Natural clustering:
get in the party mood with next-level segmentation

Customer segmentation has many layers – and the ultimate level, natural clustering, is often hardest to understand. So let’s use a metaphor: you’re at a party. A big party, with live music, great weather, and a barbecue pit the size of an asteroid crater. Everyone in town’s been invited, and they’ve all turned up. 

Some people know each other; some don’t. Some are longterm neighbours, others just visiting. But as the hours go by, you notice people have gathered into definable groups. Gardening fans around the planters, car fanatics over by the garage, students near the bar tent, that sort of thing.

Nobody told them to form into groups. Nobody even knew what the groups would be. But by reasoning and inference, everybody’s joining a set of people with whom they have some commonality. Any marketer knows this as “segmentation” – seeing a population as a set of groups based on data. But what really unites each group at a party? 

Some data is easily discovered, like age and geographic location. Other data is harder, like dreams and desires. Here’s the thing: that “harder” data is where the real value is. And at Distil we’ve built a business on putting that less-obvious data to work. 

Here’s how it works – plus some background on segmentation itself.

Party on: the four levels of segmentation

Let’s recognise this first: “the customer” is not a personality type. Or a database record. Or a bunch of Excel cells. Customers are people – and people are more complicated than that. 

When they take a decision to buy, it may have one obvious driving factor, or a thousand little ones. The decision that drove a first purchase may be totally different to the reasons second time round, or they may buy a totally unexpected product on a whim. The same customer can even look like different people, depending on the time of day or the mood they’re in. 

Which means segmenting customers isn’t a one-off exercise. It’s ongoing and constant. The good news: segmenting them in the right way can make all these varying behaviours useful – if you apply technology in the right way. 

Key is to understand segmentation has four different levels. Short version: the higher you climb, the closer you’ll get to customer segmentation nirvana.

First time at the party: beginner level segmentation 

Going somewhere for the first time can be difficult, but it’s still possible to have fun even if you don’t know anybody. In this situation people tend to gravitate to “kindred spirits” – people they feel are like them, for example who are roughly the same age or similarly dressed. 

Any marketer will recognise this as the oldest form of segmentation: demographics, based on age, sex, family status and income. It’s still useful to many marketers – if you’re selling stairlifts and funeral plans, you’ll never be sacked for focussing on people 65+.  

In the world of ecommerce, this basic level of customer segmentation can also be applied using aspects of a customer’s record such as product type and new or repeat purchase. 

The great thing about a demographic approach is that even if there’s an element, like age or income, missing from your data, you can infer based on location. 

That’s where Distil comes in: our Customer Data Platform (CDP) can join different bits of data into a single view of your customer across all their interactions with your business, often keyed off a postcode. Simply put, it’s called Distil Data Enrichment.

Public databases such as Britain’s voting register give fine-grained detail on average ages and incomes in each postcode, letting marketers reach out to separate demographic groups with ease; many marketers never use anything more sophisticated, because it keeps things simple. 

There are, however, problems with stopping here. This approach to segmentation segmentation divides not unites.

Plenty of older adults love music and movies, many women are as excited by fast cars and barbecue grills as men. So focus only on stereotypes, and you’re missing a big chunk of your marketing opportunity. That’s why Distil goes a lot further. 

Experienced partygoer: engaged conversations

Demographics was all the rage until the 80s or so. But since the 90s the psychographic approach to segmentation has taken priority. Not when customers were born or where they live – the stuff that separates them – but what brings them together, like shared interests and similar behaviours. 

Psychographic grouping makes for more enjoyable parties. Because you’ve got people from different walks of life talking excitedly about their mutual interests, making new connections and friends. But it’s hard to see these connections at a glance, which is why the best party hosts are obsessive about seating plans and introductions. They’re constantly trying to create value by sparking new conversations. 

This is what makes the psychographic approach higher-level: it treats people as human beings, not bits of data. 

The customer isn’t “Women 18-30”, it’s “Loving Thai takeout” or “Always down the DIY store.” The things that make life meaningful. And the Distil CDP adds value here, too, with marketing analytics that show you who’s really buying your stuff. The built-in AI can tell you which groups look like your best customers, your highest value purchasers or are at risk of leaving you.

But there’s a fresh problem with going psycho: it gives you the “what”, but not the “why”. Why do those people buy packaged drinks on a Wednesday but not a Friday? Why do people spending £100 with you in April suddenly stop in May? Why do people make a big first purchase but a much smaller second one? 

The Distil App for Shopify – which is free by the way hints at the answers. But in our experience, once you have a taste, you’ll want to go further. So let’s get back to our four-step pyramid and climb to the third level: advanced segmentation.

Life and soul of the party: making new connections

Take the ultimate party, Hollywood’s Oscars. What seems like a star-studded extravaganza of glitter and glamour isn’t random at all. Every square-jawed superstar’s table plan, every fresh-faced sensation’s route down the red carpet is planned in incredible detail. Shake hands with this producer here; pose for pictures at this precise spot; meet this journalist one minute later. (And one minute doesn’t mean 50 or 70 seconds.)

Why are all these interactions planned? Because each one is designed to maximise exposure for a specific star in a specific way, showing them off to precisely the right audience so their movie will be a success. The footage from one red carpet walk will be cut into over a hundred clips for different TV stations and streaming services, each tailored in dozens of ways. 

Organisers can do this because they use advanced segmentation: actively maximising your marketing opportunity based on insights from data.

That’s why advanced segmentation is our next-level. When you can see multiple customer interactions in a single view, as with Distil’s CDP, it lets you see insights and inferences from that Big Picture, things you hadn’t thought of before. Imagine being able to pick out your premium buyers, those who might be ready to become your social fan boys and girls, and those customers who prefer chicken korma over chicken tikka masala despite having bought both in the past? Now that would be worth knowing, right?

For instance, is the young male audience you imagined you had actually 70% female? Are your treasured Millennials low-margin profit-eaters, and richer Boomers actually your best market? Distil can tell you – enabling you to focus your marketing spend on the people with the greatest profit potential.

But there’s one level above advanced. And for web retailers today, it’s the most exciting.

Boss level: throws all the best parties

Woodstock ’69. Manchester in the 80s. Glastonbury Pyramid Stage on a Sunday night, well, any year. The greatest parties of all time are legendary because they aren’t just events or venues; somehow, they created something new. (Even if in reality Glastonbury some years is mud and sunburn in equal measures!)

And that’s our fourth level of segmentation: teasing out that secret ingredient that makes it special and different. It can’t be planned for; it’s an emergent effect of all your activities interacting. But by putting Artificial Intelligence to work, it can be discovered.

We call it natural clustering – and it’s the Boss Level of segmentation, letting machine learning algorithms work out entire new audience definitions across your customer base so you can target your marketing ever more precisely. 

Is there a correlation between time of day and basket abandonments? Or a connection between source of lead and total spend per year? Or between the background colour of your product shot and its appeal to different demographic and psychographic groups? These clusterings are real – and with Distil AI, you can use them. 

Unknown to you, there are ideas within your data that can add $$ to your profit margins, boost your conversion rates into the double digits, raise your sales by a third. These benefits come not from pre-planned audience personas and social groups, but from letting whole new categories of segmentation bubble up naturally – including ones you’d never imagined your customers had in common.

And that’s natural clustering. It’s the ultimate segmentation, combining the capabilities of the Customer Data Platform with the power of AI.