From Mainframes to AI: Why Geodemography Keeps Winning (ft. Jan Kestle, Environics Analytics)

In 2019 I flew up to Toronto to meet Jan Kestle , CEO of Environics Analytics . Since then I have admired Environics from afar. As I have dug into conversations with the early developers of Geodemographic / Psychographic segmentation like Richard Webber (Mosaic), Michael Weiss (PRIZM), I have come to fully appreciate that you can’t have a serious conversation about the science of consumer segmentation without hearing her name.

I had the opportunity to have a conversation with her that is part autobiographical and part retail insights. I walked away inspired, as does everyone who works with her.

Here are my favorite takeaways from our conversation on the enduring power of geodemographics: 👇

THE FOUNDERS OF GEODEMOGRAPHICS STARTED IN THE SAME UNLIKELY PLACE - AND THAT’S NO ACCIDENT

Most people assume the science of consumer segmentation was invented by marketers trying to sell more stuff. It wasn't. Three people responsible for building the foundational systems; Richard Webber, Jonathan Robbin, and Jan Kestle,  all started their careers doing government deprivation research. No way that is a coincidence.

Webber built what became ACORN and Mosaic by studying Liverpool neighborhoods to figure out which types of social services actually matched which types of communities, rather than giving everyone the same intervention regardless of local need. Robbin, who created PRIZM at Claritas, came out of a national deprivation study for the US government. Jan started at the Ontario Statistical Centre running federal-provincial negotiations on what data should even be collected, representing health departments, social services, economists, and finance simultaneously at the same table.

What they all had in common was the ability to look at a geography, understand the people living in it, and argue that a one-size-fits-all solution was the wrong answer.

That argument, made first in the context of social services, housing, and public health — is structurally identical to the argument geodemographics makes to retailers today. The right product, right place, right person was originally a social policy principle.

"I really got to understand how to tell stories with data. I had to represent the health department, the social services department, the economists, the financial transactions departments in negotiating with the federal government about what kind of data would be collected."

The tools these founders built weren't designed to sell more shampoo. They were designed to stop wasting resources on the wrong people in the wrong place. Retail inherited that logic, and the best practitioners still think about it that way.

Lesson: The common thread is their projects focused on quantifying and codifying everyone regardless of commercial value. This large scale picture gave them the ability to focus on the denominator first (total population) rather than the numerator (who I can sell to).

A 60 YEAR OLD FOUNDATIONAL PREMISE THAT DOESN’T EXPIRE

Although it is used today for retail targeting and site selection, geodemographics has its root in academic theory. Writings out of the University of Chicago School of Human Ecology in the 1960s established the core claim: if you know where people live, you can make statistically reliable inferences about what they're likely to do.

Every segmentation system built since: PRIZM, Mosaic, ACORN, Panorama, Tapestry, PersonaLive is an application of that same theory at increasing levels of granularity and data richness throughout the years. The reason this matters for retailers today is that the premise doesn't expire. Tools change. The theory doesn't.

"Whether you're using AI or whether you're using traditional statistical tools, the key thing is to make sure that the methodology embedded in your technology is giving you best practice."

Lesson: The platform is the vehicle. The methodology is the asset. Don't confuse upgrading one for replacing the other.

PEOPLE CHOOSE NEIGHBORHOODS - THEIR NEIGHBORHOOD SHAPE THEM BACK

The foundational premise of geodemographics is that people who live near each other tend to behave similarly. But Jan learned something sharper from Robin Page, the ad executive who joined Claritas in its earliest days.

People don't just move into neighborhoods that reflect who they are. They move into neighborhoods that reflect who they want to become. Aspiration is baked into the spatial signal. And then they tend to adapt to the neighborhood over time. The neighborhood shapes them back.

"People live in locations that have attributes and people and shopping and convenience that either are the attributes that are like them — so they move into a neighborhood because 'this is my kind of neighborhood' — or they move into a neighborhood because this is a neighborhood that they want to be like."

If you're doing trade area analysis or site selection and treating location as purely descriptive, a snapshot of who's there today, you're missing the forward-looking signal embedded in where people chose to live.

GEODEMOGRAPHY IS THE DENOMINATOR. MOST RETAILERS ARE ONLY USING THE NUMERATOR

In the late 1990s, as CRM systems proliferated and brands finally got their first-party data organized, the conventional wisdom was that geodemographics would be replaced. You had actual customer data now, why would you need neighborhood inference?

Jan and Bruce Carroll (Claritas co-founder) wrote a paper to answer exactly that question: Why in the World of One-to-One Marketing Would Anyone Ever Use Geodemography? The core argument: first-party data tells you about your customers. It cannot tell you about the people who are not your customers yet. Geodemography is the denominator, market share, untapped potential, competitive white space, that no CRM can generate on its own.

"Geodemography is the denominator. If you know who your customers are, you want to calculate market share. You want to calculate untapped potential."

This is probably the most under-appreciated characteristic of geodemography. If your brand is making site selection and marketing decisions purely on loyalty data, you're optimizing within your existing customer base with no view to where the growth actually lives.

The denominator is the growth map.

GEODEMOGRAPHICS AND PSYCHOGRAPHICS ARE POWERFUL WHEN INTENTIONALLY COMBINED

Segmentation systems use where people live as a proxy for how people think. Jan's co-founder Michael Adams, the pollster who documented Canadian-American attitudinal divergence in Fire and Ice, combined values-based surveys to Environics Analytics in a powerful way.

The methodology: build extremely granular geographic micro-segments called nanoclusters, then regroup them against Adams' social values survey data based on how they actually split on values, not just demographics. Two neighborhoods identical on income and age can land in different segments because their residents hold measurably different values.

"When you look at the propensity to hold certain values at the postal code level, and then you add it up over your whole customer database, we see trends of a correlation between demographics and values."

For retail, the difference matters most at the edges: two neighborhoods that look demographically identical but hold different values will respond to the same business, advertising and assortment in measurably different ways.

GEODEMOGRAPHY IS KEY TO PRIVACY COMPLIANCE

Here's the counterintuitive one. As US brands scramble to adapt to a post-cookie, post-IDFA world — trying to figure out how to do personalization without crossing into creepy — Jan's been solving this problem in Canada for 25 years under GDPR-equivalent privacy law.

The answer? Geography. Aggregating to a geo, or to a small cohort defined by a geodemographic cluster, gives you statistical reliability without personal exposure. The segmentation system becomes a privacy-enhancing technology, not just a targeting tool.

Brands that figure out how to use place and segment as the activating layer — instead of individual identity — will move faster and face fewer regulatory headwinds than those still trying to rebuild one-to-one stacks.

GETTING DEPARTMENTS TO SPEAK THE SAME LANGUAGE

If there is one thing we can learn from the biblical story of the tower of babel is innovation happens at the speed of communication. And if everyone is speaking a different language about their consumer things are going to slow down.

Real estate uses one dataset. Marketing uses another. Finance uses a third. They can't argue from the same facts because they're not looking at the same map. Geodemography fixes this by providing a common lens that can reference their consumer.

Jan described what's happened at Environics' client base over the past decade: users used to live in real estate or marketing. Now they span real estate, marketing, finance, advertising, corporate strategy, and brand — all using the same segmentation system as a shared language.

"One of the things about geodemography — it levels the playing field. It allows you to look at disparate data sources through the same lens. Data that you couldn't previously join together gets joined together."

This is an organizational capability, not just an analytical one. When a site selection team and a marketing team can argue about the same customer segment in the same vocabulary, the gap between "we found the location" and "we know how to open it" collapses.

WHY GEODEMOGRAPHICS HAS SURVIVED EVERY TECHNOLOGY WAVE THAT WAS SUPPOSED TO KILL IT

Every decade since the 1970s has produced a new technology that was supposed to make neighborhood-level inference obsolete. Desktop computing. CRM. Digital advertising. Programmatic. Mobile. Each one triggered the same panic inside the geodemographics industry.

"My God, you're dead now that people have got their own customer data. Then they said, my God, you're dead now that we're going digital. And now AI."

It didn't die. Each time, it absorbed the new technology and got more useful. Desktop GIS made segmentation maps accessible to non-analysts. CRM integration gave geodemographics a customer file to append, turning neighborhood inference into individual-level enrichment. Digital advertising created a new activation layer, letting practitioners take a postal code segment and find those same people in programmatic, connected TV, and social. Mobile movement data added a temporal dimension, not just who lives where, but who shows up where and when.

The reason geodemographics survives every wave is structural: it's the only methodology that works as a denominator across all of them. CRM needs a population baseline to calculate market share. Digital targeting needs an audience definition that doesn't rely on third-party cookies. Mobile data needs a profile layer to mean anything beyond a headcount. Geodemography provides all three.

Lesson: Geodemographics has outlasted every platform built on top of it because it's not a platform. It is not a technology. It is a reflection of something more fundamental. We are social creatures, and we relate to those around us. Where we live shapes who we are. And our aspirations shape where we live.  It has been true these past 60 years. And it was true thousands of years before that.

IN THE AGE OF AI, BAD DATA DOESN'T JUST GIVE YOU A WRONG ANSWER, IT GIVES YOU A CONFIDENT WRONG ANSWER AT SCALE

Jan mentioned a newsletter from her 1990s writing. The article could have run last week: the quality of data determines the outcome of the model, not the sophistication of the modeling technique. AI made this more dangerous to ignore.

The implication for any retailer building AI-powered tools on top of geodemographic or behavioral data: the model's confidence will scale with its training, not with its accuracy. If the input data has gaps, the AI will propagate those errors at a speed and scale that a human analyst running a regression would never reach.

"The quality of the data is the thing that really determines the outcome of a model. It is the data itself that make a difference."

The adjacent trap Jan flagged: organizations that build research on the most sophisticated methodology available, then realize they can't actually activate against it. You have to level your insights to what you can execute on. A $500K segmentation model that can't connect to your media buying desk is a shelf trophy.

MARKET RESEARCHERS SHOULD STOP APOLOGIZING FOR WHAT THEY DO

I loved this. After getting sideways glances when I tell people I am in the data and market research industry, I have just started telling people I am in “data science”. In our industry we tend to soften or apologize for what we do.

Jan doesn't do that.

"If you've got to pick up dinner on the way home, it's important for the right products to be in the right place. It's important for you to get the right message. I sometimes get in trouble when I argue that the business and the marketing purposes of data make people's lives better. But I believe it does."

She's right. The founders of this field came out of government deprivation research because they believed understanding place was a form of progress, that treating every market identically was wasteful and disrespectful to the people living in it. That argument was social before it was commercial.

The nation's retail infrastructure, public health systems, and housing policy all run on the methodology this industry built. We should own it.