Did the TV Spend Do Anything?

CTV is the easiest ad spend a dealer will ever approve and the hardest to defend three months later. The pitch writes itself: your commercial, on the big screen, in the living rooms of your own zip codes, without buying the whole DMA like the broadcast days. Then the quarterly review arrives, someone asks what it did, and the room goes quiet in a very specific way / because the channel with the most beautiful creative has the least convincing receipts.

Here is the honest version of that conversation: what CTV measurement actually is, which numbers are theater, and how a dealer can know / really know, at budget-decision quality / whether the spend did anything.

Why this channel resists measurement by design

Every hard problem in ad measurement shows up in CTV at once, structurally. There is no click / nobody taps a sofa. The screen is shared, so an "impression" landed on a household, not a person, and the person it influenced may be three devices away from the one that saw it. The inventory is fragmented across streaming apps and platforms that do not share user-level data with each other or with you. And the outcome you care about happens weeks later, offline, at a desk in your showroom. Digital measurement grew up assuming a clickable, individual, trackable journey. CTV has none of the three. This is not a maturity problem that another year of ad tech fixes. It is the shape of the medium.

The numbers that are theater

Completion rate is delivery QA, not proof of impact. CTV decks lead with it because it is always gorgeous / streaming ads are mostly unskippable, so of course completion is high. It tells you the file played. A number that is high for everyone, every time, regardless of outcome, is a receipt that the vendor did the shipping. It says nothing about whether anyone drove anywhere.

Vendor-matched conversions are the thermometer grading itself. The standard "proof" is a match report: households exposed to your ads, matched against website visits, foot traffic, or sales, delivered by the same company that sold you the media. Set aside the fuzziness of household-level matching and one structural problem remains / the people your targeting reached were selected for being likely buyers, so some of them were always going to buy. A match report counts them all as wins. Without a comparison group of similar households who did NOT see the ads, "exposed and then purchased" is not causation. It is your targeting criteria, read back to you as a victory.

Dashboards inherit everything broken about attribution, minus the clicks. Whatever skepticism you apply to self-attributed platform reporting elsewhere, apply double here, where there is no click to anchor even the weak version of the story.

How to actually know

The honest answer is incrementality, and CTV has a saving grace: the same geographic precision that makes it buyable makes it testable. You cannot run TV to one household and not its neighbor / but you can run it to one market and not its twin.

Design the test before the buy. Pick matched markets / similar stores, similar volume, similar seasonality. Run CTV in some, hold the others dark, keep everything else constant, and read the DMS and your brand gauges after enough weeks for a car-buying cycle to turn. If the exposed markets outgrow the dark ones beyond their historical relationship, the spend did something, and you can even price what. If they do not, you have learned something worth exactly as much. A dealer group with multiple rooftops can do this properly; a single store can approximate it with on/off periods over time / cruder, slower, but still evidence rather than allocation.

Watch the gauges that move when awareness moves. Branded search volume, direct website traffic, "how did you hear about us" at the desk, walk-ins who never filled a form. CTV is upper-funnel brand media; these are the instruments that flex when brand media works, and they flex within weeks, not quarters. None is proof alone. Moving together, alongside a holdout test, they converge on the truth.

Use the weak signals for texture, not verdicts. A vanity URL or QR code on screen catches a sliver of response / treat it as a floor, never a measure. Frequency reports tell you whether you bought presence or wallpaper: reach spread thin across a whole DMA did nothing for anyone, and the same dollars concentrated on your actual selling radius at real frequency is the version that has a chance. Check that before judging the channel / an unmeasurable buy and a bad buy look identical in a quarterly review.

Our position

CTV for dealers is a legitimate channel wrapped in illegitimate reporting. The medium is real: television-quality attention, bought with zip-code precision, at entry costs the broadcast era never offered a single rooftop. The default measurement is not real: completion theater and self-graded match reports, sold by the counterparty, designed to make renewal feel responsible.

The fix is to stop asking the channel to grade itself and design the exam yourself: matched-market holdouts as the backbone, brand gauges as the early warnings, DMS matchback as the scoreboard, and vendor studies filed under sales collateral where they belong. Measured that way, CTV earns a place in a dealer's mix more often than the skeptics expect / and gets caught doing nothing faster than the enthusiasts would like. Both outcomes are wins. The only losing move is the current default: beautiful creative, gorgeous completion rates, and a quiet room every quarter when somebody asks the only question that matters.