CTV Measurement Companies Compared: Nielsen, Comscore, VideoAmp, iSpot, EDO, Samba and the Verification Vendors

Most confusion about CTV measurement comes from comparing companies that answer different questions. The table sorts them by question first.

A buyer usually needs one answer from each row. A currency number without verification can count bots. A verified impression without an audience number says nothing about who watched.

The table covers the six audience and outcome companies named most often in CTV buying, and the four verification vendors with CTV accreditation. Every MRC status was read on the MRC's lists on 9 November 2026. "Not listed" means the company appeared on neither the accredited nor the under-review lists that day; it does not mean a service is inaccurate.

Publishers and ad servers hold CTV accreditations too, for counting their own impressions. The MRC list includes rendered CTV video impressions for Netflix, Hulu, FreeWheel and Innovid, among others (MRC digital list). Those numbers are what a buyer is invoiced on, so they are the base every vendor above measures against.

Four data sources sit under almost every CTV number. Each sees something the others miss.

1. Panels. A sample of homes agrees to be metered, so the measurer knows who is in the room. Panels give person-level demographics but are small for niche programmes.
2. Automatic content recognition (ACR). Software in a smart TV matches what is on screen to a reference library, whatever the source. ACR data covers millions of sets, but records the household's TV, not the person watching.
3. Set-top box return-path data. Cable and satellite boxes report tuning back to the operator. The data is large and detailed, but misses homes without a box and apps watched outside it.
4. Server and ad logs. Ad servers, SSAI servers and verification tags record each ad call. This is the only source that sees every impression, but it identifies devices and IP addresses, not people.

Nielsen's accredited product shows how they combine. It pairs a representative panel with return-path data and ACR data from partners including Comcast, DirecTV, Dish, Roku and Vizio. It then uses the panel to calibrate the big data and assign viewing to people (Nielsen, 22 January 2025).

Nielsen measures streaming inside its national TV service, so CTV viewing of TV programming counts toward its ratings. Its accreditation covers national TV deliverables from the panel alone and from Big Data + Panel. The MRC's note adds first-party streaming data for some live events, at present Amazon Prime's Thursday Night Football, the NBA, NASCAR and Netflix's Christmas Day NFL games (MRC television list).

The exclusions matter to digital buyers. The accreditation does not yet cover Nielsen's digital TV ratings (dTVR) or its national out-of-home expansion. The out-of-home work was listed as under review on 9 November 2026 (MRC services under review, television). Nielsen ONE, its cross-media platform, had not been submitted to the MRC when Big Data + Panel was accredited (Nielsen, 22 January 2025).

So a Nielsen rating for a streamed football game can be accredited currency, while a Nielsen cross-media reach report for a CTV campaign may not be. Ask which product produced the number.

The three "alternative currencies" compete with Nielsen for TV and streaming deals. The JIC, formed in January 2023 by media buyers and video publishers, certifies them as currencies (JIC via PR Newswire). The 2026 evaluation ran 672 tests over more than 150 million telecast observations. All three met the threshold for Total Households, Advanced Audiences and Personified Demographics.

JIC certification and MRC accreditation test different things. The JIC asks whether a currency is ready to trade on, judged by its buyer and seller members. The MRC audits a service's methods and controls against its minimum standards.

• Comscore holds MRC accreditation for national and 210 local market household ratings, with persons-level and commercial estimates outside the scope (MRC letter, 31 March 2025). It sold its movies business, including box office measurement, on 27 May 2026 for $70 million (Comscore 8-K, 2 June 2026).
• VideoAmp describes "census-level, cross-screen measurement" built on direct integrations with publishers and platforms (company description, VideoAmp). It did not appear on the MRC's lists on 9 November 2026.
• iSpot is accredited by the MRC only for its ad occurrence data: which national ads aired, where and when, for spots of at least 9.1 seconds (MRC letter, 1 October 2025). Its audience and attention metrics sit outside that scope.

EDO measures what happens after an ad airs rather than who watched it. It matches every ad airing across networks and streamers to online engagement, such as searches for the brand, and treats that engagement as a leading sign of sales (EDO). Edward Norton and Daniel Nadler founded it in 2015 (EDO). EDO says its outcome signals plug into Amazon DSP, The Trade Desk and Google DV360 for programmatic TV buys (company claim, EDO).

Samba, formerly Samba TV, sits between audience and outcomes. It says it started in 2008 with software for smart TVs and now holds first-party TV and web data linked to 1.5 billion people (company claim, Samba). It bought Bestever AI, a generative AI ad platform, on 22 June 2026 (Samba, 22 June 2026).

Neither EDO nor Samba appeared on the MRC's lists on 9 November 2026. That is common for outcome tools. Outcome numbers depend on each advertiser's own conversion data, which makes them harder to audit as a syndicated service. Treat them as directional and test them with a holdout, in line with the incrementality methods buyers already use.

Verification vendors check the impression itself. In CTV, the main threat is servers that pretend to be TVs. With server-side ad insertion (SSAI), the ad request comes from a server on the TV's behalf, so a fraudster can run fake SSAI servers for TVs that do not exist. HUMAN described one such scheme, NewsJunkie, on 7 July 2026; it reached nearly two billion invalid bid requests a day per seller at its peak, by HUMAN's count (PPC Land, 11 July 2026). Our engineering guide to SSAI and CTV ad serving shows the headers that SSAI servers should pass on.

The MRC accredits each vendor's CTV work metric by metric. This matrix shows what was accredited for CTV screens on 9 November 2026 (MRC digital list).

Each profile carries the full scope and limits: DoubleVerify, Integral Ad Science, HUMAN Security and Pixalate. For what viewability and attention metrics prove in general, see viewability, attention and brand safety explained. For the fraud side, see ad fraud detection tools compared.

Two vendors also add programme context. IAS Total TV reports genre, rating and show-level data from Disney, NBCUniversal, Paramount and Prime Video, plus publishers using its Publica ad server (IAS, 27 April 2026). Pixalate's OpenEPG returns show, genre and rating fields for a CTV bid request (company claim, Pixalate).

Take a streaming campaign of 2,000,000 impressions bought at an illustrative $30 CPM, so $60,000 of media. DoubleVerify's 2026 report found that, in campaigns without its pre-bid protection, 9% of impressions were not served to a real person (company data, PPC Land, 12 May 2026). At that rate, 180,000 impressions and $5,400 of spend would reach no one.

Now look at the three reports the buyer receives. The publisher's ad server bills 2,000,000 impressions. The audience currency reports reach based on panel, ACR or box data, which may not flag the 180,000 invalid calls. The verification vendor reports 9% invalid, but only on traffic it could measure. None of the three is wrong; each counts something different. The fix is to agree before the campaign which number is billed, which is net of invalid traffic, and who settles disputes. Our guide to log-level data explains how to keep your own record to check against.

1. Name the question. Decide whether you are buying audience, outcomes, attention or verification, and do not let one vendor's number stand in for another's.
2. Pick the currency with your sellers. Agree whether the deal trades on Nielsen or a JIC-certified alternative, and which product and demographic.
3. Read the accreditation letter, not the logo. Check the metric, the screens and the notes, such as "certified apps only".
4. Add an accredited CTV fraud check. Use a vendor accredited for SIVT on CTV, and pre-bid filtering where your DSP supports it.
5. Ask how SSAI traffic is handled. Require SSAI headers to be passed on and ask whether the vendor's SSAI metrics are accredited.
6. Test outcome tools with a holdout. Treat outcome scores as hypotheses until a controlled test confirms them.
7. Keep your own logs. Store impression-level data so you can reconcile vendor reports.

Measurement costs and gaps also show up in how much of a CTV budget counts as working media. Our page on working media vs non-working media has the CTV cut, and the RTB architecture hub links the related engineering pages.

Until now, the TV currency and the CTV verifier were different companies. On 6 August 2026 Nielsen's parent agreed to buy DoubleVerify for $13.60 a share, an enterprise value of about $2.15 billion (Nielsen and DoubleVerify, 6 August 2026). DoubleVerify's stockholders approved on 29 October 2026, and the company expects to close by the first quarter of 2027 (DoubleVerify Q3 2026 release).

The companies say they will keep both operating structures independent and aim to score media "on both audience delivery and media environment quality" (company statements, same source). PPC Land noted that one company would then produce both the audience number and the quality check on the impressions behind it (PPC Land, 7 August 2026). With IAS owned by private equity since December 2025, buyers who want a verifier separate from their currency can still choose IAS, HUMAN or Pixalate. Each holds its own MRC accreditation.

1. Media Rating Council, accredited television services (accessed 9 November 2026). https://mediaratingcouncil.org/accreditation/television
2. Media Rating Council, accredited digital services (accessed 9 November 2026). https://mediaratingcouncil.org/accreditation/digital
3. Media Rating Council, television services under review (accessed 9 November 2026). https://mediaratingcouncil.org/accreditation/services-under-review/television
4. Media Rating Council, Nielsen National TAM accreditation letter (22 January 2025). https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/Nielsen/GI012225_Nielsen%20National%20Accreditation%20Letter.pdf
5. Media Rating Council, Comscore TV accreditation letter (31 March 2025). https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/Comscore/GI033125_Comscore%20TV%20Accreditation%20Letter.pdf
6. Media Rating Council, iSpot.tv ad occurrence accreditation letter (1 October 2025). https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/iSpot%20TV/GI100125_iSpot%20Ad%20Ocurrence%20Service%20Accreditation%20Letter.pdf
7. Media Rating Council, DoubleVerify accreditation letter (13 April 2026). https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/DoubleVerify/GI041326_DoubleVerify%20Accreditation%20Letter.pdf
8. Media Rating Council, Integral Ad Science accreditation letter (9 July 2026). https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/IAS/GI070926_Integral%20Ad%20Science%20Recurring_Google%20_Meta%20Accreditation%20Letter.pdf
9. Media Rating Council, HUMAN accreditation letter (14 April 2026). https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/HUMAN/GI041426_HUMAN%20Accreditation%20Letter.pdf
10. Media Rating Council, Pixalate accreditation letter (8 April 2025). https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/Pixalate/GI040825_Pixalate%20Accreditation%20Letter.pdf