What is CTV measurement?

The definition is short. What buyers’ interviews add is which of six methods companies actually use to measure streaming and TV ads, and where each one stops.

Based on verified interviews with the marketing, media and growth leaders who buy and measure TV and streaming advertising, at DTC and enterprise brands. Buyers are anonymized before publication; vendor names and views are reported as given. No vendor paid to appear or could edit this page.

CTV measurement is how a brand assesses what its connected-TV and streaming ads delivered: which households or devices they reached, what those households did afterward, and whether the ads added sales that would not have happened anyway. These are separate questions: exposure does not identify who watched, and a purchase credited to an ad is not automatically one the ad caused. The hard part is that nobody clicks a TV. A streaming-service marketing director puts it plainly: “CTV usually is a black box, it's really hard to track without a partner like Kochava.” Buyers agree on the problem. They solve it in six different ways.

What is CTV measurement?

CTV measurement covers the methods that tell a brand what its connected-TV advertising delivered. Connected TV means ads shown inside streaming apps on a television, through a smart TV, a streaming device such as a Roku or Fire TV stick, or a console. It answers three separate questions: exposure (which households saw the ad, and how often), outcome (what those households did afterward: a site visit, a sale, a store visit) and incrementality (whether the ad caused the outcome, or the household would have bought anyway).

Most CTV measurement products answer the first two by connecting reported exposure to later activity within a set window, using IP addresses, household or device graphs, publisher data, first-party identifiers or clean-room matching, depending on what is available and permitted. Those matches can be incomplete or wrong, and a household connection does not prove the same person saw the ad and bought. The result is an attributed relationship, not necessarily an incremental one: the household might have bought anyway. Controlled lift experiments, which compare exposed groups with comparable controls, give the strongest direct evidence of cause when they are designed well. Mix models and other statistical methods can also estimate incremental contribution, but they depend on assumptions, data quality and separating advertising from other demand drivers.

It is not a ratings currency, the audience figure TV is bought and sold on, though buyers here still call Nielsen that currency. The same questions apply to scheduled linear TV, and most of the methods below measure both.

The second half of the question is the one buyers can answer that a definition cannot:

In these interviews companies measure TV and streaming ads in six ways: the buying platform’s own reporting, a marketing mix model, an independent TV measurement company or a publisher’s exposure data matched to their sales, an attribution or analytics tool, response signals such as traffic spikes after an airing, and controlled experiments. Of the companies that describe how they measure TV, about a third use or are testing the platform’s reporting, and a similar share a mix model. Many combine methods, so the shares overlap. They are not adoption rates.

The complaint that recurs across all six is the missing click. View-through credit, the only outcome many platforms report, shows an exposed household converted, and several buyers say they cannot tell from it whether the ad worked.

What is the difference between CTV measurement, digital attribution and MMM?

Each answers a different question, and CTV measurement borrows from all of them. Digital attribution asks which tracked touchpoints preceded a conversion; a marketing mix model, how much of total sales each channel’s spend explains over time; an incrementality test, how much a channel caused in one controlled period; and CTV measurement, who saw the TV ad and what they did next, often with one of the other three on top.

Against digital attribution, the difference is the click. Multi-touch attribution follows clicks and tracked visits; a TV ad produces neither, so CTV tools substitute view-through attribution, which credits a conversion to an ad the household saw within a window such as seven days. A consumer-goods subscription brand passed on multi-touch tools because it spent heavily on linear and streaming TV, and said of them “there's no way for them to understand that impression.”

Against MMM, the difference is the unit. A mix model estimates advertising’s contribution from aggregated data, such as spend, impressions, sales and other demand factors, and needs no household matching, so it covers linear, streaming and offline media alike. It can estimate incremental contribution, but the result depends on its assumptions, enough variation in spend and controls for seasonality; where TV runs in peak weeks, it can mistake TV for the season.

Buyers draw the line themselves. A home-furnishings retailer buying streaming inventory programmatically says “the attribution is just so poor because the platforms can't attribute that CTV app to web conversion pathway”, and reads those buys through a data clean room instead. A casual-apparel brand’s paid-media director explains why a multi-touch tool misses part of the picture: on YouTube, which it calls half of CTV inventory, “Google is not passing you view through data at all.” Those are the gaps the TV-specific methods below exist to fill.

Against incrementality testing, the difference is cause

Matching a household that saw the ad to a later sale measures exposure followed by outcome. A geo-lift or holdout test compares regions or households that saw the ad with comparable ones that did not, so it estimates the lift the ad caused rather than crediting conversions to exposed households. A direct-to-consumer apparel brand running streaming TV is doing both: the platform reports lift through its analytics tool, and a separate geo-lift study is under way “to prove incrementality”.

The terms also differ from the vocabulary. OTT, over-the-top, means video streamed over the internet to any screen, and buyers here use it interchangeably with CTV. Linear TV is scheduled broadcast and cable television. ACR, automatic content recognition, identifies the content playing on participating smart TVs by matching audio or video samples against a reference library of programmes and commercials; measurement providers use consented ACR data to estimate exposure across the devices it covers. That is not every TV or viewer, so wider reach needs weighting or estimation, and ACR does not show which person watched, whether they paid attention, or whether the ad caused anything.

Who measures CTV advertising?

The media, performance and brand leads who own the TV budget, sometimes through an agency. The companies that describe measuring TV run from direct-to-consumer brands moving spend beyond Meta and Google to national retailers, financial-services companies, health insurers, entertainment networks and streaming services measuring their own subscriber campaigns.

Two groups recur. Several direct-to-consumer brands start in CTV through a self-serve buying platform and measure with what that platform reports, sometimes fed into an ecommerce analytics tool. The companies carrying a dedicated TV measurement company are larger advertisers with linear budgets, and several also run a mix model. Retailers with mostly in-store sales sit apart, because their question is store traffic: a regional retailer whose sales are almost all in its stores was testing CTV campaigns “to measure credit card attribution of people who saw the CTV then went to the store”.

The agency is often the operator. A consumer-health device brand’s growth director says its iSpot relationship was agency-run and the brand was taking it over; a personal-care brand reads “cost per spike new user for TV” from its TV agency’s analytics.

How do brands measure CTV advertising?

In six ways, and many companies use more than one, so the shares below overlap and do not sum to 100%. A company still piloting a method counts as testing it, which is not the same as a finished implementation or a good result.

  1. Read the buying platform’s own reporting
  2. Put TV into a marketing mix model
  3. Match exposure data to sales outside the platform
  4. Ask the attribution or analytics tool to cover TV
  5. Read response signals
  6. Run a controlled experiment

About a third of the companies that describe how they measure TV use or are testing the platform’s reporting, and a similar share a mix model; about one in five match exposure outside the platform or in a clean room; about one in six each use an attribution tool, response signals or an experiment. These are reported methods, pilots included, not market penetration, and a company can appear in several.

01
Most common

Read the buying platform’s own reporting

The CTV platform the brand buys through, MNTN, tvScientific, Vibe.co, Tatari, Keynes Digital or a DSP such as The Trade Desk, places a pixel on the site, matches exposed households to visits and sales, and reports return on ad spend. It is the cheapest route and the one buyers question most, because the seller of the media is also grading it.

A footwear brand buying through MNTN says “We do have basic pixel tracking in place for that to justify the return, which surprisingly works really well.” An outdoor-apparel brand that buys streaming through The Trade Desk is less sure: “you tend to see a lot of conversions attributed to view-through attribution, which we don't have a lot of confidence in.”

02
Equally common

Put TV into a marketing mix model

Built in house, on open-source code such as Google’s Meridian or Meta’s Robyn, or bought from a vendor such as Mutinex or Recast. The model needs no household data, so it covers linear and streaming TV beside the other channels; it is slow, and it needs spend that varies enough to separate TV from everything else.

A European mattress brand runs an in-house mix model and plugs in a third-party TV attribution feed from Innovid: “For TV, we also plug that on top from this third-party client.” A vacation company paused its in-house model because it ran most of its TV and radio in peak season, so the model credited them with the season’s sales: “the results were very misleading in a way.”

03
Common

Match exposure data to sales outside the platform

An independent TV measurement company, iSpot, VideoAmp, Innovid or Data Plus Math, measures which households saw the ad across linear and streaming and links them to web visits or sales; or a brand brings a streaming publisher’s exposure logs into a clean room and matches them against its own transactions. Either way the analysis happens outside the platform that sold the media, though it may still use publisher, identity-partner or seller data: a different source, not necessarily an unbiased or causal one. It costs a contract and, for clean rooms, a data partnership, and its coverage and matching need checking as well as the commercial relationship.

A consumer-health device brand says “We use iSpot for our linear TV measurement”, to see which spots and networks drive traffic and sales. A financial-services company that has used iSpot for about seven years credits “their integration with connected TV partners”. The home-furnishings retailer has “used data clean rooms to evaluate CTV buys with direct publishers”, and saw far more conversions credited to those buys once the logs met its own transaction data. That is attributed credit, not measured lift.

04
Less common

Ask the attribution or analytics tool to cover TV

Brands with a multi-touch attribution or ecommerce analytics tool, Rockerbox, Northbeam or Triple Whale, try to fold TV into it. It puts TV beside the other channels in one view; where the tool imports the CTV platform’s own figures, it carries that platform’s view-through credit rather than measuring TV independently.

A fitness company moved to Rockerbox because its previous attribution tool was digital-only and its radio and linear TV “did not tie into the MTA at all”. A home-appliance brand says of Northbeam that its TV and streaming are awareness channels, and “Northbeam is very click-driven.” A wellness brand found in mid-2026 that its Vibe.co spend “wasn't on Triple Whale”; Vibe.co’s help center now documents an integration that pushes its spend and attributed conversions to Triple Whale, with views added once switched on.

05
Less common

Read response signals

The oldest TV methods: watch site traffic in the minutes after an airing, give each spot its own phone number or promo code, or ask customers how they heard of the brand. They need no new software and work for linear TV; they read the response that shows up quickly and miss the rest.

The casual-apparel brand above buys its TV through Tatari and says “we just try to look at the logs and when things run and if there's a correlation between the lift and incremental traffic directly to the site”. A health insurer gives TV ads their own tracked phone numbers, and an online gaming company used promo codes for TV and radio.

06
Less common, and tests incremental lift directly

Run a controlled experiment

A geo-lift test turns TV on in some regions and holds others back; a holdout withholds the ad from a share of households; a matched-market test compares similar markets. Each compares outcomes under different exposure, so it estimates lift rather than crediting conversions to exposed households. Randomization protects the comparison best; matched markets are only as good as the match. All need variation in exposure, reliable outcome data, and enough budget, time and sales to read a result.

The direct-to-consumer apparel brand’s geo-lift study on streaming TV, which a third party runs for its CTV platform, is “going to run for about 42 days to hit statistical significance.” A multi-brand beauty company is running a geo-lift test with a CTV provider beside a mix-model pilot. A beauty group whose mix includes TV runs matched-market tests through Measured to read each platform’s incremental contribution.

The routes are layered more often than chosen. The financial-services company pairs iSpot with a Mutinex mix model; the mattress brand blends Innovid into its in-house model; the multi-brand beauty company tests and models at once. A company that describes one method may still use others, particularly where an agency runs the TV buy.

This is the average buyer's picture. What does your stack need?A free, peer-powered checkup returns a shortlist for your stack and your next steps.
Start my checkup →

Which CTV measurement tools do buyers use?

iSpot is the most-named independent TV measurement company in these interviews, used for linear and streaming TV by a financial-services company, an automotive-information company and a consumer-health device brand. VideoAmp, Innovid, Data Plus Math, Nielsen, Comscore, Samba TV and EDO are also named. Every TV measurement tool here is rated by too few buyers to publish a figure.

Built to measure TV

iSpot · VideoAmp · Innovid · Data Plus Math (LiveRamp) · Samba TV · EDO · Nielsen and Comscore (ratings currency)

Where TV also gets measured

CTV buying platforms: MNTN · tvScientific · Vibe.co · Tatari · Keynes Digital · The Trade Desk · mix models (in-house, Meridian, Robyn, Mutinex, Recast) · attribution tools (Rockerbox, Northbeam, Triple Whale) · clean rooms with publisher data · Kochava (mobile measurement)

The independent companies are praised for what platforms cannot give: measurement across linear and streaming in one place. The automotive-information company has used iSpot for “Seven, eight-plus years” to measure TV against web conversions, and names one gap: “we still are not getting enough visibility into competitive data as to how our ads are performing versus competitors' ads.” It was looking at EDO, Samba TV, VideoAmp, Comscore and Nielsen, an evaluation rather than a move. A media and streaming company uses VideoAmp for cross-platform measurement and finds the access slow: “we have to request analysis, or wait a long time before we can get any sort of data.”

Several of these companies have changed hands. Innovid completed its acquisition of the TV measurement company TVSquared in March 2022, and Mediaocean completed its acquisition of Innovid in February 2025, combining it with Flashtalking. LiveRamp bought Data Plus Math in 2019. Among the buying platforms, Pinterest completed its acquisition of tvScientific on February 17, 2026, and the platform now markets itself as tvScientific by Pinterest; Walmart completed its acquisition of Vibe.co on August 4, 2026, making it part of Walmart Connect. An ownership change does not by itself change measurement quality or contract terms, but it is worth checking when a contract names one of these companies. Where the question is the clean room rather than the measurement, see what a data clean room is.

Where does CTV measurement fall short?

At the step from exposure to cause. Most of the methods above can show that a household saw an ad and later bought; buyers’ complaints are about what that does not prove, and about the data that never arrives.

View-through credit is the most-named gap. An enterprise software company chose tvScientific because it ties conversions back to the ad, then found “Since it's view-through, it's still been difficult to determine if tvScientific campaigns have been successful or not.” An outdoor-sports manufacturer happy with Vibe.co wanted a holdout group built into the tool in August 2024, because the platform’s return on ad spend counted everyone who saw the ad and bought: “I don't have a control group.” That describes its setup then; Vibe.co’s site said in June 2026 that incrementality testing is built in, run through an integration with Haus. A lifestyle apparel retailer advertising on Hulu says “we can get data back from Hulu on who saw the ad, but there's no click and buy from Hulu.”

The screen and the sale are different devices. The media and streaming company has a decent read of its connected-TV audience and of its mobile audience, but says “We do not have a great understanding of tying those two things together into a single user”. The regional retailer wants a single view of impressions, reach and frequency across YouTube, CTV, Meta and TikTok, and does not think one exists. Exposure is observed at the household or device level, while the conversion happens on a phone, a laptop, a payment account or in a store, so linking them needs shared or inferred identifiers. What share of exposures and outcomes were matched, how unmatched records are treated, and whether a result reports observed matches or estimates extended to a wider population all decide how much it says.

Tests cost more than smaller brands can spend. A direct-to-consumer wellness-equipment brand with a long consideration window dropped a planned incrementality test on CTV because “the amount of data that was needed and the budget that was needed to even get a stat sig read was way too high”, and launched CTV with tvScientific and a different measurement approach. A jewelry brand ran a linear TV test whose result was too close to call, and pulled back. A health brand’s analytics director says view-based channels such as linear TV are much harder to run experiments in than walled-garden platforms with built-in tests. None of these is a verdict on TV; each is one company’s result.

Want this checked against your own stack?

Start my checkup →

Should you buy a CTV measurement tool?

Only for a question your buying platform cannot answer, such as reach across publishers or across linear and streaming. Independent analysis is more transparent, but it does not by itself make attribution unbiased or prove lift, and whether CTV earns its budget takes a controlled test or a calibrated model.

Our recommendation, not a practice buyers describe as a method: start from the measurement question, not the vendor. For exposure, decide whether you need impressions or deduplicated household reach, frequency, and coverage across publishers and linear TV. For attributed outcomes, decide which conversions count, how households or devices are matched, and over what window. For incremental impact, decide whether a controlled experiment is feasible, or whether a modelled estimate with calibration and transparent assumptions will do. A separate provider is most useful for what the buying platform cannot give: measurement across publishers or across linear and streaming, more exposure data, analysis by someone other than the seller, and control over the method. Which one fits depends on the question left unanswered and the data the provider can reach. Keep platform-reported conversions apart from measured lift, and do not add a CTV platform’s view-through conversions to other channels’ conversions as one count, since one sale can be credited twice.

Run the test as a test: define the outcome and the design before launch; randomize households or regions where you can, or document how matched markets were chosen; estimate the sample size, sales volume and duration needed to detect an effect worth acting on; watch for contamination between groups and hold other marketing constant; read the outcome in your own transaction data; and report the lift with its uncertainty, including an inconclusive result. A planned length does not guarantee a significant read. Where a powered test is beyond budget, as it was for the wellness-equipment brand above, which also judged itself too small for a mix model, treat platform attribution as a directional signal rather than as incremental return. A test that shows no distinguishable lift may be inconclusive rather than proof that TV did nothing; carry that uncertainty into the spending decision.

In a trial of any measurement provider, ask where its exposure data comes from, which publishers and devices it covers, how it deduplicates reach and frequency, and which households are observed rather than modelled; how it matches exposure to your web, transaction and store data, with its match rates, windows, treatment of unmatched events and privacy limits; whether you can see its assumptions, outputs and validations; and how quickly results arrive. After an acquisition, check whether data access, contract terms or workflows changed.

What the interviews do not show is that any one method is accurate, that independent measurement produces better budget decisions than platform reporting, or that the brands reporting good CTV results would see the same lift in a controlled test. The shares above count companies that describe their measurement, not the market.

For the methods underneath, see what marketing mix modeling is, what incrementality testing is and what marketing attribution is; for how companies combine them, unified measurement vs multi-touch attribution; and for TV bought through a retailer, what retail media is.

Common questions

What is CTV measurement?

CTV measurement is how a brand assesses what its connected-TV ads, the ads shown inside streaming apps on a television, delivered. It answers three separate questions: which households or devices saw the ad and how often, what they did afterward, and whether the ad added sales that would not have happened anyway. Most tools answer the first two by matching exposure to later visits or purchases, which shows an attributed relationship, not necessarily cause. Controlled experiments give the strongest direct evidence on the third; mix models can also estimate it, under assumptions.

How is CTV advertising measured?

In six ways in these interviews: the buying platform’s own reporting, a marketing mix model, an independent TV measurement company or a publisher’s exposure data matched to the brand’s sales, an attribution or analytics tool, response signals such as traffic spikes after an airing or promo codes, and controlled experiments such as geo-lift tests. The first two are the most common: about a third of the companies that explain how they measure TV use or are testing the platform’s reporting, and a similar share a mix model. Many combine methods, so the shares overlap.

What is the difference between CTV measurement and digital attribution?

Digital attribution credits conversions that follow a tracked click. A TV ad has no click, so CTV measurement matches the household or device that saw the ad, using IP addresses, device graphs, publisher data or clean rooms, to visits or purchases within a set window, which is called view-through attribution. That credit shows the ad was seen before the sale, not that it caused the sale, and the match itself can be incomplete or wrong. Buyers here describe click-based attribution tools as missing TV, and some distrust view-through credit.

What is ACR data?

ACR, automatic content recognition, identifies the content playing on participating smart TVs by matching audio or video samples against a reference library of programmes and commercials. Measurement providers use consented ACR data to estimate which households were exposed to an ad on linear or streaming TV across the devices it covers, then link them to later web visits or sales. It does not cover every TV or viewer, so wider reach needs weighting or estimation, and it does not show which person watched, whether they paid attention, or whether the ad caused anything.

Should you use MMM or incrementality testing for TV?

They estimate TV’s effect in different ways, and some buyers here use tests to check the model. A mix model estimates TV’s contribution from aggregated spend and sales over time, needs no household matching, and covers linear and streaming alike, but its estimate depends on its assumptions and can confuse TV with seasonality when TV runs in peak weeks. A geo-lift or holdout test gives more direct evidence of the lift TV caused in one period, but it needs enough budget, time and sales to read a result, which stopped one smaller brand from running one.

Which CTV measurement tools do buyers use?

iSpot is the most-named independent TV measurement company in these interviews; VideoAmp, Innovid, Data Plus Math, Nielsen, Comscore, Samba TV and EDO are also named. Many companies instead rely on the reporting inside the platform they buy CTV through, such as MNTN, tvScientific, Vibe.co, Tatari, Keynes Digital or The Trade Desk, or on a mix model. Each TV measurement tool here is rated by too few buyers to publish a figure.

Why is CTV hard to measure?

Because there is no click, and the person who sees the ad often buys on another device. Buyers here describe view-through credit they cannot trust, attribution tools built around clicks, platforms that do not pass viewing data, exposure and mobile data they cannot tie to one person, and experiments too costly or slow to reach significance. Several combine methods rather than trusting one.

Get a checkup on your tech stack

A checkup for your marketing stack, built from hundreds of verified buyer interviews: what matters most to you, a shortlist for your stack, and your next steps.

Start my checkup

Free · no password · your answers are anonymized before they inform research like this.

Not ready for an interview?Get the one-page outline — the six ways companies measure TV and streaming ads, and the question each one can and cannot answer.
Methodology. Alium conducts verified interviews with software buyers — the marketing, media, growth and analytics leaders who buy and measure advertising. This page draws on the buyer speech in every published interview that mentions connected, streaming, OTT or linear TV, or names a TV measurement or CTV buying platform (iSpot, VideoAmp, Samba TV, Innovid, Data Plus Math, MNTN, Tatari, tvScientific, Vibe.co, Keynes Digital and others), read with the interviewer question above each answer so that every quote is about the product named. Method shares count the companies whose interviews describe how they measure, or tried to measure, TV or streaming ads; each company is counted once for each method it describes in use or under test, and a company can use several, so the shares do not sum to one. Methods a company only considered are left out. These are not market shares or adoption rates. Product descriptions come from each vendor’s own site, and acquisitions from company filings and newsrooms, as of November 2026. The TV measurement tools on this page are rated by too few buyers to publish a figure, so it carries no ratings. Buyer identities are verified at interview time and anonymized before publication; vendor names are reported as given. No vendor paid to appear or was able to edit this page.