Customer lifetime value: fix it by purchase cycle

September 13, 2026 · 5 min · Free, with full transcript

Can a single lifetime value formula cover groceries, mattresses and seasonal clothing? The answer here is no. Averaging across categories with different purchase rhythms makes a mattress buyer look churned by month three, while a grocery shopper who slips from weekly to fortnightly disappears into the blended number. The result is win-back discounts sent to customers who were never leaving. You come away able to tag products by natural repurchase interval, set churn windows per category instead of a fixed 90 days, and treat mixed-basket customers as a portfolio of category relationships, an approach covered by MIT Sloan Management Review. Also covered: the Bain 5 percent retention claim and how to sanity-check SEMrush waste figures against your own campaign logs.

Listen to the podcast

4 min

Chapters

Key takeaways

  • Tag every product line in your catalog with its natural repurchase interval in days, months or years.
  • Define churn relative to each category's cycle rather than a fixed 90-day window.
  • For mixed-basket customers, compute a value stream per category and sum them instead of blending one LTV.
  • Group products into three buckets, fast, seasonal and rare, before reaching for a machine learning model.
  • Suppress win-back discounts for seasonal categories so you do not train buyers to wait for a coupon.
Read the full transcript

Host:Welcome to the Leaders Insights Podcast. Today's episode, calculating customer lifetime value when purchase cycles vary by category. A furniture retailer told me last month they'd been valuing their customers using the same lifetime value formula as their online grocery arm. Someone who buys a sofa every eight years scored like someone buying milk every Tuesday. How badly does that break?

Expert:Catastrophically and quietly, which is the worst combination. Lifetime value, or the total profit you expect from a customer across the whole relationship, collapses the second you average across categories with different rhythms. That mattress buyer looks like a churned debt account by month three because they haven't come back. They haven't come back because nobody buys two mattresses in a quarter.

Host:So the number isn't just wrong, it points you the wrong direction.

Expert:Exactly the direction that costs you money. You look at the mattress customer, see no repeat purchase, and you flood them with win-back discounts. You're bribing someone who was never leaving. Meanwhile, your grocery customer, who's genuinely slipping, buying every two weeks instead of weekly, gets no attention because the blended average smooths their decline into invisibility.

Host:Walk me through what category-aware actually means in practice, because it sounds like an excuse to build 12 spreadsheets.

Expert:It starts with one honest question per category. What's the natural purchase cycle? Grocery, days? Seasonal clothing, twice a year, spring and autumn? Mattresses, once a decade if you're lucky. You define churn, a customer walking away, relative to that cycle, not a fixed 90-day window. A grocery shopper silent for a month is a fire alarm. A mattress buyer silent for a year is Tuesday.

Host:And the retailers with mixed baskets, someone buying groceries in a coat in a mattress from the same brand.

Expert:That's where it gets interesting and where most models cheat. You don't compute one LTV for that person. You compute a value stream per category and sum them each with its own cycle and its own churn definition. MIT Sloan Management Review has been decent on this, treating the customer as a portfolio of category relationships rather than a single lump. The coat relationship can be healthy while the grocery one is dying. A blended number hides that entirely.

Host:Give me a figure. What is getting this wrong actually cost?

Expert:The retention economics are brutal. The oft quoted line is that lifting retention 5% raises profit by 25 to 95%, that traces back to Bain research from the 90s, still cited because it's roughly held up. On tooling, SEMrush pegs a chunk of retail marketing waste at misfired re-engagement campaigns. But flag it. They sell analytics software, so they're incentivized to make your targeting look broken. Cross check against your own campaign logs before you believe the scale.

Host:Let's be uncomfortable. Is this just complexity for its own sake? A small retailer with three product lines. Do they need this?

Expert:If those three lines have different cycles, yes. And it's not hard. You're not building a machine learning model. You're grouping products by how often people buy them. Three buckets. Fast, seasonal, rare. The failure isn't insufficient math. It's applying one average where three would do.

Host:And the small retailer feels this harder. They can't afford to discount a customer who is going to buy anyway. Where does the standard approach break most invisibly? The thing nobody catches.

Expert:The seasonal category, because it fakes churn twice a year. Your swimwear customer disappears every October. The naive model marks them lost, triggers a discount, and you've just trained a loyal seasonal buyer to expect a coupon before every summer.

Host:You taught them to wait.

Expert:That's revenue you set fire to, trying to save a customer who was coming back regardless of what you did.

Host:So what's the one thing a listener does Monday morning?

Expert:Pull your product catalog and tag every line with its natural repurchase interval. Days, months, years. Just that column. Then look at your churn rules and see how many categories you're judging against a single window. I'd bet it's most of them. That one column tells you exactly where your lifetime value numbers are lying to you, and you can fix it before you send another wasted discount.

Host:The lie in the average. Good place to end. Sources for today's episode. SEMrush vendor. SEO analytics tools. DigiDay MIT Sloan Management Review. That's your five. The full CMO track, structured no fluff, is at MBA-training.com.

Read and go deeper