When Common Sense Fails: Why Evidence-Based Frameworks Beat Intuition in the Age of AI

In 2007, Microsoft launched a $300 million advertising campaign for Windows Vista featuring Jerry Seinfeld and Bill Gates. The ads were quirky, unexpected, and starred two of the most recognisable figures in their respective domains. They generated buzz. They were memorable. And they were … completely ineffective at selling Vista. Microsoft pulled them within weeks.

That same year, Apple's "Get a Mac" campaign, featuring a nerdy "PC" character and a cool "Mac" guy in a series of simple comparison ads, was dominating market share growth despite costing a fraction of Microsoft's media budget.

What made the difference? Common sense would tell you both campaigns should have worked. Both featured recognisable figures. Both were witty and unexpected. Both had substantial media budgets. Yet one is now studied as a cautionary tale, while the other is taught as one of the most effective tech campaigns in advertising history.

The difference wasn't star power, humour, or budget. It was whether the campaigns aligned with evidence-based predictors of what makes marketing stick in memory and drive behaviour – or not.

The Digital Age Paradox

Fast forward to 2025. New York City's mayoral race pitted the 34-year-old democratic socialist Assemblyman Zohran Mamdani against former Governor Andrew Cuomo, an experienced politician with name recognition, major endorsements, unlimited access to political consultants, and a campaign budget that dwarfed his opponent's. Cuomo's campaign deployed the latest AI tools to generate content at unprecedented scale across every platform.

Common sense said Cuomo should win. Experience matters in politics. Name recognition matters. Resources matter. Cuomo had all of these advantages, plus cutting-edge technology.

Mamdani won decisively, capturing 50.4% of the vote and becoming NYC's first Muslim mayor and youngest in over a century.

Mamdani's approach was entirely different. The October 2024 launch video showed Mamdani walking through the doors of a New York City bodega or convenience store, ordering a typical chopped cheese sandwich, and immediately distancing himself from the cartoonish corruption and donor-centric politics of Andrew Cuomo and the other mayoral candidate, Eric Adams. In one minute and forty seconds, Mamdani defined the contours of working-class struggle in New York City, aligned himself with it, and introduced a set of policies designed to alleviate ordinary people's material burdens.

After Mamdani's primary victory, when Andrew Cuomo was asked what he did wrong during his primary run, he responded, "I did not do enough on social media … I think the assemblyman did a better job on TikTok".

But the difference wasn't just "doing more on TikTok." It was understanding the evidence-based principles of what makes messages stick and what makes them drive action – principles that Cuomo's well-funded, AI-powered campaign completely missed despite having access to unlimited consultants and technology.

The Trap of Marketing Intuition

If you're a marketing director today, you face a version of this challenge at unprecedented scale.

AI tools like ChatGPT, Claude, and Midjourney can generate months' worth of content - social posts, ad copy, video scripts, banner designs - in hours. Your team can produce more creative assets this quarter than your entire department generated in all of last year.

But there’s a paradox. While creating content has never been easier, ensuring it works has never been harder. Why? Because when everyone has access to the same AI tools, the differentiator isn't production capacity; it's strategic judgment about what to produce in the first place. And this is precisely where intuition fails us.

Consider three email subject lines for a digital marketing executive education programme:

• Option A: "New Executive Programme: Digital Marketing Strategy at LSE"
• Option B: "Join LSE's Most In-Demand Course: Digital Marketing Strategy"
• Option C: "The London School of Economics is launching a course in digital marketing. Be part of a new class of marketers."

Which would you choose? More importantly, which would your target audience remember 24 hours later, let alone act upon?

Common sense might lead you toward Option A or B. They're clear, professional, descriptive, and emphasise either newness or social proof - everything your instincts tell you a good subject line should be. Option C feels wordier and less punchy. Surely brevity wins in our attention-deficit world?

Why Intuition Misleads Us

The problem with marketing intuition is that it's built on assumptions that feel self-evident:

• Be clear and comprehensive. If people understand what you're offering, they'll engage.
• Make it punchy and brief. In our attention-deficit world, brevity wins.
• Lead with credentials. The LSE brand should do the heavy lifting.
• Emphasise scarcity or social proof. "Most in-demand" creates urgency.

These aren't unreasonable assumptions. They're what any competent marketer might conclude from experience. They're what feels professionally safe. But "reasonable" and "effective" aren't the same thing.

Research into what actually makes marketing messages stick, across thousands of campaigns, products, and contexts, reveals that the reliable predictors of success are quite different from what intuition suggests.

What the Evidence Actually Shows

Let us reveal two of the key evidence-based predictors that separate campaigns like Apple's "Get a Mac" and Mamdani's bodega video from their competitors. (There are three additional factors and together they create a comprehensive framework for evaluating marketing effectiveness. We will explore all of these in depth in our forthcoming LSE programme, Digital Marketing Strategy: AI, Analytics and Innovation.)

 

Predictor #1: CONCRETENESS

Messages that create specific, tangible mental images dramatically outperform abstract messages, even when the abstract version conveys the same information.

Consider our email subject lines again through this lens:

• Option A ("New Executive Programme: Digital Marketing Strategy at LSE") is abstract. "Programme," "strategy"—these are conceptual terms that don't create visual images.
• Option B ("Join LSE's Most In-Demand Course") is also abstract. What does "most in-demand" look like? What mental image does it create?
• Option C ("The London School of Economics is launching a course in digital marketing. Be part of a new class of marketers") is more concrete. "London School of Economics" evokes a specific place. "Launching" suggests action. "Class of marketers" creates an image of a group—you can visualise joining a cohort.

Now let's apply this to our campaign examples:

Apple's "Get a Mac" campaign was maximally concrete. Abstract technical differences ("compatibility with enterprise software" vs. "superior user experience") became concrete human characters you could visualise. PC wearing a stiff business suit: you can see the rigidity and corporate constraint. Mac in casual clothes: you can see the relaxed, creative approach. When PC got a virus or crashed, you saw a human character physically sick or falling down. It’s a concrete visual representation of abstract technical failures.

Mamdani's bodega video was maximally concrete. Walking through bodega doors, ordering a chopped cheese—these create specific mental images. When he talked about "fast and free buses, universal childcare, a rent freeze, and city-owned grocery stores," each policy created a tangible visualisation. You could see what he was proposing, not just understand it conceptually.

Predictor #2: UNEXPECTEDNESS

Messages that violate our expectations in surprising but meaningful ways are significantly more likely to be remembered and acted upon than messages that simply convey information clearly.

This is because unexpected elements force our brains to hit pause, pay attention, and figure out what's going on, creating the cognitive engagement that transfers information into memory.

But here's the crucial distinction: unexpectedness must be strategic, not random.

Apple's "Get a Mac" was unexpected in a strategically meaningful way. Personifying operating systems as human characters wasn't just novel, it created a schema violation that forced engagement. You expected tech ads to show products or features. Instead, you got a conversation between two people. This unexpectedness served the strategic goal of making technical differences emotionally resonant and memorable.

Mamdani's entire campaign violated political messaging conventions. One digital strategist noted that Mamdani's social presence "dismantles the ivory tower" that so many politicians keep themselves in. "He is so smiley, he's so giggly. He's always hugging people," Jain said. "He's just running a grassroots and community-driven campaign, and I think his body language embodies that. Like, I've never seen Cuomo hug anyone in my entire life".

Political convention says: be polished, measured, emphasise experience, moderate your positions. Mamdani violated all of these expectations: explicitly socialist positions, casual bodega settings, personal warmth on camera. Each violation was strategically aligned with his message about being different from establishment politics.

Our subject line example demonstrates this principle at smaller scale. Options A and B follow the template you expect from course announcements. Option C violates that template slightly - it reads more like a news announcement than promotional copy - creating a brief moment of cognitive engagement.

Why These Two Factors Matter More Now

In the age before AI, creating content required time and resources. The constraint was production capacity, so clever shortcuts and conventional wisdom were valuable for efficiency.

In the age of AI, production capacity is unlimited. When you prompt ChatGPT to "write 10 subject lines for an executive education programme," you'll get grammatically perfect, professionally worded options that look exactly like subject lines. The AI defaults to what's typical - which means neither concrete nor unexpected. 

The same applies to every AI marketing tool. They're trained on patterns - that is, what content looks like - and not on principles of what makes content work.

The Other Three Factors

Concreteness and unexpectedness are two of five key evidence-based predictors of what makes messages stick and drive action. The other three factors interact with these two in complex ways, and together they create a comprehensive framework for evaluating marketing effectiveness - whether you're generating content via AI, commissioning it from agencies, or creating it internally.

We deliberately haven't revealed all five factors here because understanding how they work together, how to apply them across different contexts (B2C vs. B2B, digital vs. traditional, product vs. service), and how to use them to evaluate AI-generated content requires the depth of exploration we provide in our upcoming programme.

What we can tell you is this: these aren't theoretical constructs. They're backed by decades of research across psychology, neuroscience, and marketing science. They predict campaign success across categories, cultures, and time periods. And they're precisely what separates "Get a Mac" from competitors with larger budgets, and Mamdani's campaign from Cuomo's well-resourced, high-tech challenge.

The New Imperative: Strategic Judgment Over Production Capacity

In the age of AI, the organisations that win won't be those that produce the most; they'll be those that produce content strategically aligned with what actually drives attention, memory, and action.

This requires a fundamental shift in how you think about your role as a marketing leader.

You're no longer primarily managing production workflows. Today, you're managing judgment: decisions about what to produce, how to evaluate what AI generates, and when to override your intuitions in favour of evidence.

Consider a practical scenario. Your team uses AI to generate 50 variations of a social media campaign for a product launch. Common sense evaluation criteria might be:

• Is it clear what we're selling?
• Does it sound professional and on-brand?
• Is it brief and punchy?
• Does it include our logo and tagline prominently?

Evidence-based evaluation asks different questions:

• Does it create concrete mental images or traffic in abstractions?
• Does it violate category expectations in meaningful ways or follow predictable patterns?
• [Three additional questions based on the other factors we'll explore in the programme]

Most AI-generated content will score poorly on evidence-based criteria until you've learned to prompt for these specific elements and, more importantly, learned to recognise when they're present or absent.

What This Means for Your Organisation

If your competitors have the same AI tools you do, your competitive advantage won't come from better software. It will come from better frameworks for deciding what to create and how to evaluate it.

After Mamdani's victory, digital strategists warned that "the upcoming midterm cycle is going to be very funny. Lots of politicians trying to do stuff like [Mamdani's videos], not realising why it isn't hitting the same way". The same will be true in commercial marketing. Organisations will try to replicate successful campaigns without understanding the underlying principles that made them work.

This is why merely training your team on "how to use ChatGPT for marketing" isn't sufficient. They need to understand:

• What makes content stick in memory and drive action (not what makes it feel professional or on-brand)
• How to evaluate AI outputs against evidence-based criteria (not just brand guidelines and intuition)
• When to trust research over instinct (and when human judgment still matters)
• How to prompt AI to generate strategically valuable content, not just grammatically correct content

These aren't technical skills. They're strategic capabilities that sit at the intersection of marketing science, consumer psychology, and AI tool literacy.

An Invitation

The digital marketing landscape has fundamentally changed. AI hasn't just given us better tools; it's changed the nature of marketing advantage itself. The organisations that will thrive aren't those with the most advanced AI subscriptions. They're those whose leaders understand what actually makes marketing work, and can deploy AI in service of those principles rather than letting AI's defaults drive their strategy.

In Digital Marketing Strategy: AI, Analytics and Innovation at LSE this June, we explore exactly this challenge. You'll learn the complete five-factor evidence-based framework that predicts campaign success across contexts. You'll gain hands-on experience with cutting-edge AI tools, from custom GPTs to digital twins to predictive analytics platforms.

And crucially, you'll learn to evaluate AI outputs critically, distinguishing content that feels right from content that performs right. You'll learn when common sense misleads and when evidence provides better guidance. You'll learn to see the difference between campaigns that succeed and those that fail, before spending resources to find out which is which.

This isn't a programme about AI tools. It's a master class in strategic judgment in an age when everyone has access to the same tools - in building competitive advantage that your rivals can't easily replicate because it's based on deeper understanding, not just better software.

We'd be delighted to have you join us on campus in June.

 

So, Which Subject Line Actually Works?

Remember our three email subject line options from earlier?

• Option A: "New Executive Programme: Digital Marketing Strategy at LSE"
• Option B: "Join LSE's Most In-Demand Course: Digital Marketing Strategy"
• Option C: "The London School of Economics is launching a course in digital marketing. Be part of a new class of marketers."

In testing across similar professional audiences, Option C generates significantly higher open rates than Options A or B.

Why? It's more concrete (creates a clearer mental image of a specific institution, an action - launching - and a group you'd be joining) and slightly more unexpected (reads like a news announcement rather than promotional copy, violating your schema for course marketing emails).

Options A and B traffic in abstractions ("programme," "strategy," "most in-demand") that don't create vivid mental pictures. They also follow the exact template you expect from course announcements, so your brain processes them on autopilot.

This feels counterintuitive to most marketers. Option C is wordier. It doesn't lead with LSE in the most prominent way. It doesn't emphasise scarcity or social proof. By conventional marketing wisdom, it should underperform.

That's the challenge of modern marketing: The evidence often contradicts what your experience tells you should work. And in an age when AI can generate infinite content based on conventional patterns, knowing when to violate those patterns - and having frameworks to guide those violations strategically rather than randomly - becomes your competitive advantage.