Most ecommerce brands still rely on broad segments like engaged customers, recent buyers, VIPs or lapsed buyers. They’re useful, but each one tells you only one thing about the customer.
The problem is that one thing is rarely enough to decide what someone should actually receive.
McKinsey’s research found that 71 percent of consumers expect personalized interactions and 76 percent get frustrated when they don’t get them. A campaign to “all engaged” is not personalized communication. It’s a broadcast with a filter on it.
The audiences actually worth sending emails to, sit inside your Shopify store data. Getting to those requires reading several metrics at once, and that’s precisely where manual segmentation stops being practical.
One-dimensional segments hide the differences that matter
Two customers in the same segment can need different things.
Take the “engaged” audience. Inside that bucket sits a high-LTV customer who opens everything and hasn’t bought in ninety days, and a low-LTV customer who opens everything and bought last week. The first is drifting and worth protecting, the second needs nothing at all right now, and sending them a discount costs you margin on a purchase that was already coming your way.
That’s the structural problem with one-dimensional segmentation. It groups people by one shared fact and treats every other difference between them as irrelevant. The differences it discards: purchase recency against LTV, category interest against browsing behavior, and many more are usually the ones that determine what the right message is.
It also means the biggest segment in most programs is the least useful one. “Engaged” typically covers a large share of the active customers, which makes it a distribution list rather than a targeted segment.
The metrics that actually define a customer
A few signals read together describe someone well enough to know what to send them.
- Engagement - opens and clicks, and more usefully the direction of engagement. Engagement declining over six weeks should be treated as a different signal from engagement that’s steady and low.
- Category and product interest - what they’ve actually bought and browsed, at the level of the product.
- Recency - days since last order read against that customer’s own cadence rather than a fixed 30/60/90 day period.
- Lifetime value - what the relationship is worth, which determines how much margin is worth spending to keep that customer.
- Purchase frequency - how often they buy, which is what makes recency interpretable in the first place.
- Discount sensitivity - whether their orders cluster around promotions or happen regardless. This one is rarely tracked and it’s the difference between a discount that buys a sale and a discount that gives one away.
Why doing this by hand doesn’t work
The problem isn’t knowing which segments to build. It’s keeping them updated.
Once you start combining engagement, LTV, recency, purchase frequency, category interest and discount sensitivity, the number of possible customer groups grows quickly.
A retention team then has to decide which ones matter, build each segment inside the ESP, make sure customers aren’t being targeted twice, and keep updating the rules as customer behaviour changes.
That’s why most teams eventually fall back on a few broad segments.
In our experience, most brands come to Nexie with five or six segments they use regularly. Not because five or six is the right number, but because that’s what the team can realistically manage alongside campaign planning, creative, reporting and everything else.
The more useful segments often exist in a planning document somewhere, they just never get built.
What AI changes about segmentation
AI makes it easier and practical to make the most out of the customer data you already have.
Instead of looking at engagement, recency or LTV separately, AI can read those signals together and find groups of customers who are behaving differently.
That changes a few things.
It can look at several signals at once.
A customer isn’t just “engaged” or “lapsed.” They might be a high-LTV customer whose engagement is still strong but whose last purchase is later than usual. That combination is much more useful than any one of those labels on its own.
Segments can keep updating themselves.
As customers buy, browse, stop engaging or move past their usual reorder window, they can move into and out of the relevant audiences automatically.
That matters because customer behaviour doesn’t stay fixed between planning meetings. A retention marketing agent can keep watching for these changes instead of relying on someone to rebuild the segment every week.
You can actually use more segments.
A team that could only maintain five or six broad audiences can now work with a larger set of specific ones without creating more manual work.
That doesn’t mean creating hundreds of tiny segments.
It means having enough useful audiences to make different decisions for customers who actually behave differently.
Multi-dimensional segmentation comes into play
What that looks like in practice: Take a pet brand with a healthy customer base.
Instead of sending one monthly promotion to most of the engaged list, the brand could find very different groups inside that same audience.
1. High-LTV dog food buyers who are past their usual reorder date
These are good customers who are slightly late to reorder.
They don’t necessarily need a discount. A simple replenishment reminder may be enough.
If they were sitting inside an “all engaged” segment, they might receive the same 15% promotion as everyone else, even though they were likely to buy anyway.
2. Repeat treat buyers who are browsing supplements
These customers already buy from the brand and are showing interest in another category.
That’s a clear cross-sell opportunity.
Instead of sending another general promotion, the brand can explain why the supplement is useful and recommend something relevant to the customer.
3. Mid-LTV cat owners who haven’t bought in 60 to 90 days but still open emails
They’re still paying attention. They just aren’t buying.
That calls for a different message from someone who has stopped engaging completely.
A reactivation campaign could introduce a different product, benefit or reason to come back instead of simply sending another discount.
4. First-time buyers who are still highly engaged
The next purchase matters a lot here.
Someone who has bought once should not be treated the same way as someone who has been buying for three years.
The goal is to help them make that second purchase while interest is still high.
| Segment | Signals | What it tells you | Campaign |
|---|---|---|---|
| High-LTV buyer past reorder date | LTV, engagement, category, recency | Good customer, slightly late | Replenishment reminder |
| Treat buyer browsing supplements | Purchase history, browse behaviour | Clear cross-sell interest | Educational cross-sell |
| Lapsed customer still engaging | LTV, recency, engagement | Still interested, but not buying | Reactivation |
| First-time buyer still engaged | Order count, engagement, recency | Opportunity to drive purchase two | Second-purchase campaign |
None of these are complicated ideas, it’s just that you only see them when you look at several customer signals together.
What this does to campaign planning
The question stops being what to promote and becomes who to promote to.
Most ecommerce email calendars start with what the brand wants to send. There might be a promotion coming up, a new product to launch or a category that needs an extra push, and once the campaign is decided, the team chooses an audience for it.
There is nothing necessarily wrong with planning this way, but better segmentation gives you another place to start.
Instead of only asking what the brand wants to promote this week, you can also look at what customers are doing. Some repeat buyers may be approaching their usual reorder date. Another group may have started browsing a category they have never bought from. Some first-time buyers may be reaching the point where a second purchase usually happens, while previously reliable customers may suddenly be buying less often.
Each of those groups gives you a reason to run a campaign.
The difference is that the campaign starts with something happening in the customer’s behaviour rather than something that was already sitting on the marketing calendar. You know who you’re speaking to and why you’re speaking to them before deciding what the email should say.
It also makes performance easier to understand. If several types of customers are grouped together and receive the same campaign, the final result is an average across all of them. You know how the campaign performed overall, but you learn much less about why different customers did or didn’t respond.
More specific audiences give you more specific results, which makes it easier to see which campaigns are actually worth running again.
McKinsey has linked effective personalization with a 5 to 15 percent increase in revenue and a 10 to 30 percent improvement in marketing ROI. Personalization at that level isn’t only about changing the content of an email. Who receives it, and why they’re receiving it, matters just as much.
The constraint was never strategy
Retention teams already know there are better ways to segment their customers.
They know that a valuable customer who is starting to lapse should probably be treated differently from someone who was never a frequent buyer. They know a customer browsing a new category is worth separating from someone who isn’t.
The difficulty has always been doing this at scale.
Every additional segment has to be defined, built and kept updated, and someone still needs to decide what campaign that audience should receive. Once a team is doing this across enough customer groups, the work piles up.
AI removes a lot of that manual work. It can look at several signals together, keep audiences updated as customer behaviour changes and help turn those audiences into campaigns.
This means teams don’t have to make segmentation complicated anymore. They can simply leverage AI to read the data and understand which customers should be treated differently.
Nexie is a retention marketing agent for ecommerce brands. It reads customer data across engagement, recency, category interest, purchase behaviour and lifetime value to identify useful audiences, then builds campaigns for your team to review and approve. See your segments →