Most TV ad creative is a guess. Here’s how to change that.
-
Marketers predict the winner of their own creative tests about as often as a coin flip, and on TV that guesswork is expensive.
-
Strong creative earns attention, builds memory, stays distinctive, and pairs emotion with a reason to act.
-
A useful pretest measures attention, recall and brand linkage, emotional response, clarity, believability, and purchase intent.
-
The earlier you test, the cheaper it is to change, and AI and synthetic audiences now deliver a read from a script in minutes instead of weeks.
-
Pretesting predicts performance, and in-market testing proves it. The brands that do both compound their advantage.
Marketers tend to trust their gut about what creative works. The trouble is their gut is only right about half the time, no better than a coin flip.
Creative pretesting was supposed to close that gap. For decades it rarely did, because it was slow, costly, and easy to dismiss. That's all changed. AI and synthetic audiences can now predict how a TV ad will perform before production starts, in minutes instead of weeks.
We brought the topic to The Marketing Architects Podcast, with Angela Voss, Elena Jasper, Rob DeMars, and special guest Chief Creative Officer Steve Babcock weighing in on a question marketers have spent decades trying to answer: how do you know if an ad works before you've paid to find out?
Why did marketers distrust creative pretesting for so long?
Rob: I didn't believe in pretesting earlier in my career. Back then it meant bringing people into a room for a focus group, or later, running online surveys that still took weeks and a real budget. Either way, it was slow, and it didn't tell you anything useful. The concern was always the same. A consumer might say they like an ad, but that didn't mean they would actually buy the product.
So for years, we only tested in-market. That approach gave us something focus groups never could. It built us a real database of outcomes. Once we had enough of that data, we could build a pretesting methodology and check it against what actually happened. When the predictions started matching the results, that's when I was bought in.
Elena: The instinct to trust your gut is where the problem starts. Marketers predict winning ads only 52% of the time. That's barely better than a coin flip. Pretesting earns its place the moment it predicts the market better than you can.
What does the research say creative actually needs to do?
Elena: Stuart Mitchell's WARC analysis of more than 5,600 campaigns over a decade found that only 21% of creatively awarded ideas also won on effectiveness. The ones that crossed over had three things in common. They ran longer, used more channels, and balanced emotion with information. Winning on both creativity and effectiveness comes from sustained investment in the work and having the support in place to let it do its job.
Steve: Growth-driving creative is built to scale, to repeat, and to live in the real world for as long as possible. Emotion gets you the attention, and the rational message earns the sale. We need to stop treating them as opposites. They're two parts of the equation.
What should you actually measure when you pretest a TV ad?
Elena: A strong pretest doesn't return one score. It answers several different questions, and each maps to a different dimension of performance. Does the ad get noticed? Will people remember it, and will they credit your brand instead of a competitor? Does it make them feel something? Do they understand and believe the message? Does it move them toward buying? In practice,, that means tracking attention, recall and brand linkage, emotional response, clarity, believability, and purchase intent. Aided recall targets around 40% are a reasonable bar for breakthrough.
Rob: One metric people underweight is attention, specifically audio attention. TV is now an audio-first medium. People's eyes drift to a second screen, so if the work can't hold attention through sound, it loses the room. A pretest that ignores the audio strategy is measuring the wrong thing.
Steve: The other one is distinctiveness. Your characters, color, and sonic identity should look and sound like no one else in your category. There are studies linking distinctiveness directly to longevity and effectiveness. When we test, we want to know the ad is unmistakably ours, because that's what gets credited to the brand later.
When in the production process should you test?
Angela: Shift from thinking like a hunter to thinking like a farmer. Hunters chase immediate conversions. Farmers build something over time that pays off consistently.
Rob: Invest in creative that's memorable and emotionally engaging. And use your distinctive brand assets consistently. Your logo, colors, characters, and sonic identity should be repeated frequently.
The patience piece is real. It’s staring at the dirt waiting for things to grow. But the brands that build mental availability before the buying moment are the ones that win when it arrives.
How has AI changed creative pretesting?
Elena: A Brigham Young University study titled "Out of One, Many" showed that a language model could mimic human survey responses with surprising accuracy across political, demographic, and attitudinal questions. And a Qualtrics report found that 95% of senior marketing and insights leaders are already using or planning to use synthetic data within the next year, mostly for speed and depth of insight.
Rob: That's the foundation of ScriptSooth, our AI creative pretesting tool built on synthetic audiences. The test no longer has to be an animatic. It can be a Word document. We've stress-tested it against our own database of known in-market outcomes, and it's come back more accurate than the online survey panels we used to run in predicting performance. Speed went from weeks to minutes.
Elena: Generic LLMs are powerful, but ScriptSooth works because we had years of pretesting data plus a database of what actually drove results in market. That history is what makes it proprietary.
Can you actually trust a synthetic audience?
Angela: Skepticism here is healthy. The reasonable version asks whether a synthetic respondent gives you the same answer a real person would, then checks. You settle it by comparing synthetic results to human results, and to outcomes you already know happened. Do that across enough categories and the pattern holds. It's still human data. Just a lot of it, surfaced faster.
Rob: That's exactly how we built confidence in ScriptSooth. We didn't take it on faith. We checked it against outcomes we already knew, the same way Angela's describing, and it held up. Once the predictions started matching reality, the skepticism took care of itself.
Where does pretesting stop and in-market testing begin?
Elena: Pretesting predicts. In-market testing proves. You pretest to de-risk the work before you spend on production and media, and you confirm in market with incrementality and real performance once it's live.
Angela: The trap on the in-market side is overreacting to early noise. The first few days of data are mostly noise, not signal. Strong measurement is about time in market, not timing the market. Pretesting helps you commit with confidence so you're not pulling a good campaign before it's had a chance to work.
Advertising rewards conviction. The brands that trade the coin flip for evidence, before they spend and after they launch, are the ones that pull ahead and stay there.
About Marketing Architects
Marketing Architects is an agency helping brands drive growth through efficient, accountable advertising across linear and streaming TV.
The Marketing Architects Team
Curated by our leaders, creatives, analysts, designers, media buyers and more at Marketing Architects.