LTV cohort analysis from your order CSV

The formula

For each cohort and period: retention = customers who ordered in that period divided by cohort size, and cumulative LTV = the cohort's cumulative revenue through that period divided by cohort size. An average cohort curve is computed as the size-weighted mean across cohorts, and an insight callout names the strongest cohort a few periods in.

In yearly view the tool also projects current-year and next-year revenue: unobserved periods follow the average cohort curve scaled by each cohort's strength factor (its observed revenue per customer versus the average at the same period, clamped between 0.1x and 3x), and incomplete years are annualized linearly. A backtest re-runs the projection on historical data only and colors the result green when it lands within 10% of what actually happened.

Expected input

A CSV with a header row, up to 1,000,000 rows, parsed in the browser. Three columns are required and auto-detected by name:

• –Date: matches date, order_date, purchase_date, created_at, timestamp, created, or day.
• –Customer: matches customer, user, email, client, buyer, member, or account.
• –Value: matches value, amount, revenue, total, price, spend, gmv, or sales.
• –Optional: an order ID column (order, transaction, txn, invoice, receipt). Without one, multiple same-day orders from a customer collapse into one.
• –Every unmapped column automatically becomes a filter and split dimension.

Assumptions and limits

• –Orders with zero or negative value are excluded, so refunds never enter LTV or retention.
• –A customer's cohort is fixed by their first order within your active filters.
• –Retention is activity-based: a customer can go quiet and count again later.
• –Results hand off directly to the Cash Flow Projection tool, prefilled with the retention curve and growth rate the data implies, plus a CPA placeholder to replace with your real acquisition cost.