July 28, 2026 · The Evolution team

Customer Scoring for Shopify: Predicting Who's About to Buy (or Leave)

You have a list of customers, and you're treating all of them the same way: same emails, same discounts, same follow-ups. But some of them are one email away from buying again, and others already checked out mentally three weeks ago. You just can't tell which is which without digging through data you don't have time to dig through.

That's what customer scoring solves, and it's becoming a lot more accessible for small stores in 2026.

What customer scoring actually is

Customer scoring assigns each person on your list a number, often out of 100, based on how likely they are to do something specific: buy again soon, respond to a discount, or churn entirely. It's built from behavior you already have in your store's data:

• How recently they purchased
• How often they've purchased
• How much they've spent
• Browsing activity since their last order
• Email engagement (opens, clicks, ignores)

Bigger platforms have used versions of this (RFM scoring: recency, frequency, monetary value) for years. What's changed is that AI-driven predictive scoring, the kind that looks at patterns across your whole customer base and flags "likely to buy in the next 7 days" or "likely to churn," used to require a data team and a six-figure budget. Now email platforms are building it directly into tools small stores already pay for.

Why this matters more than another blanket sale

Here's a simple, realistic example. Say you have 4,000 people on your list. Without scoring, a typical move is to blast everyone with 15% off to drive a slow week's revenue. That works, but it trains your best customers to wait for discounts, and it wastes the discount on people who were never going to buy anyway.

With scoring, you'd instead see something like:

• 300 customers scored "high, ready to buy": send them new arrivals or a small nudge, no discount needed
• 800 scored "medium, needs a push": this is where a discount or urgency actually earns its keep
• 900 scored "low, drifting toward churn": a win-back sequence, or just a "we miss you" check-in, is more appropriate than a coupon
• The rest: low engagement, best left alone rather than burning list health

Store owners who've segmented this way typically report they can cut discount volume by a third while keeping revenue flat, because the discount is going to people who actually needed it to convert, not to people who would have bought anyway.

How to start without a data team

You don't need custom modeling to get most of the value. Here's a version any $50k-$1M store can run:

1. Pull recency and frequency first

Even a basic customer list export with last order date and order count gets you 70% of the way there. Anyone who hasn't ordered in 90+ days and used to order every 30-45 days is a churn risk, full stop. That's rule-based scoring, and it works.

2. Layer in engagement

Add email opens and clicks over the last 60 days. Someone who stopped opening emails is a stronger churn signal than someone who's just been quiet on purchases but still reads what you send.

3. Use your platform's built-in scoring if you have it

Several email platforms rolled out "likely to buy" or predictive send-time features this year that do the recency/frequency/engagement math for you automatically. If your platform has this, turn it on before building your own spreadsheet version. Check the settings for anything labeled predictive analytics, propensity scoring, or customer insights.

4. Act on the tiers differently

This is the step most stores skip. Scoring is useless if everyone still gets the same email. At minimum:

• High scorers: no discount, just relevant new product or restock alerts
• Medium scorers: your best converting offer
• Low scorers: a win-back sequence, not a generic newsletter

What to do with each tier, beyond email

Scoring isn't only useful for deciding who gets a discount. It should change how you spend your time and attention across the whole store, not just what lands in someone's inbox.

For high scorers, consider early access to new drops or restocks before the general list sees them. This costs you nothing and reinforces that being a good customer gets rewarded with access, not just discounts, which is a stronger long-term retention lever.

For medium scorers, this is where a personal touch matters more than automation. A short, specific message referencing what they bought last time ("thought you might like this since you got the X last month") converts better than a generic offer, and it doesn't take long to write once you know who's in this tier.

For low scorers, be honest about the math. Chasing a customer who hasn't opened an email in six months with increasingly aggressive discounts usually costs more in margin and deliverability risk than it recovers in revenue. A single well-timed win-back attempt, then letting them go quiet on your list (or suppressing them entirely), protects your sender reputation for everyone else.

Where this connects to profit, not just marketing

Scoring isn't just a marketing exercise. It changes your CAC vs LTV math directly: every dollar of discount you stop wasting on customers who'd have bought anyway is a dollar back in margin. It also feeds into how you think about true profit per order, since discount-heavy segments always look better on revenue than they do on actual profit.

The honest limitation

Scoring models get better with more data and more history. A brand-new store with 200 customers won't get much signal from this yet, rules of thumb (recency and frequency) will do fine until you have more volume. Don't buy a fancy predictive tool before you have enough customers for it to learn from.

There's also a difference between a score being accurate and a score being useful. A model that's 80% accurate at predicting who'll buy in the next week is still wrong one time in five, so treat scores as a prioritization tool, not gospel. Use them to decide who gets attention first, not to write off customers entirely based on a single number.

For stores past that early stage, though, this is one of the highest-leverage things you can set up, because it doesn't require more traffic or more spend. It just means spending your existing list more intelligently.

Keeping this running well, checking scores, adjusting tiers, catching a segment that's silently sliding, is exactly the kind of ongoing maintenance that gets skipped when you're running the store solo. It's a natural fit for an AI COO to monitor in the background so nothing drifts for months before you notice.

Want a free audit of your store? Get yours here.