The three kinds of numbers in a CTV report
| Layer | Examples | How real it is |
|---|---|---|
| Counted | Impressions, households reached, frequency per household, completion rate, the apps the spot ran on, cost | Measured by the platform. Tells you the ads ran and where. |
| Matched | Website visits, leads and showroom visits from households that saw the spot | Inferred by matching IP addresses, device graphs and mobile location data to the exposed households. A correlation. |
| Caused | Extra calls, searches, visits, leads and sales compared with a holdout | The only layer that answers whether the ads changed anything. |
Vendor reports tend to lead with the counted layer because it always looks good. Completion rates on streaming run above 90% because most spots cannot be skipped, so a high completion rate describes the format, not the campaign.1 Some dealer CTV sellers leave clicks off their reports entirely, because most TV screens are not clickable and click numbers would describe a small slice of the audience.2
Set up a ZIP-code holdout before you launch
- Pick comparable ZIP codes. From the ZIP codes where your buyers and service customers live, choose a few pairs that look alike in households, sales history and distance from the store. One of each pair gets the ads; the other is held out.
- Exclude the holdout ZIP codes from the streaming targeting on every platform, and keep them out of any other new campaign during the test.
- Record a baseline. Four weeks before launch, log calls, branded searches, website visits, leads and showroom visits by ZIP code.
- Run the flight for six to eight weeks. Shorter flights rarely produce enough calls and visits to compare.
- Compare the change, not the totals. If exposed ZIP codes rose 20% over their baseline and holdout ZIP codes rose 5%, the difference is your best estimate of what the ads added.
A holdout will not be perfect: people drive across ZIP lines and see ads on other screens. It is still far better evidence than a report that only counts impressions.
What to track at the store
| Metric | Where it comes from | Why it matters |
|---|---|---|
| Phone calls | Call tracking numbers on the website and in Google Business Profile | TV makes the phone ring first; calls are the quickest signal. |
| Branded searches | Google Search Console and Google Ads search terms for your store name | People search the store they just saw on TV. |
| Direct and organic site visits by city | Google Analytics 4 | A lift in the exposed area that the holdout does not show. |
| Vehicle detail page views | Google Analytics 4 and the Meta Pixel | Shopping intent, and the audience for Meta inventory ads. |
| Leads and showroom visits by ZIP | Your CRM and the showroom log | The comparison the holdout is built for. |
| Sales by ZIP | Your DMS, 30 to 90 days after the flight | The slowest signal and the one that pays for the next flight. |
When a vendor matches visits or sales back to exposed households, ask for the attribution window. Automotive vendors commonly use 14 to 30 days because buyers take time to decide.3 A longer window credits more visits to the ads; it does not make the ads work better.
Questions to ask any streaming TV vendor
Dealers say reporting is where streaming TV breaks down: in a 2026 survey of dealership marketers, 64% said they were frustrated by reports listing channels they did not recognize.4 Ask these before you sign:
- Which apps and publishers will my spot run on, by name, and can I see that list every month?
- What is the CPM by app, and what is the minimum spend?
- Whose account is the campaign in, and can I log in and see it myself?
- Is frequency capped per household across all apps, or per app?
- Will you run a holdout, and who chooses the holdout ZIP codes?
- How are website and showroom visits matched to households, and with what attribution window?
- Can I export the log-level data when the contract ends?