What a First CTV Buy Actually Costs, and How to Judge It

In September 2024 Roku launched a self-serve ads manager with a Shopify integration for shoppable campaigns, aimed explicitly at growth marketers rather than agencies (Roku’s announcement). Paramount and Hulu run their own self-service buying platforms. Whatever else is true about connected TV, access is no longer the hard part, and anyone still writing about it as an emerging opportunity has not opened one of these interfaces.

The question that replaced access is duller and more useful. What does a first CTV campaign actually cost, and how would you know whether it worked?

The media is not the expensive part

Three costs sit inside a first buy, and most first-time advertisers budget for one of them.

The media is the line everyone plans for. The creative is the line that surprises them, because assets built for feeds do not work on a television: vertical crops, burned-in captions, corner logos and phone-sized text all announce themselves as repurposed the moment they appear on a large screen in a living room. If your video library was made for social, the honest budget includes producing something built for the screen it will run on, and that cost arrives before any impressions do.

The third line is measurement, and it is the one that decides whether the first two were worth spending. It costs money in two ways: any verification or incrementality tooling you buy, and the revenue you deliberately forgo by holding markets out of the campaign so you have something to compare against. Both are real costs. Neither appears on a media invoice.

The platform minimum is not your budget

Self-serve platforms have pushed entry minimums low enough that they no longer tell you anything useful. A number you can put on a card is not the same as a number that can move a business, and treating the minimum as the budget is how most first tests end inconclusively.

The floor that matters is arithmetic on your own numbers. Pick one contained market rather than a country. Estimate the households you want to reach in it. Decide how many times each needs to see the spot across the flight for the message to register, which for an unfamiliar brand is several rather than one. Multiply through to get impressions, then apply whatever CPM you have actually been quoted. That figure is your floor, and if it exceeds what you were willing to spend, the correct response is a smaller market rather than a thinner buy.

The classic first-buy failure is the opposite. A modest national budget produces a whisper everywhere and a voice nowhere, delivers a frequency of roughly one against a huge audience, and returns a flat result that gets read as proof the channel does not work.

Ask what share of the dollar becomes working media

The CPM quoted to you is rarely the whole cost. Depending on the route you buy through, some combination of demand-side platform fees, audience data fees, ad serving, verification and agency management sits between your budget and the impression.

Buying direct through a publisher’s self-serve manager removes several of those layers and limits you to that publisher’s inventory. Buying through a DSP gives you reach across publishers and adds intermediaries. Neither is wrong, but the comparison is meaningless until you ask both routes the same question: of every hundred dollars, how many become working media, and what does the rest buy?

Ask for that breakdown in writing before signing, alongside device and app-level reporting. A partner who cannot produce a gross-to-net figure is either not in a position to see it or would rather you did not.

Decide how you will judge it before you spend anything

Connected TV cannot be judged the way search is judged, because there is no click and no clean path from exposure to conversion. That constraint, and what the terminology actually means when a vendor uses it, is the subject of the difference between a screen and a delivery method, and it is worth settling before a media conversation rather than during one.

For a first buy, the practical design is a matched-market holdout. Run in your chosen markets, deliberately do not run in comparable ones, and compare totals rather than campaign-reported numbers. Track branded search volume in both, because people looking you up afterwards is the most reliable fingerprint a television campaign leaves, and it is the connection between the living room and the search bar that makes CTV legible to a performance team at all. Watch direct and organic sessions the same way.

Vendors are aware this is the weak point and are building toward closing it, which is what the streaming platforms are now selling. Use those tools, and hold them to the same standard as everything else: a measurement method that cannot return a negative result is not measuring anything.

What a sane first test looks like

One or two contained markets. One offer, stated plainly, with a QR code if you want a direct path. A control group of comparable markets that stays dark. A flight long enough for a lagging signal to appear rather than four weeks judged against a lower-funnel benchmark. And a decision rule written down before launch: what result makes you spend more, what result makes you stop, and who decides.

That last item is the one people skip, and it is the reason so many CTV tests end in a meeting where everyone argues about a dashboard. Deciding the threshold in advance costs nothing and it is the difference between a test and an expenditure. Where this sits alongside search, social and everything else you already run is a planning question rather than a channel question, and it belongs in the same media program as the rest.

Frequently Asked Questions

Can a small business genuinely afford CTV now?

Access is affordable; a meaningful test is a separate question. Self-serve platforms will accept small budgets, but a campaign spread thinly across a large area produces a frequency too low to register and a result too flat to read. A small business is usually better served by concentrating a modest budget on one city or region, where the same money buys enough repetition to matter and enough contrast against a held-out market to prove something.

Do we need a proper TV commercial, or can we use video we already have?

Existing video can work if it was shot for a horizontal frame and holds up without sound-off captions and interface furniture. Repurposed social cuts generally do not, because everything that makes them effective in a feed marks them as out of place on a television. The realistic minimum is one well-produced spot in the standard lengths, made to be watched by more than one person, in a lit room, from across it.

Should we buy direct from a platform or through a DSP?

Direct self-serve is simpler, has fewer fees between your budget and the impression, and confines you to one publisher’s inventory. A DSP buys across publishers, which matters for reach and for frequency management, and adds fee layers plus a steeper operational learning curve. For a first test in one market, direct is usually the faster path to an answer. For sustained buying across several services, the DSP route starts to earn its overhead.

How long should a first flight run?

Long enough for the effect you are measuring to appear, which depends on your own sales cycle rather than on a standard. Set the flight so that a customer who sees the ad in the first week could plausibly have bought by the last, then add time for the signal to accumulate. Anything shorter measures the speed of your purchase cycle rather than the value of the channel.