Creative Is Advertising’s Biggest Untested Variable

I’m writing this from Cannes, where, unsurprisingly, the industry is having a thousand conversations about creativity, storytelling, AI, and the future of advertising. The Croisette is packed, and everyone is fired up.

And yet, walking through those conversations, I keep coming back to one uncomfortable truth. For an industry obsessed with optimization, performance advertising still treats creative like a sacred black box.

We measure audiences, bids, and incrementality. We track CAC, ROAS, and LTV down to the decimal. But creative, arguably the single largest driver of whether a campaign actually performs, is still often governed solely by instinct and internal debate.

Teams often hear phrases like, “I like this cut better,” or “This pacing feels stronger.” Sometimes those instincts are right. Sometimes they are spectacularly wrong. Either way, billions of media dollars get committed before marketers have meaningful evidence about whether the creative itself is actually built to perform.

That frustration has followed me for over 20 years of building in performance advertising. Everyone in this industry acknowledges the problem, but almost nobody has solved it. Last week, we launched Creative Advisor, and sitting here at Cannes, I want to tell you what it does, why we built it, and what early results are already showing.

The last frontier of performance advertising

The measurement side of television advertising has improved dramatically. Deterministic attribution, outcome-based buying, pixel-level conversion tracking — these things work now. tvScientific by Pinterest has been at the forefront of making them standard, and across the industry, the back end of the performance equation has gotten genuinely rigorous. But the front end (i.e., the creative) has remained stubbornly, almost embarrassingly opaque.

Performance advertising has spent two decades building toward accountability. Search brought measurable intent, social brought scalable optimization, and Performance TV brought measurable outcomes to television itself. Every stage of the media stack has been put under the microscope, except creative. When a campaign underperforms, the first instinct is still to interrogate the audience targeting or the media strategy. Creative is where smart people shrug and say, "Well, it tested well internally." That is not good enough when every other dollar in your stack is accountable.

Creative performance is not actually a mystery. Its impact is driven by identifiable, measurable elements: how early your brand appears on screen, whether there's a voiceover or just music and visuals, how clearly your offer is communicated, how your end card is structured, whether your call to action lands. These are knowable variables. We just hadn't built the infrastructure to analyze them at scale, against real outcome data, before campaigns go live.

The industry developed a coping mechanism instead: taste masquerading as strategy. I want to be careful here, because I am not arguing for a world where AI generates interchangeable creative sludge and humans are optimized out of the process. Human creativity matters enormously. What I'm arguing is that creative deserves the same rigor we already apply to every other part of media. Creative judgment informed by evidence is a fundamentally stronger operating model than creative judgment alone.

Applying the scientific method to creative

One of the ideas I keep coming back to is how the scientific method transformed media buying. You start with a hypothesis, isolate variables, and test. From there you learn and iterate. That framework is overdue for creative.

Instead of asking, "Which ad does the room prefer?" (which is basically just voting), marketers should be asking sharper, testable questions. Does voiceover improve message comprehension and downstream action? Does a tagline with a clear value proposition outperform a URL-only approach? Does visual pacing help or hurt conversion behavior in a 30-second spot?

These are empirical questions, but the reason they haven't been answered systematically is that until now, the tools didn't exist to analyze creative at the element level, connect those elements to concrete performance outcomes, and deliver that analysis before a campaign launches.

That's exactly what Creative Advisor is built to address.

What Creative Advisor does

Creative Advisor uses tvScientific AI to evaluate your TV ad creative and predict how it is likely to perform before a single impression is bought. It assigns each ad a predictive Creative Strength score correlated with conversion performance, then surfaces the specific elements driving that score and generates actionable recommendations for how to improve it before any media dollars are spent.

The data foundation here is what actually makes this different from anything else in the market, and I want to be specific about it. Creative Advisor is powered by tvScientific by Pinterest's proprietary creative intelligence dataset, built from tens of thousands of CTV creatives, trained on millions of creative elements, and validated across hundreds of millions of household impressions.¹

In other words, Creative Advisor was trained on every creative asset ever run across tvScientific's platform history (more than two years of real-world CTV performance signals) to identify which specific attributes actually drive performance. That kind of systematic, outcome-linked creative analysis has never been done before at this scale in performance TV.

The tool evaluates creative across a wide range of signals, including:

• Messaging clarity
• Voiceover and audio effectiveness
• Logo placement and brand visibility
• Scene structure and visual composition
• Offer visibility and promotional clarity
• Direct response elements (QR codes and end cards)

It also generates specific, prioritized recommendations you can act on before the campaign runs.

What makes this different from a standalone diagnostic tool is how this fits into the larger performance flywheel we've been building. Creative Advisor lives inside the tvScientific by Pinterest platform, which means the recommendations feed directly into campaigns. You optimize before launch, run with stronger inputs, and then real performance data feeds back into the model over time. That is a compounding advantage and one that only exists because we built the measurement infrastructure first.

Without the outcome data, you can't train the model. Without the model, you can't predict pre-flight. It all connects.

What we're already seeing in the market

I don't believe in launching tools and declaring victory on theoretical benefits alone. We ran Creative Advisor with advertisers across categories before this launch, and what came back gave me genuine confidence that this works in practice.

The most consistent and repeatable finding across tests: branding that appears early and stays visible is one of the highest-impact creative levers available on CTV. That might sound like conventional brand marketing wisdom, but there's a meaningful difference between knowing something as a best practice and proving it with a controlled in-market experiment. We saw advertisers take Creative Advisor's branding recommendations, implement them (in some cases with relatively modest production changes), test the updated creative against their original, and come back with statistically significant lifts in site visits. They got clean results they could act on confidently.

These tests also surfaced something important about creative measurement methodology that I think deserves more attention. Site visits are a better KPI for creative learning experiments than purchases are, in most cases. At typical test budgets, purchase events are too sparse to reach statistical significance in a reasonable time window. Site visits give you a higher-volume, faster-feedback signal that lets you iterate quickly and make decisions with actual confidence before you scale your spend. It's a more honest way to run creative experiments.

What this means for the future of Performance TV

Here at Cannes, a lot of the AI conversation is framed as a tension: technology versus creativity, efficiency versus expression, data versus art. I find that framing genuinely tiresome because it misses the point. Creative Advisor is not a tool that generates your TV commercials. It's here to give creatives and the performance marketers who deploy their work a stronger foundation for decision-making before money hits the market.

Creative Advisor represents the completion of something we've been working toward for a long time: a genuinely closed-loop Performance TV system. You now have predictive intelligence before launch, outcome-based optimization during your campaign, and deterministic measurement after the fact. Every stage of the campaign lifecycle is accountable and improvable. That is a materially different operating model than where this industry was even three years ago.

And when you layer this on top of what we've built with tvScientific by Pinterest, combining Pinterest's early-intent signals with our AI-powered optimization engine, the effect compounds. You're reaching audiences who are already in a planning and discovery mindset, with creative that's been evaluated against real outcome data, running in an environment optimized toward measurable business results. That is the full-stack Performance TV capability that performance marketers have been asking for and haven’t fully had until now.

The creative problem has always been the last unsolved piece. Cannes celebrates what creativity can do at its peak, and that is great. But the best creative teams in the years ahead will make beautiful work that they can also understand, improve, and prove. Creative talent paired with stronger intelligence, faster learning loops, and measurable feedback is a different level of competitive advantage.

Creative Advisor is available now through tvScientific by Pinterest's creative and performance teams, with self-serve capabilities on the way. If you want to get into it, reach out.