposure, not a radius around a billboard pin.
Buyers who run only delivery-based OOH or reach-based CTV reporting are optimizing for inputs the market already commoditized. The differentiator in 2026 is defensible outcome proof: did exposed devices visit relevant stores at rates above control, after polygon QA and exclusion rules both parties can replay? That question forces POI into the architecture early, not as a line item added after the IO is signed. Agencies pitching store visitation without naming the POI vendor, geometry tier, and refresh date are asking brands to trust an outcome they cannot audit.
Media owners and DSPs define geofences around inventory: transit shelters, roadside units, mall media, and place-based screens. Historically, teams drew circles around lat/long pins supplied by the owner. In dense corridors, those circles capture devices that never saw the creative but drove past the block. Polygon-primary POI from GSDSI POI data lets buyers measure visit lift among exposed devices versus a holdout, using adjacent retailer footprints as the outcome geography rather than the billboard radius alone.
Transit and roadside inventory deserve different outcome geographies than mall media. A shelter unit may warrant a tight retail strip footprint; a highway board may need a destination polygon set miles away at the brand's store network. Pre-register those mappings in the pilot charter so week-two results are comparable to week-one assumptions. When media owners supply only lat/long pins, require a polygon upgrade path in the contract or accept that visit lift will be contested in every QBR.
IAB measurement guidance and buyer RFPs increasingly expect location proof beyond delivery logs. Pair OOH reads with global mobility only after POI geometry passes a false-positive test on your chain list: mobility without place truth is un-auditable.
ACR or ad-log exposure at the household level must join to MAIDs or other device keys observed inside store polygons. Weak POI geometry shows up as inflated attribution: the classic fix is polygon upgrades, not a bigger panel. Teams licensing CTV/ACR should confirm householding assumptions, dedupe rules across devices, and content-category exclusions before tying exposures to foot-traffic outcomes.
Frequency capping and competitive separation live on the exposure side; visit proof lives on the place side. Conflating the two in one dashboard without documenting join rates invites false confidence. Require match-rate tables from exposure grain to device grain to store polygon, each with aggregation floors, so stakeholders see where uncertainty enters. When lift disappears after polygon QA, the story is geometry, not creative fatigue.
For identifier basics and decay expectations, see CTV ACR 101. For governed joins, pair this workflow with clean room measurement.
Activation teams use POI to build recent visitors, competitive conquest, and lifestyle proximity segments. The difference between a usable segment and a noisy one is usually polygon quality and refresh, not audience platform UI. Require stable POI IDs, brand hierarchy for franchise rollups, and change-delta files so DSPs do not keep targeting closed locations. Audience targeting programs should document minimum cohort sizes and exclusion lists before export to ad platforms.
Activation should respect minimum cohort sizes, sensitive locations, and purpose limitation. Pair POI with documented exclusions before building lookalikes from visitation. See privacy-safe targeting and what privacy-safe means for location. The FTC's 2024 location-data orders made sensitive-place exclusion an enforcement baseline, not a nice-to-have.
Scope a pilot with your top 50–200 store polygons, a two-week flight, and pre-registered lift thresholds. Ask for polygon WKT samples, hierarchy fields, and refresh dates on the same extract you will use in production. Compare vendors on identical geometry before comparing panel marketing numbers. When ready, request a scoped sample via contact with chain list and DMA attached.
Media mix models that treat OOH and CTV as awareness-only channels leave money on the table when place data is available. The operational shift in 2026 is to require store-anchored outcomes in every location-heavy RFP, not as a post-campaign forensic. Document holdout construction, device graph vendor, and POI vendor in the same appendix so QBR narratives stay consistent when results are challenged. Teams that separate exposure logs from place truth often discover attribution inflation that no amount of panel scaling fixes: polygon upgrades routinely move lift estimates more than ten points in dense retail DMAs.
Finally, align legal review with measurement design. Permitted use for visit-derived segments must match how the DSP or clean room will receive IDs, and sensitive-location exclusions must be tested on the same week as lift readouts. Audience targeting exports should inherit aggregation floors from the measurement environment so activation does not outrun governance.
Pre-register holdout design in the pilot charter: mixing holdout types within one readout produces arguments, not insights.