How to Run CTV Ads for DTC Brands in 2026

The DTC CTV playbook: four phases

The rest of this guide is the part you can run against. A CTV program for a direct-to-consumer brand moves through four phases:

• Plan sets the goal and the budget.
• Target decides who sees the ad, and it is where most of the ROAS is won or lost.
• Measure proves what the spend did.
• Optimize compounds the result over time. Treat them in order.

Let’s dive in to the four phase outline and playbook.

Phase 1

Plan the campaign

Start with the job the campaign has to do, because it dictates everything downstream. CTV serves three distinct goals for DTC brands, and they are not measured the same way.

• Prospecting: reach net-new households that look like your best customers, to grow the top of the funnel beyond what saturated social and search can deliver.
• Retargeting: re-engage people who visited your site or engaged with your brand but did not convert, this time on the living-room screen.
• Retention: stay in front of existing customers to drive repeat purchase, or deliberately suppress them to keep prospecting budget clean.

Set the budget to the goal. A prospecting test needs enough impressions to reach a real audience at a frequency that registers, generally three to six exposures per household across the flight.

Spreading a small budget too thin across too many markets is the most common way a first CTV test fails to show signal. Pick your strongest geographies, concentrate spend, and give the campaign room to be noticed.

Creative basics

CTV creative is usually a 15- or 30-second video at broadcast quality.

The 15-second spot is the most efficient for a product-focused message: open with a sharp hook in the first three seconds, show the product solving a problem, and close on the brand, the offer, and a clear next step.

The 30-second format buys room for storytelling and emotional connection, which suits brand building. Production no longer requires a six-figure budget; capable teams produce strong CTV spots in the low five figures, and AI-assisted tools for marketing have pushed the floor lower still. Whatever the budget, the creative has to earn the full-screen attention the format gives you.

Phase 2

Target the right audiences

This is the phase that decides your return, and it is where most CTV guides go vague. They tell you to “use first-party data” and “build lookalikes” without explaining the layer that makes any of it work.

That layer is identity resolution. A CTV impression is only as good as the platform’s ability to know which real household it is reaching, and to connect that household back to a person you actually care about.

Here is the problem with renting that layer from a buying platform, and the reason the Walmart-Vibe deal should make you think. When the identity graph belongs to the platform, your audiences are only as portable, as precise, and as durable as that platform allows.

Untitled is built the other way around. The identity layer is yours, and CTV becomes one of several channels you can activate on top of it.

Step 1: Unify your customer data in the identity graph

Your first-party data is scattered by default: a customer list in one place, email engagement in another, site behavior in a third, purchase history in a fourth. The same person shows up as several disconnected records. Untitled’s identity graph resolves those fragments against a graph of more than 300 million US individuals, collapsing the duplicates into a single, person-level view of who your customers actually are.

That unified base is what every good audience is built from. Feed a messy, fragmented list into a CTV platform and you get messy, fragmented targeting back.

Step 2: Build better segments and lookalikes

Once your customers are resolved to real people, customer segmentation gets sharper and more useful. You can separate one-time buyers from repeat purchasers, high-value from low-value, recent from lapsed, and target each with the right message instead of one blunt audience. Lookalikes improve for the same reason: a lookalike model is only as good as the seed it learns from, and a clean, deduplicated, person-level customer base produces a far more accurate prospecting audience than a raw list full of duplicates and partial records. Better seed, better lookalike, better prospecting ROAS.

Step 3: Expand to net-new demand with intent data

Lookalikes find people who resemble your customers. Intent data finds people who are showing the behavior of someone about to buy in your category right now. Untitled’s buyer intent data lets you expand beyond your known audience to reach in-market households beyond the reach of your CRM and your lookalikes, the demand that exists but that your first-party data cannot see on its own. For a DTC brand trying to grow past the ceiling of Meta and Google, this is how CTV adds genuinely incremental reach rather than re-touching the same people.

Step 4: Fuel retargeting with the Untitled ID Tag

Most site visitors never identify themselves. They browse, they leave, and a cookie-based pixel loses them almost immediately. The Untitled ID Tag identifies anonymous website visitors at the person level by resolving them against the identity graph, which means the audience you retarget on CTV is built from real people who were genuinely on your site, not a decaying pool of cookies.

That closes the loop: a visitor who looked at a product and left can be re-reached on the living-room screen, and because they were resolved to a person, that exposure can be tied back to what they do next.

Phase 3

Measure what actually happened on CTV

This is where the identity-led approach pays for itself. The standard CTV measurement playbook, the one most guides describe, leans on proxy signals because it has no other option: watch for a lift in branded search, watch for a bump in direct traffic, compare conversion rates between exposed and unexposed regions.

Those methods are not wrong, and geo-based incrementality testing in particular is a legitimate gold standard. But they are indirect. They infer that CTV worked rather than showing who it worked on.

When identity is resolved before the campaign runs, measurement changes character. Because the household that saw your CTV ad can be connected to a person, and that person can be connected to a conversion on your site, you get person-level attribution instead of regional inference.

You can see that the ad reached real prospects, that some of them came back, and that some of them bought. That is measurable ROAS on a channel that, for most advertisers, is still measured by vibes and search-volume charts.

None of this means you throw out incrementality testing. Run holdouts when the budget supports them; they remain the cleanest proof of causal lift.

The point is that person-level attribution and incrementality reinforce each other, and most DTC brands have never had access to the first one on CTV because they never owned the identity layer it depends on.

Phase 4

Optimize and compound your CTV result

A CTV program gets better the longer it runs, because each flight produces data that sharpens the next one. With person-level measurement in place, optimization stops being guesswork.

• Reallocate by audience, not just creative: when you can see which segments converted, you can shift budget toward the lookalike, intent, or retargeting pool that actually drove revenue.
• Manage frequency: cap exposures so you reach more households instead of fatiguing a few, and use suppression to keep recent converters out of prospecting audiences.
• Feed CTV into the full funnel: CTV creates awareness and intent; search and shopping capture it; retargeting closes it. Suppress and sequence across channels using the same identity layer so the channels stop competing for credit and start compounding.
• Refresh seeds continuously: as new customers convert, they flow back into the identity graph and improve the next lookalike. The program gets smarter on its own.