Customer Churn Analysis: Set the Line Per Product

Customer churn analysis finds who stopped buying and where. In a store, nobody cancels, so you set the line. Set it per first product: the usual reorder gap plus 14 days. Then count churn by cohort.

That last step is the part most guides skip. They give you one churn number for the whole store. One number cannot tell you which product is losing people.

What is customer churn analysis for an ecommerce store?

It answers 2 questions. How many customers stopped buying? And where did they stop?

A churned customer is one who will not buy from you again. A subscription business knows the day this happens. The customer clicks cancel.

A store has no cancel button. A customer who has not ordered in 60 days might be gone. Or they might still have half a bottle left. So the first job is a rule. When does "late" turn into "gone"?

Every page we read sets that rule in a different way.

Where do the guides disagree on churn?

Here is how 4 sources on page 1 decide that a customer has churned.

Source When a customer counts as churned What it leaves out ThoughtSpot Customers lost ÷ customers at the start of the period What "lost" means when nobody cancels LoyaltyLion No 2nd order within the store's average repeat timeline. Its example uses 6 months Different products in the same store Polar Analytics Silent for more than 2 times the median gap between orders One line per store, tuned by category at most Klaviyo A churn risk score built from the number and timing of orders Which products the customer bought

The first one is a subscription formula. It assumes you can see who left. In a store you cannot, so the formula has nothing to count.

LoyaltyLion and Polar are much closer. Both use your store's own timing. Both draw one line for the whole store.

Klaviyo scores each customer, which sounds better. But its help center says its expected date of next order does not look at which products someone bought. A customer who bought a 30 day supply and one who bought a 90 day supply can get the same clock.

Why does one churn line hide the problem?

Because your store has more than one buying cycle. A serum that lasts 30 days and a cleanser that lasts 60 days do not run out on the same day.

Here is a made-up example to show the math. One month, 200 first-time buyers. Half bought the serum. Half bought the cleanser.

First product First-time buyers Usual reorder gap Churn line Past the line, no 2nd order Churn Serum 100 30 days Day 44 60 60% Cleanser 100 60 days Day 74 30 30% Whole store 200 90 45%

The whole-store number says churn is high. That is true, and it tells you nothing to do. The product rows say the serum is the problem. That tells you where to look on Monday.

One store-wide line also gets the timing wrong. Say your line is 90 days. A serum buyer who left on day 44 still counts as active for 46 more days. Your analysis reads low, and your winback starts late.

How do you run a customer churn analysis, step by step?

You need your order export and a spreadsheet. No data team.

1. Export your orders. In Shopify, export orders as a CSV. Keep the customer email, order date, product name and quantity.
2. Find each customer's first order. Note the date and the main product. If the first order had more than one product, pick a rule and keep it. The product that runs out first is a good one.
3. Find the reorder gap for each first product. Take the customers who came back. Work out the days from their 1st order to their 2nd. Use the median, not the average. One customer who came back after a year drags the average up. If only a few people bought a product, group it with a similar one.
4. Set each customer's churn line. First order date + reorder gap + 14 days. The 14 days is when our winback email flow starts. If they bought 2 units of something that gets used up, double the gap. The replenishment email flow explains why units matter.
5. Leave out customers whose line has not passed. They have not had a fair chance to come back. This is the same rule as in how to calculate repeat purchase rate.
6. Count churn by first product and by month of first order.

Churn rate = customers past their line with no 2nd order ÷ customers whose line has passed

You end up with a grid. Products down the side. First-order months across the top. Each cell is a churn rate.

What does the churn table tell you to fix?

Read the grid in 2 directions. Down a column compares products. Along a row compares months.

What you see What it likely means Where to look One first product churns far more than the rest The product did not match what the buyer expected, or the reorder reminder comes late The emails between order and delivery and the replenishment flow for that product Churn goes up month after month, for every product Newer buyers are a different crowd. A new ad channel or a big discount can do this Where those months' customers came from Many customers order a few days after their line Your line is too tight for that product Move that product's line out and run it again Churn is flat for every product and every month Nothing after the first order is changing Which flows you have live after the first order

What is a good churn rate for ecommerce?

There is no public number you can hold your store to. LoyaltyLion's 77% average comes from flipping one 23% retention figure from Omniconvert. Polar quotes 70% to 75% a year as a "widely cited" range and names no study behind it.

We checked the numbers on page 1 for retention in our ecommerce retention benchmarks article. Few of them had a sample you could trace.

So use your own grid. A good churn rate for your serum is lower than last month's churn rate for your serum.

Can Klaviyo predict churn for you?

Yes, if your account qualifies. Klaviyo's help center lists the bar. At least 500 customers who have placed an order. At least 180 days of order history. Orders in the last 30 days. Some customers with 3 or more orders.

The score is useful for sorting a list. It is not a churn analysis. It does not look at products. Klaviyo also says the predictions are "not expected to be exact for any single individual."

Use the score to find people to watch. Use your own grid to decide what to fix.

Where does churn analysis stop?

It tells you where customers leave. It does not bring them back. The flows after the first order do that. The ones that send between order and delivery, then replenishment, cross-sell and winback.

For how those fit together, start at the ecommerce customer retention hub.

Get the whole system

Your churn grid shows where customers leave. The flows after the first order are how you keep them.

12 flows, 37 emails, one job: repeat purchases. Every trigger, every send delay, every exit condition, written out so you can build them this month.