Connected TV promised the measurability of digital with the impact of television. What it delivers is messier. CTV measurement is broken in three places: inventory is spread across many platforms, each platform counts in its own way, and the link between an ad seen on a living-room screen and a purchase made days later is still hard to prove.

Brands that run CTV next to performance channels feel the gap most. Search and affiliate report sales line by line. CTV reports impressions. When budget is split by what each channel can prove, CTV loses money it may deserve, or wins money it does not.

CTV sits between two measurement worlds

Traditional TV measurement was imprecise but understood. You bought ratings points, accepted that attribution was loose, and measured brand lift over time. Digital performance channels went the other way: click-level attribution, ROAS by the day, deterministic tracking.

CTV inherits the worst of both. It promises digital measurement because it is delivered over the internet. But the viewing context, a shared screen in a living room, with no click and no cookie, makes deterministic attribution close to impossible.

So platforms report impressions as delivered and completed. Linking them to purchases takes probabilistic matching, household graphs built on IP addresses, or panels. None of these is as reliable as a click, and all of them cost more than most mid-market brands can justify.

Every platform measures its own way

Each large CTV platform measures differently and shares different data:

  • A retailer's streaming service reports the sales made in its own store, not the ones on your site.
  • A large video platform reports view-through inside its own ecosystem and struggles beyond it.
  • A TV maker's own channels have deep data, about their own screens only.

For a brand running CTV on several platforms at once, there is no single view of performance. The pieces of the puzzle come from different boxes.

That has a cost. The platform that reports the most conversions tends to get the most budget, whether its method is more accurate or simply more generous. A platform that can attribute purchases inside its own marketplace looks better than one that cannot. That is a measurement advantage presented as a performance advantage.

What you can do now

The industry will not fix CTV measurement soon. These four practices work without waiting for it.

  • Test incrementality instead of chasing attribution. Run CTV in some markets, hold it out of comparable ones, and measure the difference in sales. This is how TV measurement has always worked at its best. Geo-lift tests and matched-market analysis are well established, and they depend on no platform's own measurement.
  • Watch branded search. If CTV works, people who see the ad search for the brand afterwards. Track branded search by market, comparing markets with CTV and markets without. It is not perfect, but it is a signal you own.
  • Manage frequency across platforms. Frequency is the one thing you can control everywhere. Without a cross-platform cap, some households see the same ad dozens of times while others see it once. Even a rough cap, deduplicated by household, improves efficiency.
  • Allocate by verified reach, not reported conversions. Fund the platforms that deliver the most unduplicated households for the money. It removes the bias of each platform's method.

Before any of this, agree with each seller what counts: whose numbers are used, the tolerance between counts, and what a shortfall is. CTV insertion order terms lists what to settle before a campaign runs.

Close the loop where you can

The link most brands miss is the one between CTV and their own commerce. An ad that makes a viewer search for the brand is valuable. An ad that sends them to a specific product page, on a path you can track, is more valuable, because it closes the measurement loop.

Feature a specific offer, a short vanity URL or a QR code in the CTV creative, and route it through the tracking you already use for affiliate or performance. It will not capture every sale CTV influences. It adds a deterministic layer on top of a channel that is otherwise probabilistic.

Used together, the two kinds of signal make a budget case:

  • direct visits and sales from the vanity URL or QR code, counted one by one;
  • the lift in branded search in CTV markets against holdout markets.

CTV creates the intent. Your own tracking captures what follows.

Judge CTV on its own terms

A brand that judges CTV with the ROAS framework it uses for paid search will conclude that CTV does not work. It works differently: higher in the funnel, over a longer window, across a household rather than a device.

A geo test also needs enough spend per market, and enough weeks, to rise above the noise. Size the test before launch, not after, and agree in advance what result would justify scaling it.

Where Oderra starts

Oderra does not attribute sales. It fixes the layer underneath: what was bought and what was delivered. Skopa, the buying agent from Oderra, sends one brief to every matching seller and returns plans in the same structure, so reach and price compare line by line. Fanera, the selling agent from Oderra, reports delivery against the confirmed plan on a shared record both sides read, shortfall first, and people on both sides confirm each plan before it runs.