AdTech Innovation: 2.3x ROAS in 2026

Everyone in adtech is chasing efficiency and numbers that prove the spend is worth it. A startup, MurphyCobb, just gave us a great example with a recent campaign that shows what real adtech innovation looks like for production, especially with dynamic creative. They totally automated their personalized ad delivery and the results were nuts, proving that good automation changes how you think about scaling campaigns and talking to individual customers. So, how exactly did they get that kind of jump in performance?

Deconstructing the “Hyper-Personalized Home Decor” Campaign

The “Hyper-Personalized Home Decor” campaign by MurphyCobb ran for 12 weeks between March and May 2026, on a total budget of $850,000. Their main goal was simple: sell more home decor online by showing people ads they’d actually want to look at. They knew that in a visual market like home goods, generic ads just don’t cut it. Their whole strategy was built on using some advanced startup tools for AI-driven dynamic creative optimization (DCO) to make every single ad impression personal, at scale.

Strategy: AI-Driven Dynamic Creative at Scale

The foundation of their plan was a custom DCO engine that was plugged directly into their customer data platform (CDP). This wasn’t just about swapping out product photos. The engine built entire ad layouts from scratch, backgrounds, where products were placed, color schemes, even the ad copy, all based on what it knew about a specific user. So if you were just browsing minimalist Scandinavian stuff on their site, you’d get an ad with a clean, modern living room. If someone else was into bohemian decor, they’d see a totally different ad with a lively, layered look. This was predictive creative, a big step up from basic retargeting.

They pushed the campaign out through a bunch of programmatic platforms, mainly Google Display & Video 360 and The Trade Desk, which gave them wide reach and tight control over their bids. The media buying team set it up so that impression-level data fed right back into the DCO engine. This created a closed loop that let the system learn and get better on its own. This thing required constant attention and algorithmic tweaks from the team. It wasn’t a “set it and forget it” job.

Creative Approach: Billions of Potential Variations

Getting the creative assets ready was a huge undertaking. MurphyCobb had to create a massive library of thousands of high-res product shots, lifestyle photos, and different background options. Their AI engine would then pull from this library, using a combination of set rules and machine learning to assemble the ads. The number of possible unique ads was theoretically in the billions. In practice, the system was smart enough to focus on making the right ad for the right user segment. The ad copy was built the same way, pulling from a bank of pre-approved phrases and adjusting the tone and call-to-action based on user intent signals. The “creative” was the system itself, a massive departure from an agency handing over a few final files.

Their “A/B/n” testing was a huge part of their success. Instead of just testing A vs B, their DCO platform was testing hundreds of different, dynamically created ad versions at the same time across all their segments. This let them figure out tiny things that made a big difference, like the best spot for a “Shop Now” button or the most effective lighting for a certain type of product, and they learned it way faster than you could with normal testing. With programmatic ad spend continuing to climb (as per a 2025 IAB report), being able to test at this level is becoming non-negotiable if you want to compete.

Targeting: Beyond Demographics

Sure, they started with basic demographics like age and location, but their targeting went so much deeper. MurphyCobb used psychographic data, what they could infer about interests, past purchases, and what users were doing on the site right now to build incredibly specific audience segments. They weren’t targeting “women 30-45 interested in home decor.” They were targeting people like “urban professionals, 30-38, recently viewed mid-century modern sofas, frequently interact with design blogs.” You have to get that specific for the DCO engine to have enough information to make a truly relevant ad. They also built lookalike audiences from their best customers to find new people who were likely to convert.

Pulling this off meant a serious investment in their data infrastructure and analytics team. Being able to take in, process, and act on that much user data in real time is a huge technical challenge. It really brings home a core point of modern adtech: managing and understanding your data is just as important as delivering the ad. If you have bad data, even the smartest DCO engine is just making pretty, random pictures.

Campaign Performance Metrics

The numbers from the “Hyper-Personalized Home Decor” campaign speak for themselves and really show what this kind of adtech can do.

Getting a 1.15% CTR on display ads is pretty wild, since that channel is known for low engagement. That strong CTR is what led to such a high number of conversions. With a Cost Per Conversion of $12.50, MurphyCobb spent its money very efficiently, especially given the average price of their stuff. But the killer metric here is the 3.8x ROAS. It’s proof that the personalized strategy drove real revenue growth that blew their old, less dynamic campaigns out of the water.

What Worked Well

• Real-time Creative Adaptation: The fact that the DCO engine could change ad creative based on live performance data was a huge win. If an ad element wasn’t working, the system would just stop using it and give more exposure to the combinations that were. That constant learning cycle means you don’t waste budget on bad creative.
• Hyper-Segmentation: Getting away from broad demographic buckets let them actually connect with people. The ads felt right because they were built for individual tastes, not for a giant, generic category of person.
• Smooth Platform Integration: The clean integration they built between their DCO platform, CDP, and programmatic buying tools was critical. It let data move instantly, which meant the system could make decisions and act on them fast. That kind of smooth operation is a signature of good adtech innovation.

What Didn’t Work as Expected

It wasn’t all perfect. In the beginning, the creative engine would sometimes spit out some really weird looking ads, clashing product styles, bad lighting. It took a period of intense human review and rule-writing to teach the AI what looked good. We also ran into an issue where some of our super-niche segments, while being highly relevant, just didn’t generate enough impressions to get statistically significant data. That meant the optimization cycle was a lot slower for those tiny groups. When you’re pushing personalization this far, that’s the trade-off: the smaller your target, the harder it is for the machine to learn quickly.

Another snag was the setup time. Building out that huge asset library and writing all the initial DCO rules took way more time than we planned. The long-term payoff was worth it, but it’s a huge barrier for smaller teams or companies that don’t have the technical people on hand. This kind of tech is a powerful tool that requires a big upfront investment of time, money, and expertise. It’s not a silver bullet.

Optimization Steps Taken

We made several key adjustments as the campaign was running:

1. Refined AI Rules: Early on, a team of our designers and data scientists worked together to tighten the AI’s creative rules. They put in more constraints to stop the weird aesthetic mistakes by manually reviewing thousands of generated ads and giving direct feedback to the machine learning model.
2. Segment Consolidation: For those tiny segments that weren’t getting enough traffic, we made a strategic decision to combine some of them into slightly larger, but still very relevant, groups. This gave us the right balance of personalization and actionable data, which let the DCO engine learn faster.
3. Budget Reallocation: We were constantly looking at ROAS in real-time and shifting budget to the best-performing platforms and segments. For example, after the first three weeks, we moved about 15% of the budget from some display networks over to social channels that were converting better for certain product lines. That kind of agile budget-shifting is only possible if you have great tracking.
4. Iterative Copy Testing: We were also constantly A/B/n testing copy, not just visuals. We tried out different calls-to-action and value propositions. We found that for users we’d identified as home renovators, “Transform Your Space” worked about 10% better than “Shop Latest Trends,” so the system automatically started using that phrase more for that segment.

The MurphyCobb campaign proves something we all know deep down about digital advertising: one-size-fits-all just doesn’t work anymore. Real production efficiency comes from building systems that can adapt, personalize, and learn at a massive scale. That kind of intelligent automation is the future of adtech and it’s changing how brands talk to their customers.