Connected TV ad spend surpassed $30 billion in 2025, and IAB's Internet Advertising Revenue Report places CTV among the fastest-growing digital channels. Yet most advertisers still struggle with the fundamental question: did my CTV campaign actually drive people into stores? The attribution gap between TV exposure and real-world outcomes remains one of the largest unsolved problems in advertising measurement — and it is the direct reason GSDSI built the Euclidean Feed, a deterministic bridge between Smart TV ACR signals and device-level mobility data.
Key Takeaways
CTV spend topped $30B in 2025 but attribution stayed probabilistic — correlating aggregates, not measuring people.
Deterministic attribution joins ACR (household saw ad) to MAID (devices in household) to POI visits (actual store visit) — answers the measurement question at the device level.
GSDSI's Euclidean Feed links all three layers against ~13–14M monthly unique CTV IDs and ~15K+ daily gz files of visit events.
Early deterministic-attribution adopters report 3–5× measurement-confidence gains vs. probabilistic panels and use the lift to re-allocate DMA / daypart / creative spend in near-real-time.
Why Probabilistic Attribution Hit Its Ceiling
Probabilistic CTV attribution typically correlates aggregate viewing patterns (reported by DSP or measurement vendor) against aggregate foot-traffic trends (from a mobility panel) and infers causation. The approach is directionally useful and cheap to run, but it fails the CMO/CFO question: 'of the people who saw my ad, what share visited my store?' Aggregation removes the causal link. The Media Rating Council's cross-media measurement guidance has, over the last two years, steadily raised the bar on what counts as a measurement-grade attribution read — and aggregate-only methods increasingly fall short of it.
What Deterministic Attribution Changes
Deterministic attribution joins three identifier classes at the row level rather than at the aggregate level. Smart TV ACR identifies which household saw which creative at which second; on-device Mobile Advertising ID (MAID) signals identify which devices are present in that household (via IP-cohabitation, Wi-Fi network join, or probabilistic household graph); dwell-confirmed POI visits identify which of those devices subsequently appeared at the advertiser's locations. When all three layers are present and joined with deterministic keys, the measurement read becomes a closed-loop count rather than a correlation. For an overview of how identity graphs underpin this kind of cross-signal resolution, see identity graphs 101: MAID to HEM, CTV IDs, and household resolution.
What the Euclidean Feed Actually Delivers
GSDSI's Euclidean Feed is purpose-built for this workflow. At the data layer it delivers:
Smart TV ACR exposure records at the second-level creative-identifier grain, drawn from ~13–14M monthly unique CTV IDs.
MAID-to-household resolution, mapping device IDs to the ACR households they cohabitate with.
Dwell-confirmed POI visit events across the GSDSI global mobility panel (~15K+ daily gz files).
Join keys that let advertisers run a household- or device-level attribution query without probabilistic inference.
The advertiser's measurement stack then asks the concrete question: 'of the households exposed to Creative A on Tuesday, how many had at least one resident device show up at a store by Saturday, relative to an unexposed control?' That's a real incremental-lift read, not a correlation.
What Early Adopters Do With It
Advertisers running deterministic CTV attribution report 3–5× measurement-confidence gains compared to probabilistic panels — and the more interesting effect is in optimization. Once the device-level lift is visible, spend moves toward the creative executions, dayparts, and DMAs that actually drive incremental visits. For companion reads on the broader measurement architecture, see cross-channel attribution without walled gardens and cross-channel measurement for privacy-first advertisers. On the advertiser-education side, the Advertising Research Foundation's Cross-Platform Measurement Initiative is the canonical industry-body reference for what deterministic cross-channel measurement should look like at scale.
Frequently Asked Questions
What's the difference between probabilistic and deterministic CTV attribution?
Probabilistic methods correlate aggregate ACR exposure trends with aggregate foot-traffic trends and infer a relationship. Deterministic methods join household-level ACR exposure records to device-level MAID signals to dwell-confirmed POI visits, producing a row-level incremental-lift read rather than an inferred correlation.
What data layers does the Euclidean Feed link together?
Three: Smart TV ACR exposure (second-level creative grain, ~13–14M monthly unique CTV IDs), MAID-to-household resolution (mapping device IDs to the ACR households they cohabitate with), and dwell-confirmed POI visit events from the GSDSI global mobility panel (~15K+ daily gz files). The join keys are built in, so advertisers do not have to stitch them post-hoc.
How does deterministic attribution handle privacy?
All inputs run under the same consent architecture described in the GSDSI privacy center — IAB TCF-compliant CMPs at collection, device-level consent state preserved downstream, sensitive-category polygons enforced in the mobility pipeline. Aggregated outputs to advertisers use minimum-cohort thresholds; no individual device is exposed.
What measurement gains do advertisers typically see?
Advertisers report 3–5× gains in measurement confidence vs. probabilistic panels. The more significant downstream effect is optimization — once device-level lift is visible, buyers re-allocate spend to the creative, daypart, and DMA combinations that actually drive incremental visits, often within one or two campaign cycles.