I used to market chemistry journals to scientists. I know it sounds dull, but it actually taught me a lot. Yes, about chemistry (you’d be amazed at what you can make with shaving cream and food coloring), but also about relationship marketing.
Thanks to how science programs are structured, it’s easy to track the evolution of a scientist’s research needs and interests. Starting in undergrad, these individuals gradually become more focused on a specific area of interest. They start with a general field, like Chemistry, and narrow that expertise as they progress through to PhD.
From a marketing perspective, then, it was pretty easy to know some key, yet basic, information about my audience. Add to that face-to-face interactions at conferences and workshops, website searching behavior, and content consumption, and I was getting a well-rounded view of my customer.
Doesn’t that sound like segmentation? Taking a big group and making it smaller around common elements? I thought so, too.
But let’s just call that a simpler time, where my intel was good, my customers were happy, and I embarked on targeted multi-channel programs to drive success. What I was missing—and what remains a potential blind spot for marketers—is that simply making a group smaller doesn’t mean you’re creating segments.
But enough about chemists, let’s apply this lesson to ecommerce.
What is a segment?
First, let’s be clear on what a segment is. A segment is a subset of people who you can identify within your audience, that you can describe via their attributes and behaviors, and that you can reach with precision across channels.
Personalization platforms come with many “out-of-the-box” segments and, while it’s great to be able to distinguish your new visitors from your returning visitors (and you definitely should be doing this), these targets only address part of the interactions you’re having with a customer. As a marketer, you need a way to reconcile their behavior across channels, which is where segmentation becomes more powerful.
To create better ecommerce customer segments, you need to use three categories of information:
- Context. This is real-time data—like location, weather, and device type—and demographic data—like age and gender. All of these data points impact the mindset of your customer, and how they interact with your brand.
- Behavior. This is observed data—what your customer is doing on your site, what products, brands, categories they express interest in, and what actions they take, like adding items to carts, wishlists, or proceeding 90% through checkout only to abandon cart. All of this information gets you closer to understanding a customer’s buying intent.
- Relationship. This is the historical and cross-channel data that makes it possible for you to recognize and relate to your customer. Taking a backward look at the customer will also help you determine stability of your segment. Are they tied to a particular purchase cycle? Do they only shop with you during a particular season or time of day? Do they only make a purchase when presented with a promotional offer?
These categories, you might recognize, progress in the same way a scientist progresses from Chemistry undergrad to PhD: they start broad, then narrow in scope. And it’s not until you make your way through all three criteria that you can create better ecommerce customer segments.
Of course, not all of your customers are going to fall into a segment. That’s why there are natural breaking points.
Context provides you with the break points to form groups—a common thread of rain, female, college educated, midwestern, etc.— while behavior will show you how people differ within the group. This is also where you start to form better segments. Examine what types of behavior are taking place within the group, how often, and when—and you’ll start to zero in on your most valuable customers.
The next step is to analyze the relationship, which will bring forward the attributes, behavioral patterns, and even changes in context over time that help you understand when, and with what urgency a customer is ready to convert.
When it’s all said, done, and hopefully purchased, you’ll have a defined set of information on that previous group of customers, and a strong customer segment. And don’t worry if not all of your customers map to a segment, just give them time to evolve.
What is a group?
Before you can get to a customer segment, you need to start with something larger. That’s a group.
A group is a collection of visitors that I can tie together with a single characteristic, like chemists, visitors, and humans. But don’t let the simplicity fool you, even if you refine the details, you still wind up with a group, for example:
- Biochemists, Organic Chemists, Physical Chemists
- First time, returning, frequent visitor
- Men, women
So far, you’ve gathered some data about who they are, and how you can classify your customers. And if you were to apply these broad categories within campaigns, you’d wind up with some traction, but these groups will still be quick to point out how, and why, they are different.
When should a group become a segment?
A lot of that decision will come down to how homogeneous your group is.
Within the larger classification—for example, female—are behavior and decisions largely the same? If so, it’s probably safe to stick with groups for the time being, but keep an eye out for significant differences in behavior.
An example:
When a woman visits your site, and you direct her to dresses, you’re anticipating some basic category needs. While the characteristics (and stereotypes) may be right, the behavior and intent could be wrong.
I, for one, can’t remember the last time my fiance bought his own clothes. And as a result, in the scenario above, I am a woman who logically should be presented with dresses, but I am also a woman who purchases men’s jeans, shirts, sweaters, and, sometimes, shoes. (I highly doubt I am the only woman in the world who does this, by the way.)
This type of behavior, appearing in a subset of your group, may point you toward forming a segment. Let’s break it down into each data source:
- Context. You’ve formed some preliminary groups based on gender. In this case we’re focusing on female shoppers. There may be other elements to break it down further, like mobile, or located in New England, that are common threads within the group.
- Behavior. Behavior will start to poke holes in your original hypothesis. I know my recent session on a certain retailer’s website looks like this: Recently Browsed: Women’s Shoes, Recent Category: Petites, Recently Purchased: Men’s Sweaters, Abandon Cart: PID12334 (Women’s Boots). What I’ve done, by simply shopping as I normally would, is provide four different ways to break down the group in favor of segments.
- Relationship. This drives patterns, affinity, and loyalty. All of which influence online habits. Looking into historical data can uncover this information. I tend to purchase when there is a promotion for free shipping or a minimum of a 20% discount. It’s also been one year since I last purchased women’s jeans, which I’ve bought around this time for the past 4 years, and I have built an affinity for the petites category and men’s sweaters and t-shirts.
Thanks to this mish-mash of context, behavior, and relationship, we can start to look at my biggest motivators when shopping and which behavior is translating into repeat purchases for your brand. What emerges will be the attributes of a new segment, like “Relationship Shoppers.” Based on any one class of information above, context, behavior, and relationship, I would have been entered into campaigns that weren’t very relevant. But by combining them, you’ve determined that I tend to purchase more items for my finance on your site, and just browse for the occasional purchase for myself. I can now be targeted by appropriate campaigns.
But the most important thing about a segment is that they should prove value to your brand. Scientists would say publish or perish (a fancy way of saying “Innovate or die”) and I think it applies here, too.
Do the research to form meaningful segments. Look for the patterns, form a hypothesis (or, in this case, a segment), and validate through experimentation. And if all else fails, use the hard facts to form a new hypothesis and start the process again.
