If you are the head of media at a large advertiser, CTV ad measurement has one job: tell you how much of the budget should go to connected TV. The problem is that every streaming platform reports its own impressions, on its own definitions, and none of them knows what the others delivered or what linear TV already covered. The direct answer is to count the people your ads actually reached, not the impressions you paid for. Measured on real people across TV, streaming, YouTube, audio and out-of-home, you can see how many different people CTV really added, and how often. This guide explains why CTV is hard to measure, the metrics that matter, the main measurement approaches, and how to choose one for 2026 planning.
Connected TV (CTV) is any TV content streamed over the internet to a television screen, whether through a smart TV app, a streaming stick or box, or a games console. It covers broadcaster video on demand (BVOD), ad-supported streaming services (AVOD) and the ad tiers of subscription services. As Beatgrid has argued in CTV Advertising Measurement, CTV is better treated as part of the TV audience than as a digital channel planned by a separate team.
CTV ad measurement, then, is the set of methods used to answer four questions about that audience:
Three things make CTV harder to measure than linear TV.
Fragmentation. Viewing is spread across dozens of apps, devices and services. Each one reports its own impressions, and no single platform sees the whole picture.
Walled gardens. Many of the largest streaming platforms report on their own inventory only, using their own definitions. Reach claims from different platforms cannot simply be added together, because the overlap between them is unknown.
No shared currency. Linear TV has long relied on household level ratings. CTV reporting is usually device level or impression level. Neither tells you how many individual people were reached across both, which is what a frequency cap or an incremental reach calculation needs. Beatgrid's guide to managing CTV incremental reach goes into why person level data matters here.
A useful CTV measurement plan tracks a small number of metrics that connect exposure to outcomes on the same people.
Unduplicated 1+ reach tells you how many people saw the campaign at least once across every screen. Frequency tells you how often. Measured at the person level, these are the base for everything else. In Beatgrid's Domino's UK case study, brand metrics stopped moving once people were exposed more than four times across linear TV, CTV, YouTube and digital video, which let the brand reinvest spend into new reach.
Incremental reach is the share of the CTV audience that linear TV did not reach. Beatgrid's real campaign data over two years shows CTV delivering roughly 30% to 50% purely incremental reach compared with linear TV, with a higher share among younger audiences. In the McDonald's Germany case study, 20% of incremental viewers were reached exclusively through CTV and 31% of CTV-exclusive viewers were under 30.
Attention measures whether the exposed person was engaged. On an ACR panel this is collected through surveys with the same people whose exposure was detected, so it can be read alongside reach and brand response.
Brand lift compares awareness, consideration and preference between exposed and non-exposed people. Because the exposed group is identified from detected exposure rather than claimed recall, the lift can be broken out by channel and by frequency.
For retailers and restaurant brands, the outcome that matters is store visits. Beatgrid's FootfallBeat compares store trips between exposed and non-exposed panelists for a campaign, flight or channel, using opt-in location data.
Panel based ACR. An opt-in panel of real people runs a passive app on their phones. Automatic content recognition (ACR) matches ambient audio against an ad signature database to detect exposure on any screen, including linear TV, CTV, BVOD, online video and radio. Because every exposure is detected on the same person, reach, frequency, overlap and outcomes are all single source and deterministic, with no modeling between them. This is how Beatgrid's ACR panel works.
Ad server logs. Impression logs from your ad server or DSP show what was served and where. They say nothing about linear TV or about who was in the room, and they count devices rather than people.
IP matching. Smart TV or set-top box data matched to digital exposure through IP addresses. This scales to large household counts, but it is household level, limited to the TV brands or operators that supply data, and often depends on probabilistic device graphs.
Clean rooms. Advertiser first-party data and platform exposure data are joined inside a privacy-safe environment. Each walled garden has its own clean room, so cross-platform overlap is still hard to see.
Beatgrid's ReachBeat product delivers reach, frequency and incremental reach per campaign and per channel on this basis, and the wider Beatgrid platform adds BrandBeat and FootfallBeat on the same panel. For a metric by metric view of effectiveness, see how to measure CTV advertising effectiveness.