How to Predict Customer Churn on Shopify: A Practical Guide for 2026

Customer churn is the silent revenue killer in ecommerce. Unlike a subscription business where a customer cancels and you know immediately, ecommerce churn is invisible. A customer just stops coming back, and you usually have no idea it is happening until you run a report months later.

The good news is that churn is predictable. There are clear signals in customer behaviour that indicate someone is at risk before they fully disengage. This guide explains how to read those signals and what to do with them.

Why churn is more expensive than you think

Acquiring a new customer costs five to seven times more than retaining an existing one. But the real cost of churn goes deeper than acquisition cost.

A retained customer has increasing lifetime value. They buy more frequently over time, their average order value tends to grow as trust builds, and they are more likely to refer others. When you lose a customer to churn, you lose not just their current spend but all of their future spend and referral potential.

For a store with a 30% annual churn rate, roughly a third of your customer base needs to be replaced every year just to stay flat. Most of your marketing budget goes toward replacing customers you already had rather than growing your base.

Understanding RFM: the foundation of churn prediction

RFM is a customer scoring method that has been used in direct marketing for decades and translates very well to ecommerce. It stands for Recency, Frequency, and Monetary value:

• Recency: how recently did this customer last place an order
• Frequency: how often do they buy
• Monetary: how much do they spend in total

A customer with high scores on all three is your best customer: they bought recently, they buy often, and they spend a lot. A customer whose Recency score is declining while their Frequency and Monetary scores are historically high is a classic churn risk. They used to be a great customer but they have not bought in a while, and that gap is widening.

That pattern is the early warning signal. If you catch it while the customer is still within their normal purchase window, you have a real chance of re-engaging them.

Identifying your churn threshold

Before you can identify at-risk customers, you need to define what "at risk" means for your specific store. This depends on your average purchase frequency.

Calculate it like this: take your total number of orders over the last 12 months and divide by the number of unique customers who placed those orders. That gives you average orders per customer per year. Divide 365 by that number and you get your average days between purchases.

A customer who has not bought in 1.5 to 2 times your average days between purchases is entering the at-risk window. A customer who has not bought in 3 or more times that window is likely already churned.

For example, if your average days between purchases is 45, customers who have not bought in 70 to 90 days are at risk. Customers who have not bought in 135 or more days are probably gone.

The churn signals to watch

Days since last purchase is the primary signal but not the only one. Other behavioural indicators that a customer is drifting:

• Declining email open rates: they are still on your list but no longer engaging with it
• No response to recent campaigns that historically drove purchases
• A return or refund on their most recent order, since dissatisfaction is a strong churn predictor
• Browsing without buying: sessions on your store with no add-to-cart activity
• Purchasing less frequently than their historical average

Any one of these signals is weak on its own. Multiple signals appearing together in a customer who was previously high-value is a strong indicator that warrants intervention.

How to automate churn prediction

Running RFM analysis manually in a spreadsheet is possible for a small customer base but does not scale. As your store grows, you need the analysis running automatically so at-risk customers are flagged in real time rather than in a quarterly report.

Miko AI runs RFM scoring automatically across your entire Shopify customer base and updates scores as new orders come in. It identifies customers in each risk category and syncs those segments directly to Klaviyo as tags, so your win-back email flows can trigger automatically when a customer enters the at-risk window.

Building a win-back sequence

Once you have identified at-risk customers, the intervention is a targeted win-back sequence. The goal is to give them a specific, relevant reason to come back before they fully disengage.

Email 1: The gentle nudge

Sent when a customer first enters the at-risk window. This is not a "we miss you" email. It is a useful, personalised message: "Based on what you bought last time, here is something new you might like." Reference their actual purchase history. No discount needed at this stage, because you are testing whether they are still engageable before spending margin on an incentive.

Email 2: The incentive

Sent 10 to 14 days after Email 1 if no purchase. If the nudge did not work, add a meaningful offer. Not a token 5% discount, but something that actually feels worth returning for: 15 to 20% off, free shipping on their next order, or a loyalty points bonus. Make it time-limited to create urgency.

Email 3: The last chance

Sent 10 to 14 days after Email 2 if still no purchase. This is your strongest offer and final attempt. Some merchants use a "we are about to remove you from our list" message as a way to drive either a purchase or a clear unsubscribe. This keeps your email list clean and your engagement metrics honest.

Customers who cannot be saved

Not every at-risk customer can or should be won back. Some customers only ever buy once. Some had a bad experience and have genuinely moved on. Spending marketing budget chasing customers who have definitively churned is wasteful.

If a customer does not respond to your three-email win-back sequence, suppress them from future marketing campaigns. Move them to a once-yearly reactivation attempt if you want, but stop including them in your regular email cadence.

Measuring the impact

Track these metrics before and after implementing churn prediction:

• Repeat purchase rate: the percentage of customers who make a second purchase
• Average days to second purchase: is it getting shorter after you launch the win-back flow
• Win-back rate: of customers who received the sequence, what percentage converted
• Customer lifetime value: is it growing over time across your base

Churn prediction and loyalty programs are two sides of the same coin. Churn prediction tells you who is leaving. A loyalty program gives them a reason to stay. If you want to build the proactive side of retention, read our guide on how to set up a loyalty program on Shopify.

Questions