How to build a cohort analysis that reveals real LTV

The number most brands get wrong

Ask a D2C founder their customer lifetime value and you will usually get one of two answers: a guess, or an average that hides more than it reveals. The average LTV across all customers, all time, is almost useless for making decisions. It blends your best buyers with one-time discount hunters and tells you nothing about whether things are getting better or worse.

Cohort analysis fixes this. It groups customers by when they first purchased, then tracks their spending over time. Once you see LTV by cohort, you can answer questions that actually matter: Are customers acquired this quarter better or worse than last quarter? Did that subscription push in March actually change behavior? Is your Meta spend bringing in people who come back, or people who vanish?

What a cohort actually is

A cohort is just a group of customers who share a common trait during a specific time window. For LTV analysis, the most useful cohort is first-purchase month. Every customer who made their first order in January 2024 is the January 2024 cohort.

You then track cumulative revenue from each cohort over the months that follow. Month 0 is their first purchase. Month 1 is the next calendar month. And so on.

The output is a table (or chart) that shows you:

• How much revenue each cohort generated in their first month
• How much they added in month 2, month 3, month 6, month 12
• Whether newer cohorts are tracking ahead of or behind older ones

This is the single most honest view of your business health.

How to build one in practice

You do not need expensive software. You need clean order data and a spreadsheet. Here is the process.

Step 1: Export your order data. Pull every order from Shopify (or your platform) with these fields: order date, customer email, order total (net of refunds and discounts), and whether it was the customer's first order. If your platform does not flag first orders, sort by customer email and find the earliest order date for each.

Step 2: Assign each customer to a cohort. Based on their first order date, tag them with a cohort month. A customer who first bought on March 15, 2024 belongs to the March 2024 cohort.

Step 3: Calculate months since acquisition. For every subsequent order, calculate how many months after their cohort month it occurred. An order in May from a March cohort customer is Month 2.

Step 4: Build the grid. Rows are cohort months. Columns are months since first purchase (0, 1, 2, 3...). Each cell is the total revenue from that cohort in that period. You can also express it as cumulative revenue per customer, which is more useful for comparison since cohort sizes vary.

Step 5: Look for patterns. This is where it gets interesting.

What to look for once you have it

• Month 0 to Month 1 drop-off. What percentage of first-purchase revenue shows up again in Month 1? If it is below 10%, your post-purchase experience or product is not compelling enough to drive a second order.
• Cohort curves flattening. Healthy brands see cohort revenue curves that keep climbing, even slowly, past Month 6. If your curves go flat after Month 2, your retention efforts are not working or you are a one-product brand without a reason to return.
• Newer cohorts underperforming older ones. This is the big red flag. If your February 2025 cohort is tracking 20% behind your February 2024 cohort at the same point in time, something changed. Maybe your acquisition channels shifted. Maybe a discount brought in lower-quality buyers. Maybe a product change hurt repurchase rates. You will not know until you look.
• Cohort spikes after specific campaigns. If you launched a new retention flow or a loyalty program, you should see it show up in the cohort data. Cohorts acquired before the change might show a bump in later months. Cohorts acquired after might show higher Month 1 and Month 2 numbers.

Common mistakes

Using revenue instead of contribution margin. Revenue cohorts are a start, but if your margins vary by product or channel, a revenue-based cohort can be misleading. Once you have the basic framework, swap in gross profit or contribution margin per customer.

Ignoring cohort size. A cohort of 50 customers with high LTV is not the same signal as a cohort of 5,000 with high LTV. Always look at both the per-customer value and the number of customers in the cohort.

Only looking quarterly. Monthly cohorts give you enough granularity to spot problems fast. Quarterly smooths out too much signal.

Treating LTV as a single number. Even after building cohorts, resist the urge to collapse everything into one LTV figure for your pitch deck. The value is in the trends between cohorts, not the average across them.

Why this matters for growth decisions

Every growth decision you make, from how much to spend on acquisition to whether a retention campaign is working, depends on knowing what a customer is actually worth. Not in theory. Not as a blended average. But segmented by when they arrived and how they have behaved since.

We build cohort models for every brand we work with at Argus because it is the foundation for knowing where to invest next. If you want us to build one for you and show you what it means for your ad spend and retention strategy, ask for a free growth plan.