Calculating lifetime value when purchase cycles vary by category

An Ocado customer places an order every week or two. An IKEA sofa buyer comes back for the next sofa somewhere around year seven. A Nespresso owner buys the machine once, then buys capsules two or three times a week for as long as the machine stays on the counter. The gap between the fastest and slowest repurchase interval across those three is a factor of several hundred.

Feed all three into one lifetime valuelifetime valueLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → formula with a 12-month lookback and the outputs are not so much wrong as meaningless: the grocery shopper looks like your best asset, the furniture buyer looks like a one-off, and the capsule buyer barely registers because the machine sale was the only transaction inside the window.

Where the formula bends

LTV = AOV × purchase frequency × margin % × customer lifespan

Four inputs. The category decides which one carries the answer, and which one you can actually move.

• Online grocery: basket above £100 (Ocado Retail's average basket has run comfortably north of that, roughly triple a typical in-store trip), frequency measured in dozens of orders a year, and margin that collapses once picking and delivery are charged against it. Use a contribution margin after fulfilment here, not a headline 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 →, or you will overstate LTV by a factor of five.
• Big-ticket furniture: order values in the hundreds or thousands, frequency below 0.2 a year, margin around 40 percent, and a "lifespan" that is really a bet on two purchase occasions in fourteen years.
• Consumables: a near-zero-margin hardware sale followed by a high-margin refill stream that behaves like a subscription nobody signed.

Get the frequency assumption wrong in the first case and you are out by 20 percent. Get it wrong in the third and you are out by an order of magnitude, because frequency is the whole business.

Worked example 1: Ocado and the weekly cycle

Modelled assumptions, not company disclosures:

• AOV: £115
• Frequency: 30 orders a year
• Contribution margin after picking and delivery: 5%
• Lifespan: 5 years

LTV = £115 × 30 × 0.05 × 5 = £862

Set that against CACCACCustomer 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 → measured incrementally rather than blended, the way the acquisition-cost lesson insists. At £60 to win a customer through a first-order discount plus media, the ratio is about 14:1. Comfortably past the 3:1 cross-sector rule of thumb, with the denominator caveats the benchmarking lesson raises.

The sensitivity that matters: at 5 percent contribution, one extra order per quarter adds about £29 of lifetime profit. One extra percentage point of contribution margin adds £172. In online grocery the margin lever is worth roughly six times the frequency lever, which is the reverse of the received wisdom borrowed from bricks-and-mortar loyalty schemes.

Worked example 2: IKEA and the decade cycle

Model IKEA as a furniture retailer and you get: €900 occasion, twice in fourteen years, 40 percent margin. LTV = €720.

That number is wrong, and the reason is visible in IKEA's own arithmetic. Divide Ingka Group's annual retail revenue by store visits and you land in the tens of euros per visit, not hundreds. Most IKEA visits are not sofa visits. They are storage boxes, tealights, a Billy shelf, lunch.

Add a modest accessory stream of three visits a year at €45 and 40 percent margin, and you add €54 a year, €756 over the same fourteen years. The stream you ignored is larger than the sofa cycle you modelled.

This is the failure mode for every big-ticket category with a long tail of small products: mattresses, appliances, DIY. The headline purchase anchors the analysis, the low-value repeat traffic funds it. It is also why IKEA Family, with well past 100 million members worldwide, is built to capture the €45 visits rather than the €900 ones. A retailer with no accessory tail (a standalone kitchen fitter, say) has no such rescue and has to build the relationship artificially through financing, warranties or design services.

Worked example 3: Nespresso and the annuity in between

The machine sells at close to cost, made by third parties such as Krups and DeLonghi. The capsules carry the profit.

• Capsules: roughly 2.5 a day, near 900 a year, around €0.45 each
• Gross margin on capsules: call it 60%
• Relationship: 5 years

LTV ≈ 900 × €0.45 × 0.60 × 5 = €1,215

Monthly contribution is about €20, so a €80 acquisition cost pays back inside four months, faster than the grocery example despite a fraction of the annual revenue. That is the shape of a consumables model: slow-looking category, fast payback.

The second-order consequence is where this model breaks. When the key capsule patents lapsed in the early 2010s, supermarket own-labels and rivals put compatible capsules on the shelf at a large discount. Nothing changes in the transactional data at the moment a customer switches. The machine is still there, the account still exists, and the €1,215 quietly becomes €300. Vertuo, launched later, uses a barcode-read capsule that is harder to copy, which is a defence of the LTV model as much as a product decision.

Model check: if 15 percent of your installed base defects to a substitute you cannot see, your portfolio LTV is out by roughly the same 15 percent, and it will not appear in acquisition reporting for years.

What counts as a repeat purchase

For Ocado, repeat behaviour is dense enough that the cohort curve reads cleanly, in the way the retention lesson describes. For IKEA, the same window tells you almost nothing: a customer silent for four years may be perfectly loyal.

Long-cycle categories need proxies, and they are weaker instruments than transactional data:

• attach rate on the original order (did the sofa buyer take the rug, the delivery, the assembly?)
• referral, since the customer will recommend long before they return
• brand recallbrand recallThe degree to which your target audience recognises or recalls your brand, either prompted or unprompted. It measures how present your brand is in people's minds.View full definition → at the next purchase occasion, which comes from tracking surveys, not from your database

Two known distortions. Attach rate is measured at the moment of highest intent, so it flatters. Referral survey data overstates by a wide margin, because stated intent to recommend is not recommendation. Halve what respondents claim before you put it in a model. The Baymard Institute publishes applied research on high-consideration purchase behaviour if you want to go further on big-ticket e-commerce.

Sanity-check with payback, not with the ratio

Payback (months) = CAC / monthly contribution per customer

Ocado model:    (£115 × 30/12) × 0.05 = £14.38/mo  →  £60 / £14.38 ≈ 4 months
Nespresso model: (900 × €0.45 /12) × 0.60 = €20.25/mo  →  €80 / €20.25 ≈ 4 months
IKEA, sofa only: €720 / 168 months = €4.29/mo  →  €150 / €4.29 ≈ 35 months
IKEA, with accessories: €1,476 / 168 = €8.79/mo  →  €150 / €8.79 ≈ 17 months

Two categories with nothing in common share a four-month payback. The furniture model, on the same €150 CAC, either pays back inside two years or takes three, depending entirely on whether you counted the tealights. That single modelling choice decides whether the acquisition budget clears the CFO's hurdle ratehurdle rateThe blended return a company must earn to satisfy everyone who funds it (lenders and shareholders). It sets the bar every investment has to clear.View full definition →.

Rebuilding the assumptions

LTV models rot at the speed of their own cycle. Grocery frequency moved materially through the 2022 to 2023 inflation period as shoppers consolidated trips, then partly recovered. Big-ticket furniture tracks housing turnover, which slowed across the US and much of Europe from 2023 as mortgage rates rose (see the Federal Reserve Bank of St. Louis (FRED) series for the US market). Consumables shift when a substitute appears on a supermarket shelf, which can happen in a single quarter.

Practical rule: rebuild high-frequency models annually, rebuild low-frequency models after any macro shift rather than on a calendar, and rebuild consumables models whenever a competitor changes price or compatibility. A furniture LTV built on 2019 turnover assumptions has no use in 2026.

Key takeaways

• The formula is stable; the dominant variable is not. Frequency and contribution margin decide online grocery, the accessory tail decides big-ticket, refill margin decides consumables.
• Similar LTV hides very different payback. Ocado and Nespresso models both pay back in about four months from completely different revenue shapes; a sofa-only IKEA model takes 35.
• Model the small repeat purchases around a big-ticket relationship. Leaving them out halved LTV in the IKEA example and doubled the payback period.
• In consumables, defection to a compatible substitute leaves the transactional record intact while the annuity disappears. Watch refill volume per active machine, not account counts.
• Use contribution margin after fulfilment for online grocery. Headline gross margin overstates lifetime profit by several times.