The marketing and advertising industry is completing a decades-long shift from description to prediction
Earlier this month, WIRED reporter Reece Rogers did something any Californian can do: he asked McDonald’s for a copy of the personal data the company had collected on him through its loyalty program. What he received back was a 515-page dossier — a document so extensive that commentators noted it rivals what the FBI once compiled on John Lennon.
The file contained exactly what you’d expect from a loyalty program: order history, favorite menu items, most-visited locations. But it also contained something far more interesting — and far more instructive for anyone who works in marketing. McDonald’s had algorithmically modeled Rogers’ future: his projected visit frequency, his anticipated spend, his most probable next order, and an “attrition score” of zero percent.
In the company’s mathematical estimation, Reece Rogers will never stop eating at McDonald’s.
The reflexive reaction to this story has been privacy alarm, and it should be.
We live in an age of mass (mostly automated) surveillance, where companies like Palantir and Flock monitor every thing we do in the digital, and physical world. The size and scale of your digital footprint is more substantial than you realize, and somewhere, sitting on a server in some obscure part of the US (or abroad) there is a sizable aggregation of data about who you are, where you live, which websites you visit, what apps you use, how you use those apps, what media you consume, what products you buy, and your complete financial (and potentially medical) history.
If that bothers you – it should.
I’m positive Amazon is no different in their data collection practices.
I can only imagine what Meta knows about humanity.
Google probably knows more about me than I do.
But for marketers, the McDonald’s dossier on Reece Rogers is something else entirely: a rare, unfiltered look at how a $26 billion enterprise actually operationalizes customer data. And what it reveals should reframe how every brand — enterprise or otherwise — thinks about buyer personas.
The dossier is a persona. It’s just a persona built about one person.
Strip away the surveillance framing for a moment and look at what McDonald’s actually built. The company took behavioral data (what Rogers ordered, when, where, how often), enriched it with contextual signals, and produced a predictive model of a customer: who he is, what he values, what he’ll do next, and how likely he is to churn. As Jeff Chester of the Center for Digital Democracy told WIRED, “McDonald’s secret sauce is really commercial surveillance” — but a marketer would recognize the underlying artifact immediately: It’s a buyer persona, rendered at the resolution of a single individual.
This is the direction the entire marketing industry is moving, and it has been for years. The traditional buyer persona — “Marketing Mary, 34, suburban, values convenience” — was always a composite sketch, an act of informed imagination built from surveys, interviews, and demographic guesswork. What the McDonald’s file demonstrates is that enterprise brands no longer need to imagine. They aggregate elaborate profiles of actual customers, then synthesize those profiles into segments and personas that are grounded in observed behavior rather than assumption. The persona is no longer a hypothesis.
It’s an output.
And critically, it’s a forward-looking output. The most telling detail in Rogers’ file wasn’t the record of his past Spicy Chicken Wrap purchases — it was the prediction of his next one. Loyalty data has become the raw material for behavioral forecasting, and McDonald’s leadership makes no secret of how central this is to the business: its CFO has called the loyalty program’s 210 million active users the company’s single most important digital metric.
Why this trend will only intensify
If you think the 515-page file is an outlier, the market data suggests otherwise. The global predictive analytics market was valued at roughly $18.9 billion in 2024 and is projected to exceed $82 billion by 2030, growing at a compound annual rate north of 28%. Retail consistently ranks among the leading sectors driving that growth. Every dollar of that spend is, functionally, a dollar spent making customer files thicker and predictions sharper.
Consumer expectations are pushing in the same direction. McKinsey’s landmark Next in Personalization research found that 71% of consumers now expect personalized interactions from companies, and 76% get frustrated when they don’t receive them. The same research shows that companies with faster growth rates derive 40% more of their revenue from personalization than their slower-growing peers, with personalization programs typically lifting revenue 10–15%. Deloitte’s consumer research similarly finds that shoppers report spending significantly more with brands that personalize well — and retailers responding by directing the majority of their marketing budgets toward personalization capabilities.
Read those numbers together and the logic becomes self-reinforcing: consumers reward prediction, prediction requires data, and data collection scales with every app install and loyalty enrollment.
The Wired article isn’t a scandal so much as a status report.
The 515-page file is what the median enterprise customer profile will look like — and then it will get longer. Companies will increasingly use these elaborate real-customer profiles not just to serve individual offers, but to construct the buyer personas and demand forecasts that shape product inventory, customer experience design, media planning, and creative strategy from the top down.
What this means if you’re not McDonald’s
Here’s the strategic problem for most brands: you don’t have 210 million loyalty members.
You may not have a loyalty program at all.
Enterprise ecommerce brands sit on rich first-party purchase data, but even they typically lack visibility into the part of the customer’s life that happens before the transaction — what their audience reads, watches, listens to, searches for, and trusts.
This is where modern audience research tooling closes the gap, and it’s why SparkToro has become a fixture in my own workflow with enterprise ecommerce clients. Rather than surveying a few dozen customers and extrapolating, SparkToro aggregates behavioral data — anonymized clickstream, search behavior, and public social profiles — to reveal where a defined audience actually spends its attention: the podcasts they listen to, the YouTube channels they watch, the newsletters they subscribe to, the websites they visit, the language they use to describe themselves. Its persona-generation workspace can then synthesize one or more audience reports into detailed, shareable buyer personas grounded in that observed behavior.
The philosophical throughline matters here. SparkToro’s own team has argued persuasively that demographic-first personas are increasingly irrelevant — the famous illustration being that Prince Charles and Ozzy Osbourne share nearly identical demographics while being wildly different buyers.
What predicts behavior isn’t age and zip code; it’s attention, affinity, and influence.
That is precisely the lesson embedded in the McDonald’s dossier: the company doesn’t care that Rogers is a millennial tech journalist. It cares what he ordered, when, in response to which offer — behavior, modeled forward.
When I build personas for ecommerce brands, the process now looks less like a creative writing exercise and more like intelligence synthesis: first-party purchase and CRM data establishes what customers do; audience intelligence from tools like SparkToro establishes who they are between purchases and where marketing dollars will actually reach them.
Research on audience intelligence methodology makes the stakes plain — brands that plan media against demographic assumptions systematically misallocate budget toward channels where their audience is merely present, rather than the niche podcasts, communities, and publications where it is concentrated.
The forecast is the point
The marketing and advertising industry is completing a decades-long shift from description to prediction. Descriptive analytics told us what customers did. Personas told us, in narrative form, who we believed they were. The next era — the one McDonald’s 515-page file announces — treats every customer profile as a living forecast: expected visits, expected spend, expected order, expected churn. Actively modeling and projecting future consumer behavior is no longer a frontier capability reserved for FAANG-scale data science teams; it is becoming the operating baseline for enterprise marketing, and the infrastructure to do it is compounding annually.
For marketers, the mandate is clear. Buyer personas are not a one-time deliverable to laminate and pin to the wall. They are dynamic models that should be rebuilt continuously from real behavioral data — your own first-party data where you have it, and audience intelligence where you don’t. The brands that treat personas as predictions to be tested and refined will increasingly out-target, out-message, and out-invest the brands that treat them as fiction.
Reece Rogers requested his file and discovered that McDonald’s had already answered the question every marketer asks: who is this customer, and what will they do next?
The uncomfortable truth of the story isn’t that the company knew so much. It’s that the prediction was probably right.
