THE UNDERGROUND AWARDS SHOW: WHAT IF WE JUDGED ADS BY ROI INSTEAD OF JURIES?

Every June, advertising's elite gather in the south of France to hand each other trophies. The judges deliberate behind closed doors, guided by craft, taste, and whatever unspoken politics animate a room full of peers reviewing each other's work. It's a system built on consensus among a small, self-selected group — and it has governed the industry's sense of creative excellence for decades. But while those juries convene once a year, another awards show has been running nonstop, every hour of every day, with no intermission and no shortlist. Its judges are consumers. Its ballots are clicks, conversions, and cost-per-thousand impressions. And its verdicts are ruthlessly, uncomfortably objective.

Consider what's happening right now inside social ad auctions. When AdExchanger analyzed Meta advertising activity across six of the largest U.S. insurance advertisers, the findings didn't reveal who had the cleverest tagline or the most cinematic spot. They revealed something far more consequential: Progressive wasn't simply outspending competitors — it was acquiring attention at a dramatically lower CPM despite being the category's largest spender. That combination, scale plus efficiency, is the advertising equivalent of winning Best Picture while also topping the box office. But no jury awarded it. The auction did. The significance, as the analysis made clear, isn't the CPM itself; it's what the CPM implies — that somewhere inside Progressive's system, there are structural advantages in audience precision and placement diversification that competitors may not even recognize, let alone possess. This is a competitive signal that would never surface in a creative reel or a case study video, yet it represents the kind of real-world dominance that traditional awards purport to celebrate.

Now multiply that logic across the entire ecosystem. When Unilever announced its pivot to a 300,000-creator network — with 71% of those creators using AI tools to produce content at scale — the old evaluation infrastructure immediately buckled. Human panels are too slow. A/B testing individual assets across a network that vast is logistically impossible. Traditional brand-tracking surveys capture what happened last quarter, not what's working right now. As Search Engine Journal reported, this is precisely the gap that platforms like DAIVID and ADIN.AI are racing to fill, building what amounts to a live loop between creative intelligence and media execution. Before a campaign launches, their system identifies which creative is most likely to succeed. While campaigns run, it scales winners and pauses losers in real time. After campaigns end, the performance data becomes the benchmark for what comes next. DAIVID CEO Ian Forrester identified the core dysfunction their partnership addresses: creative has been "measured in isolation, disconnected from media results" for too long.

What these systems collectively represent is the infrastructure of a parallel, meritocratic awards show that never stops judging. The "winners" aren't chosen by a panel in a screening room squinting at production values. They're surfaced by millions of micro-transactions — every bid, every impression, every scroll-past or click-through functioning as a vote cast in real time by the only jury that actually matters: the audience being asked to pay attention. This system is more democratic, more honest, and more useful than anything that happens on a stage in Cannes. It just doesn't have a red carpet. And perhaps that's exactly why the industry has been so slow to treat it as the award it already is.

Walk into any awards show judging room and you'll find a specific aesthetic orthodoxy at work. The winning ads tend to share a common DNA: arresting visuals, conceptual novelty, emotional narratives that build slowly toward catharsis, and — increasingly — a posture of social commentary that signals the brand's moral seriousness. These are legitimate artistic virtues. Nobody should pretend otherwise. A beautifully crafted brand film can shift perception, build cultural relevance, and generate the kind of earned media that no performance campaign ever could. The problem isn't that awards shows celebrate great craft. The problem is what they ignore.

In performance markets, the winning creative looks nothing like a Cannes reel. It's blunt. It's fast. It's often ugly. A direct-response ad that converts at twice the rate of its polished competitor doesn't care about typographic elegance or a three-act story structure. It cares about clarity — does the viewer understand the offer in the first two seconds? — and relevance — does the targeting match the intent? The ads that dominate performance dashboards are optimized for entirely different selection pressures than the ads that dominate festival shortlists, and the gap between these two worlds has become a chasm that the industry politely refuses to discuss.

This disconnect would matter less if the awards ecosystem didn't exert such gravitational force on how agencies allocate their best creative talent. But it does. When the path to a promotion, a raise, or a new job runs through a trophy case, creatives rationally optimize for jury taste. They build "awards entries" — work designed to win applause in Cannes, not results in market. The incentive structure is clear, and the output follows the incentive.

Meanwhile, the most sophisticated advertisers have moved to an entirely different definition of what "great advertising" even means. As Avinash Kaushik frames it, the old game was whether the human could out-manage the ad account; the new game is whether the company can feed the machine better truth, better creative assets, and better customer value signals. In this model, advertising excellence isn't about a single hero spot — it's about the quality of the inputs you provide to an optimization engine that runs millions of micro-experiments a day. The creative brief becomes a data architecture problem. The "best" ad is whatever variant the system discovers at scale, not whatever a creative director fell in love with at 2 a.m.

But before the performance zealots take a victory lap, it's worth staring hard at what happens when the machine runs without guardrails. As AdExchanger documented, Skechers has been running AI-generated out-of-home ads featuring hypersexualized images of young girls — creative so flagrantly inappropriate that no human art director could have submitted it without being fired on the spot. The same piece catalogued brands producing fake CGI product review ads, a practice that would have been genuinely scandalous just a few years ago but now barely raises an eyebrow inside the cult of performance. When platforms like Performance Max and Advantage+ Shopping are left to their own devices, they'll dredge the worst images from a product feed or generate surreal creative on the fly — whatever stops thumbs, regardless of brand safety or basic human decency.

This is the Cannes paradox in its starkest form. The awards ecosystem pretends that effectiveness is someone else's department. The performance ecosystem pretends that brand equity is a luxury it can't afford. Neither extreme — jury worship nor unchecked algorithmic optimization — produces advertising that is both effective and worthy of admiration. One side builds beautiful work that nobody outside the industry ever sees. The other builds work that millions see but nobody should be proud of. And the conversation between these two camps? It doesn't exist, because the awards show has no category for it, and the market doesn't care.

While the previous section examined the gap between what wins awards and what drives results, there's a more radical development unfolding beneath that tension: the question of who evaluates creative work is rapidly becoming moot, because machines are already doing it — faster, cheaper, and at a scale no human jury could replicate.

Consider the architecture of what MarTech calls "agentic AI" — systems that don't merely execute instructions but make decisions autonomously. These self-optimizing agents experiment continuously, reallocating budget, adjusting targeting, and refining creative without human intervention. Read that description carefully and you'll notice it contains every function of an awards jury — evaluation, deliberation, verdict — compressed into milliseconds and running around the clock. The difference is that these systems don't deliberate about whether an ad is "brave" or "culturally resonant." They deliberate about whether it works. And they do so not once a year in a ballroom on the French Riviera, but thousands of times per hour across live campaigns.

The velocity matters because the volume of modern advertising has outstripped any human evaluation mechanism. When Unilever announced its pivot to a network of 300,000 creators — 71% of whom are using AI tools to produce content at speed — the traditional infrastructure for separating good creative from bad simply collapsed under the weight. Human panels are too slow. A/B testing individual pieces across a creator network of that size is logistically impossible. Quarterly brand-tracking surveys tell you what happened last quarter, not what's working right now. The old evaluation model was designed for a world where a brand ran a handful of campaigns per year, each one painstakingly crafted and carefully assessed. That world is gone.

What's replacing it looks something like what DAIVID and ADIN.AI have built: a live loop between creative intelligence and media execution that scores creative effectiveness at scale, links those scores to real-time media performance, and surfaces signal from noise before budget gets allocated to the wrong places. Before a campaign launches, the system identifies which creative is most likely to succeed. While campaigns run, it scales winners and pauses underperformers automatically. After campaigns end, the performance data becomes benchmarks for future work. This isn't optimization in the old sense — a media planner tweaking bids on a Tuesday afternoon. This is a closed-loop judgment engine that decides which ads deserve investment and which deserve oblivion, operating at a cadence no awards jury could match.

As Avinash Kaushik frames it on his blog, the old game was whether a human could out-manage the account. The new game is whether a company can feed the machine better truth — better creative assets, better customer value signals, better data — and give it enough room to learn. The cultural implication is profound: taste, that most prized currency of the creative director, is being subordinated to signal. The new creative director doesn't have a reel or a shelf of trophies. It has a feedback loop. And it never stops judging.

This doesn't mean human creativity is irrelevant — far from it. The strategic inputs still matter enormously: brand narrative, messaging architecture, audience understanding. But the act of evaluation, the moment when someone or something decides this ad is good and that one isn't, is migrating from subjective human consensus to automated, data-driven arbitration. The jury hasn't been replaced. It's been made perpetual, algorithmic, and indifferent to reputation.

But before we get too comfortable with this thought experiment, we need to turn the skepticism inward. The premise of an "underground awards show" — one that replaces subjective juries with hard performance data — sounds appealingly rigorous. It flatters our desire for objectivity. But objectivity requires more than dashboards, and the metrics we worship may be just as fallible as the judges we're trying to replace.

The core problem is attribution itself. The systems that most marketers rely on to determine which ad "worked" do not actually prove causation. They describe sequences — a user saw an ad, then bought a product — and assign credit based on proximity or platform logic. But observation is not explanation. A last-click model rewards the ad that happened to appear closest to a purchase that may have been inevitable anyway. A view-through window credits impressions that a user may never have consciously registered. Even sophisticated multi-touch models are, at bottom, allocation heuristics dressed up as science. If the underground awards show is handing out trophies based on ROAS dashboards or platform-reported conversions, its verdicts may carry a veneer of empirical authority while being no more reliable than the gut instinct of a jury in Cannes.

The platforms themselves have every incentive to make their attribution look favorable. As AdExchanger has documented, Meta quietly introduced "engage-through attribution" in March — a new category designed to credit social ads more generously than traditional click-through models would. Around the same time, Meta also altered what it calls "safe zones" in ads, making it easier for users to accidentally tap on creative and trigger an attributed click. Meanwhile, Google's AI ad products, left unchecked, will aggressively bid on a brand's own name and related terms — essentially paying to cannibalize organic traffic, then reporting the resulting conversions as performance wins. These are not neutral measurement systems. They are revenue models with a measurement interface.

And the downstream consequences of treating these signals as gospel are not merely analytical — they are creative and ethical. When brands surrender to what AdExchanger calls the cult of performance, the machine optimizes for whatever stops thumbs, regardless of brand coherence or basic decency. Skechers ran AI-generated out-of-home ads featuring hypersexualized imagery of young girls that no human creative director would have approved. Major brands now routinely produce fake CGI product review content that, just a few years ago, any CMO would have considered scandalous. The performance metrics validated all of it. The dashboards showed engagement. The attribution models assigned credit. And the guardrails collapsed — not because anyone decided they should, but because nobody was watching anything except the numbers.

This is the trap the underground awards show must reckon with: optimizing for the measurable risks abandoning the meaningful. Brand equity, cultural resonance, long-term trust — these are real forms of value that resist neat quantification. A campaign that builds deep affinity over years will almost always lose in a quarterly ROAS comparison to a campaign running retargeting ads against people already holding their credit cards. As MarTech has noted, the brands that will ultimately succeed are those that establish governance for autonomous systems, setting boundaries that balance performance optimization against brand equity rather than letting the algorithm chase conversions into oblivion.

So the underground awards may be more honest than Cannes — but they are not honest enough. Swapping one flawed authority for another is not reform. It is redecoration.

So let's actually build this thing — at least on paper. If we're serious about judging advertising by what it achieves rather than how it makes a jury feel, we can't just swap trophies for dashboards. A credible, performance-based awards system would need an architecture as rigorous as the scientific method it claims to champion. And the moment you start sketching that architecture, you realize why nobody has successfully built one yet.

The first and most critical requirement is incrementality. Not conversion tracking, not ROAS as reported by the platform that sold you the media, but genuine counterfactual analysis: what would have happened if this campaign had never run? That means randomized geo experiments, holdout markets, synthetic control groups — the kind of measurement infrastructure that most brands don't have and most agencies have no incentive to build. Without it, every entry in our hypothetical awards show would be submitting the advertising equivalent of a student grading their own exam. The same fundamental problem applies to the broader digital ecosystem: as AdExchanger has argued, even the W3C's privacy-focused measurement frameworks risk making a critical mistake by prioritizing last-touch attribution models that confuse correlation with causation. An awards system built on those same flawed signals wouldn't be more honest than Cannes — it would just be differently dishonest.

The second requirement is temporal range. Short-term conversion metrics reward the bottom of the funnel and punish brand-building work that compounds over quarters and years. A legitimate ROI awards framework would need to evaluate campaigns across multiple time horizons — 30-day direct response, 6-month consideration shifts, 18-month brand equity effects — and weight them according to category norms. A luxury brand and a DTC mattress company cannot be evaluated on the same timeline, and any system that tries will simply crown whoever had the shortest purchase cycle.

Third, competitive context matters enormously. A 20% lift in conversions means something very different in a category with two competitors than in one with two hundred. The competitive signals embedded in platforms like social ad auctions — bid density, impression share, cost-per-outcome volatility — could theoretically provide that context, but harvesting and normalizing that data across entrants would be a logistical nightmare requiring platform cooperation that doesn't currently exist.

Fourth — and this is the requirement that would make most performance marketers uncomfortable — you'd need creative governance criteria. Pure optimization without guardrails produces dark patterns, misleading urgency tactics, and brand-corrosive clickbait that converts beautifully in the short term and destroys equity over time. As Avinash Kaushik frames it in his Google Ads maturity model, the new competitive game isn't about out-managing individual auctions but about feeding machines "better truth, better creative assets, better customer value signals." An awards system that ignored the quality of those inputs — that celebrated a campaign for its conversion rate while ignoring whether it lied, manipulated, or strip-mined brand trust to get there — would be rewarding the advertising equivalent of strip mining: impressive yields, uninhabitable land.

Finally, any credible ROI awards system would need independent verification. Not self-reported case studies, not platform-provided analytics, but third-party audited results with disclosed methodologies and confidence intervals. The moment you require that level of transparency, you eliminate roughly 95% of potential entrants — which is either the system's fatal flaw or its entire point, depending on how honest you want the industry to be.

The uncomfortable conclusion: a truly rigorous performance-based awards show is theoretically possible and practically almost impossible. Not because the measurement science doesn't exist, but because the institutional willingness to submit to it doesn't.