Connected TV advertising spend surpassed linear premium dayparts in key demos in 2025, and the IAB's Internet Advertising Revenue Report documents the structural shift. But spend runs ahead of measurement literacy — many buyers evaluating CTV datasets still don't have a working mental model for what an ACR panel actually captures, how CTV IDs differ from MAIDs, or what "~13–14M unique CTV IDs/month" buys them in analytical terms. This post is the 101 — what ACR is, what it measures, and how to use the resulting data responsibly. For the practical attribution mechanics that build on this foundation, see CTV attribution: bridging the last mile and cross-channel attribution without walled gardens.
Key Takeaways
ACR (automatic content recognition) is the smart-TV equivalent of an opt-in viewing panel — it captures what's on the screen, not what was served via an ad server.
CTV IDs are device-level identifiers scoped to a smart TV; they are distinct from mobile MAIDs and require an identity graph to link to the rest of the household.
A ~13–14M unique CTV IDs/month panel is large enough for DMA-level reach & frequency reads and national duplication analysis against linear, per MRC measurement standards.
ACR's core analytical advantage is deterministic visibility into both ad exposure AND content watched on the same device — neither set-top-box nor panel-only data matches that combination.
What ACR Actually Captures
Automatic content recognition is a smart-TV feature that matches what's on the screen against a fingerprint library. When the panel viewer opts in at setup, the TV emits a hash or glimpse of the video every few seconds, and the ACR vendor matches it against the fingerprinted library of broadcast/cable/streaming content plus whatever ad creatives have been fingerprinted. That produces a time-stamped sequence of "this device, this program, this ad creative, this time window" — the same person-level content + exposure record that the industry has wanted since the linear panel era, except done at scale on opt-in smart-TV populations. The MRC cross-media audience measurement framework is the standards document buyers should reference when evaluating ACR methodology.
CTV ID vs. MAID: What's the Difference?
A CTV ID is a device-level identifier scoped to a smart TV (Samsung TIFA, Vizio, LG, etc.) and exposed to the ACR vendor under the opt-in. A MAID (IDFA/GAID) is a mobile advertising identifier scoped to a phone or tablet. They are in different ID spaces and do not natively join — you need an identity graph to resolve them to a common household or person. GSDSI's Euclidean Feed is the deterministic linkage layer for CTV ID ↔ MAID ↔ HEM at the household grain; see identity graphs 101 for the mechanics. Without that linkage, a CTV-only panel is useful for in-home reach & frequency but cannot close cross-screen attribution loops.
What ~13–14M Unique CTV IDs/Month Actually Buys You
Panel size translates to analytical power in specific, quantifiable ways:
**National reach & frequency**: ~13–14M unique CTV IDs/month (the monthly scale of GSDSI's CTV/ACR product, drawn from ~15K+ gz files/day of raw signal) supports defensible national R/F reads with sub-5% standard error for most advertiser-sized audiences.
**DMA-level reads**: the panel is large enough to break reach by DMA for top-75 markets with reasonable confidence — below DMA you need to aggregate weeks or use hierarchical models.
**Linear duplication**: because ACR captures both linear-cable and streaming content on the same device, you can compute overlap between a linear buy and a CTV buy without a separate panel stitch.
**Incremental reach against linear**: the cleanest ACR use case — identify households exposed to your CTV buy who had zero qualifying linear impressions in the same flight.
**Creative-level fingerprinting**: if your creative is fingerprinted in the library, you get exposure data at the creative + daypart + network + device level, not just the campaign level.
What ACR Panels Can't Tell You
ACR is powerful but has structural limits buyers should price in:
**It doesn't see set-top-box content not shown on the smart TV** — DVR playback through a different device, mobile streaming, etc. The panel is TV-screen-centric.
**It can't attribute to the individual viewer** — a CTV ID identifies the device, not which household member is watching. Use household-level analysis or cross-reference with a mobile companion panel.
**Creative coverage depends on fingerprinting** — ads that aren't in the fingerprint library aren't captured as ad events. Buyer should verify fingerprint-library coverage for their own creative before relying on ad-level reads.
**It doesn't replace panels for under-penetrated demos** — smart-TV households skew toward specific income and age bands; compare to the IAB CTV revenue report demo data before using ACR as the single source.
How to Evaluate an ACR Dataset Before Licensing
Vendor-evaluation diligence for ACR tracks closely with the mobility-data diligence in the geospatial data quality framework. Before signature, verify:
Panel size and OEM mix — is the panel concentrated in one OEM or diversified across Samsung/Vizio/LG/etc.?
Opt-in mechanics — what does the consumer see at setup; is consent MRC-audited?
Fingerprint-library coverage — network/programming coverage + ad-creative coverage; both matter.
Identity-graph options — can the vendor resolve CTV ID to MAID/HEM, and at what match rate? (See GSDSI's Euclidean Feed as the reference architecture.)
Delivery cadence — raw gz files vs. aggregated hourly feeds vs. daily R/F tables; match the cadence to your analytical use case.
ACR is now table-stakes for any advertiser spending meaningfully in CTV. The panel economics are real, the analytical ceiling is higher than any legacy TV measurement method, and the compliance story (opt-in at setup, household-level scope, no inferred PII) holds up under scrutiny. Get the basics right and it's a genuinely durable measurement layer.
Frequently Asked Questions
What is ACR and how does it differ from set-top-box data?
ACR (automatic content recognition) is a smart-TV feature that opt-in viewers enable at setup. The TV fingerprints what's on the screen and matches it against a reference library of programming and ad creatives, producing a device-level record of both content and ad exposure. Set-top-box (STB) data is operator-controlled tuning data (channel + time, often with a lag) that doesn't see streaming content or ad creatives. ACR's advantage is unified content + exposure visibility on one device; STB's advantage is broader linear penetration. Most serious measurement stacks use both.
How is a CTV ID different from a mobile advertising ID (MAID)?
A CTV ID is scoped to a smart TV device (Samsung TIFA, Vizio VIDA, LG, etc.) under the OEM's opt-in terms. A MAID is scoped to a phone or tablet (IDFA/GAID). They live in different identifier spaces and do not natively join. To connect them at the household level, you need a deterministic identity graph like GSDSI's Euclidean Feed — see identity graphs 101 for the mechanics.
Is ~13–14M unique CTV IDs/month enough for national reach & frequency?
Yes — for most advertiser-sized audiences that panel depth supports defensible national reach & frequency reads with sub-5% standard error, DMA-level breakouts for top-75 markets, and incremental reach vs. linear. Below DMA or for niche demos you should aggregate multiple weeks or use hierarchical models, and cross-reference against the MRC cross-media framework to set expectations about confidence intervals.
What are the biggest limitations of ACR data buyers should price in?
ACR is TV-screen-centric (it doesn't see mobile streaming or DVR playback on other devices), it identifies the device not the individual viewer, and creative-level reads depend on your ad being in the fingerprint library. Smart-TV households also skew toward specific income and age bands — check the IAB Internet Advertising Revenue Report demo data before treating ACR as a single source for audience measurement.