Urban Sprout’s 3.1 ROAS on CTV in 2026

Connected TV (CTV) advertising gets you into millions of living rooms, but actually measuring the return on that spend can feel like throwing money into a black box. The whole system is fragmented, there are no standard IDs to track people across devices, and that makes accurate attribution a serious headache. Brands are left wondering if their spend is working at all when the old measurement models completely break down.

I was running a CTV campaign for a direct-to-consumer (DTC) brand called Urban Sprout that sells sustainable home goods. The goal was simple: get more online sales by pushing traffic to their site and hit at least a 2.5x return on ad spend (ROAS). We were going after direct conversions, a hard target for CTV which is usually seen as an upper-funnel play. We put $250,000 into CTV over an eight-week period.

Initial Strategy: Targeting and Creative

Our game plan started with a mix of contextual and audience targeting. On the contextual side, we ran our ads during eco-themed shows and lifestyle content on big platforms like Hulu, Peacock, and Tubi. For audiences, we used third-party data to find households into sustainable living and with higher incomes (HHI). We zeroed in on specific neighborhoods in Atlanta, Georgia, places like Virginia-Highland, Decatur, and Buckhead, that matched our target customer profile.

We ran two different 30-second video ads. Ad A was all about the beautiful design and quality of Urban Sprout’s products, while Ad B focused on the sustainability story and ethical sourcing. Both had a clear call-to-action (CTA) to visit urban-sprout.com. Wherever we could, we used interactive formats that let people scan a QR code or click to the site with their remote, which was our attempt to turn a passive TV-watching moment into a shopping trip.

The Measurement Dilemma: What Worked, What Didn’t (Initially)

At first, we measured everything with a last-touch attribution model, which just gives all the credit to the very last thing a customer did before buying. A lot of people do this, but it’s a terrible way to see the full picture in a campaign with many channels. For our CTV ads, it meant we only counted conversions from QR code scans, direct clicks from the remote, or post-view conversions (PVRs) that happened within 24 hours. The initial numbers looked awful:

A 0.8x ROAS was a disaster. At a $125 cost per conversion, we were way off target. The data screamed that CTV was a waste of money for direct response, but I didn’t buy it. Our targeting was too precise and the creative was solid. The problem wasn’t the channel. It was how we were measuring it. This is the classic trap of CTV advertising measurement. The cookie-based tracking we’re all used to on desktop and mobile just doesn’t exist in the cookieless world of TV sets.

The Attribution Overhaul: Embracing Multi-Touch Models

Since we knew last-touch was giving us bad data, we switched to a multi-touch attribution (MTA) model right away. We chose a time-decay attribution model, which gives more weight to the touchpoints that happen closer to the sale but still gives some credit to earlier ads like our CTV spots. Getting this to work meant pulling data from our CTV ad server, our CRM, and our Google Analytics 4 (GA4) account and stitching it all together. We also had to use a device graph to connect what a household saw on their TV to what someone in that house later did on their phone or laptop. It’s not a simple process (it often requires expensive platforms and data people), and it’s where partners like Nielsen and Innovid make their money.

Here’s what we had to do:

1. Unify the Data: We pulled all the impression data from our CTV buys and matched it up against our website’s conversion data.
2. Match Devices: We used anonymized IP addresses and other household IDs to connect a CTV ad view to a later website visit or purchase from a different device, like a phone.
3. Model the Attribution: We ran our time-decay model to spread the conversion credit across every touchpoint we could see. So if someone saw a CTV ad, then clicked a social ad, and then bought something, the CTV ad would get a piece of the credit instead of zero.

Revealing the True Impact: Optimized Metrics

Once we ran the numbers with the time-decay MTA model, the picture changed completely. It turned out CTV was contributing way more to sales than we thought, proving our theory that it was teeing up purchases that happened later. That massive 12,500,000 impression number wasn’t just noise. It was influencing behavior, even if it didn’t lead to an immediate click.

Suddenly, CTV was responsible for 5,200 conversions. That dropped our attributed cost per conversion to a very healthy $48.08. The ROAS shot up to 3.1x, blowing past our original goal and proving the spend was worth it. This is exactly why you can’t just look at last-click. If you ignore upper-funnel channels like CTV, you’ll end up cutting budgets that are actually working. And that CTR of 1.2% on our interactive ads? For CTV, that’s a great signal that a portion of the audience is ready to act right then and there.

Optimization Steps and Learnings

With real data in hand, we could finally start optimizing properly:

• A/B Testing Creative: We kept testing Ad A vs. Ad B. The ad about sustainability (Ad B) beat the aesthetics-focused one (Ad A) by 15% in driving attributed conversions. The message mattered.
• Shifting Budget: We moved 15% of our CTV spend away from platforms with weak engagement and high costs, and put it into the ones that were performing. For example, one AVOD service was a clear winner on cost per conversion, so we doubled down there.
• Refining Audiences: We took everyone who scanned a QR code or clicked through and put them into a new retargeting segment for other digital ads, which helped lower our acquisition costs down the line.
• Capping Frequency: We saw that showing an ad more than twice a day to the same person didn’t help, so we capped frequency at 2/day instead of 3. This simple tweak stopped wasted impressions and is a key lever to pull in CTV to avoid annoying people.

A quick thought on this: I get why marketers stick with last-click. It’s easy. But easy isn’t accurate, especially with something this complicated. If you’re spending money on CTV and not using some form of MTA to measure it, you are almost certainly making bad decisions, either by killing a channel that’s working or giving credit to one that isn’t. Yes, it’s hard to set up, but the payoff in knowing what’s actually going on is huge.

Conclusion

The Urban Sprout campaign is a perfect example of why you can’t measure CTV advertising with old-school, single-touch models. As soon as we switched to a time-decay MTA model, we saw the real story: a channel that looked like a loser was actually a winner, delivering a 3.1x ROAS. To get CTV right, marketers have to get serious about their measurement tools and stop guessing where their media spend is working.