CTV is hard to measure because nobody clicks a TV. Last-click reports and Google Analytics give the credit to branded search, direct traffic or social, so CTV looks like it did nothing. CTV platforms go the other way: they count any visit or sale from a household that saw an ad, including people who were going to buy anyway, so CTV can look better than it is. The way to know what CTV really adds is to compare markets or households that saw it against ones that didn't, and judge it on your total sales, not on any one platform's report.
Why CTV attribution is so hard
Most of the problems come from the same place. CTV works like TV, but it gets reported like digital, and digital reporting was built around clicks.
The sale shows up somewhere else
Someone sees your ad on the TV, then searches your brand on their phone or types in your URL a few days later. Google Analytics and last-click reports give that sale to branded search or direct, not CTV.
The TV and the phone don't always match
CTV is bought and tracked by household, usually through the home's IP address. The person who saw the ad may buy on a work laptop or on cellular data, and the match between the TV and that device is never perfect.
Platforms count sales they didn't cause
CTV platforms report conversions from households that saw an ad within a set window, often 7 to 30 days. If you target people who already know you, many of those sales would have happened without the ad.
The platform totals add up to more than you sold
Your CTV platform, Meta and Google each report the same sale as their own. Add their numbers together and you'll usually get more conversions than your business actually had.
There are smaller issues too. Reporting varies from one streaming publisher to the next, some publishers share very little data, and not every impression sold as CTV runs on a TV screen. Ask for a device breakdown so you know where your ads actually ran.
What CTV reports can and can't tell you
| Metric | What it tells you | What it doesn't |
|---|---|---|
| Impressions and completion rate | Your ad ran and people watched it to the end | Whether watching it changed what anyone did |
| Reach and frequency | How many households saw it and how often | Whether you reached new people or the same customers again |
| View-through visits | Households that saw the ad later visited your site | Whether they would have visited anyway |
| Platform-reported ROAS or CPA | The platform's own estimate of the sales it drove | How many of those sales were new, and how much of that credit Google and Meta are also claiming |
| Lift vs. a holdout group | How much more the exposed markets or households did than the ones that didn't see it | This is the number to plan budget around |
How to measure what CTV actually adds
Write down your baseline before you launch
Record sales, new customers, branded search and direct traffic by market for at least a few weeks before CTV starts. Without a clear before, there's nothing to compare the after to.
Hold out some markets
Run CTV in a set of markets and keep similar markets dark. If the CTV markets grow faster than the dark ones, that difference is what CTV added. This is the most reliable read most brands can get.
Watch branded search and direct traffic by market
These usually move first. A rise in branded searches in your CTV markets, and not in the others, is an early sign it's working.
Tighten the platform settings
Shorten view-through windows, exclude existing customers where it makes sense, and ask for reporting by device and publisher. It won't fix attribution, but it makes the platform's numbers closer to reality.
Use household holdouts where you can
Some CTV platforms can hold back a random group of households from seeing your ads. Comparing them with the households that did see ads gives you a cleaner read than view-through reporting.
Judge CTV on total results
Look at total revenue and cost per new customer across all channels in your test markets, not CTV's own report. If CTV is working, your search and social should look better in those markets too.
Give it enough budget and time
A test that's too small or too short produces noise. Plan a budget that can actually move results in your test markets, and run it long enough to cover your normal buying cycle.
- CTV's ROAS beats every other channel. Especially if total sales didn't change when CTV launched.
- Long view-through windows. A 30-day window credits CTV for a lot of sales it had nothing to do with.
- No control group. Without markets or households that didn't see the ad, there's no way to know what it added.
- Retargeting dressed up as CTV. If CTV is mostly reaching people who already visited your site, it will look great and add very little.
- No device or publisher breakdown. If the platform can't show where your ads ran, you can't tell how much ran on a TV.
Plan CTV so it can be measured
The best time to fix CTV measurement is before the first ad runs. Plan CTV alongside search and social so you can see how it affects both, pick test and holdout markets up front, and set up tracking for branded search, direct traffic and sales by market. A simple vanity URL, QR code or offer in the creative can help too, as long as you treat it as a floor and not the full count.
We'll review your CTV reporting, show you what it's really telling you, and set up a test that proves what CTV adds. CTV & streaming →