Calculating customer lifetime value when purchase cycles vary by category

Customer lifetime valueCustomer lifetime valueLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → is one of those metrics that every retail marketing team reports and almost none calculates the same way. The concept is not complicated. Multiply average order value by purchase frequency, adjust for margin and churn, and you get a figure that tells you how much a customer relationship is worth over time. The problem is that "purchase frequency" is doing enormous hidden work in that formula, and in retail it varies so wildly across categories that a single number per customer is often meaningless.

A shopper who buys coffee capsules from your grocery operation every twelve days and a mattress from your home furnishings division every eight years are both "customers." Treating them with the same LTVLTVLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → model will lead you to underinvest in the coffee buyer (whose value accumulates in small, frequent increments) and either overvalue or abandon the mattress buyer (whose single transaction looks thin until you account for the associated accessories, protection plans, and the decade-long referral window it opens).

Why this matters for a CMO specifically

Retail economics punish imprecision here in ways that other sectors do not. Margin structures differ sharply by category: fresh food margins often sit below 5%, whereas own-label household goods or branded cosmetics can exceed 40%. When you roll LTV across those categories into a single customer metric, you flatten the margin signal entirely.

Budget allocation decisions follow the LTV number. If your model tells you that a pet food buyer and a consumer electronics buyer are worth the same over three years, you will set acquisition and retention budgets based on a fiction. In reality, the pet food buyer generates predictable monthly revenue and responds well to subscription mechanics, while the electronics buyer has a much longer inter-purchase gap and is at highest risk of defection precisely when they are not buying anything at all.

Retail media is adding pressure to get this right. As Digiday has reported, retail media networks are multiplying and competing hard for ad spend, which means CMOs are increasingly being asked to demonstrate closed-loop returns at the category level. You cannot credibly show a CPG brand that its sponsored placement in your coffee aisle produces $X of lifetime valuelifetime valueLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → uplift if your LTV model does not have a coffee-specific denominator.

First-party dataFirst-party dataData collected directly from your own customers and prospects through your own channels: your most reliable and privacy-compliant source.View full definition → from loyalty programs is the input that makes category-level LTV possible, but it also introduces a compliance dimension. The loyalty card economy rests on data collection that sits inside GDPR and, in the UK post-2026, the Data Protection and Digital Information Act framework. What data you can retain, for how long, and how you can model on it is not purely a technical question.

How it actually works

The starting point is to abandon a single customer LTV and instead build a matrix: one LTV calculation per category cluster, applied at the customer level based on that individual's actual mix of purchases.

Take a simplified example from a format like Tesco or Carrefour, which span grocery, clothing, home, and fuel. You define category clusters not by department but by inter-purchase interval. Grocery replenishment items (milk, bread, coffee) have intervals measured in days. Personal care sits at two to four weeks. Clothing is seasonal, so roughly four purchases a year on average. Large appliances and furniture operate on multi-year replacement cycles.

For each cluster, you calculate:

• Gross marginGross marginGross margin is the share of revenue left after subtracting the direct cost of producing goods or services, expressed as a percentage of revenue.View full definition → contribution per transaction (not revenue, because a high-frequency low-margin category looks very different from a low-frequency high-margin one)
• Expected purchase frequency over a defined horizon, typically 24 or 36 months
• Category-specific churn probability, meaning the probability a customer stops buying in that cluster entirely within the horizon

The formula for a single category cluster per customer is: (average margin per transaction x expected transactions in horizon) x (1 minus churn probability).

A customer's total LTV is the sum across all clusters they have purchased in, weighted by recency. A buyer who has not touched the clothing section in 18 months gets a low recency weight on that cluster, reflecting genuine behavioural drift rather than assumed loyalty.

This is wherereading retention curves correctly becomes operationally useful. Most churn does not announce itself. A grocery buyer who quietly shifts their household staples to a competitor's app while still buying fresh produce in-store looks retained in aggregate data but is already churning at the category level. Retention curves built per category expose this months before it surfaces in total revenue.

The discount rate matters too. Retail planning horizons are typically 24 to 36 months for LTV purposes, partly because the category data degrades in reliability beyond that, and partly because promotional environments shift enough to make longer forecasts speculative.

When to use it, and when the tradeoffs bite

This approach is worth the complexity when you have sufficient transaction history to calculate category-level churn reliably. A rough threshold: at least 12 months of loyalty card data for a meaningful purchase-cycle cluster, with enough customers in each cluster to produce stable churn estimates (generally 1,000 or more per segment). Below that, you are fitting noise.

It is less useful in formats with thin purchase histories, including pure-play specialists with narrow ranges. A single-category retailer selling only mattresses or bicycles does not benefit from the cluster model because there is only one cluster. In those formats, the more useful extension of LTV is accessory attachment rate, extended warranty uptake, and referral behaviour, none of which appear in the basic formula.

The honest tradeoff is model maintenance. Category-level LTV requires ongoing calibration. A structural change, such as moving a product from a monthly replenishment to a subscription format (as many personal care brands did with their DTC shifts in the early 2020s), changes the inter-purchase interval and therefore the churn model entirely. A calculation built in 2024 for a category that now has a subscription mechanic will overestimate churn and underestimate value.

AttributionAttributionA framework for assigning credit to the touchpoints that contributed to a conversion, so you can measure which channels and interactions actually drive results.View full definition → of costs also gets complicated.The reasons why acquisition costacquisition costCustomer Acquisition Cost (CAC) is the total sales and marketing spend divided by the number of new customers gained in a period. It measures how efficiently you grow.View full definition → misleads in retail are partly about how category-level promotions get allocated. A price investment in coffee to drive footfall lifts basket value across the whole shop. Assigning that promotional cost only to the coffee cluster produces a distorted category LTV even if the mechanics are otherwise sound.

Category-aware LTV does not solve every measurement problem in retail, but it does stop you from treating a quarterly furniture buyer and a daily grocery shopper as financial equivalents. Build the clusters, track the margin, and calibrate the churn rates at least every six months. The precision will earn more credibility with your CFO than any single blended number ever will.

The full course on this sector:Marketing in Retail & Distribution.