Calculating lifetime value when clients buy twice a decade

A Patek Philippe client buys a Calatrava at 34 and a perpetual calendar at 47. Thirteen years sit between the two transactions. Your CRMCRMCustomer Relationship Management: software and strategy to manage and analyse customer interactions throughout their lifecycle.View full definition → will have marked him lapsed by year three, finance will ask what he is worth against this year's budget, and both answers are wrong for the same reason: nobody decided over how many years the count runs, or what a sale in 2039 is worth in a plan you are defending in 2026. Those two choices, horizon and discount rate, determine the number more than anything the client does.

Why the textbook formula breaks over a ten-year gap

LTVLTVLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → = average order value × purchase frequency × lifespan

Three things go wrong when the interval stretches past five years.

Frequency is bimodal, not average. A base blending scarf buyers (many small purchases) with haute horlogerie clients (one purchase, then silence) produces an average frequency that describes nobody. You are running two businesses and reporting one number.

Lifespan is right-censored. With 18 months of history, every episodic client looks dead. Worse, if you estimate lifespan only from clients who did come back, you are measuring survivors: the sample is selected on the outcome you are trying to predict. Fixed-horizon modelling avoids this. Cohort-decay curves from a category benchmark set (the comparative table its own lesson supplies) beat anything you can fit on your own thin sample.

The formula has no time value at all. Forty-five thousand dollars spread over thirty years is not forty-five thousand dollars, and pretending otherwise is how long-dated relationships get overfunded.

Pick the horizon before you pick the model

Match the horizon to the decision it funds, and write it on the slide:

• 3 to 5 years for media budgets and payback arguments.
• 10 to 15 years for a boutique lease or a senior advisor hire. In France the standard commercial lease runs 3/6/9 years, so a flagship decision naturally carries a nine to twelve year window.
• 25 to 30 years only for household and generational value, reported as a separate line, never blended into the headline.

An unbounded horizon is unfalsifiable. Nobody is ever proved wrong about a purchase in 2056, which is exactly why teams reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → for it when the near-term maths disappoints.

What a sale in year twelve is actually worth

Present value of a future purchase: PV = C / (1 + r)^n.

A $70,000 watch expected in year 12 is worth $44,000 today at a 3% discount rate, $27,800 at 8%, and $22,300 at 10%. Same client, same watch, half the value depending on a rate somebody chose in a spreadsheet. Set r once, centrally, at the company's cost of capitalcost of capitalThe blended rate a company pays to finance itself through debt and equity. It sets the minimum return an investment must clear to create value.View full definition →, and refuse to reopen it per campaign. Teams discounting at zero will overinvest in relationships that pay off after they have left; teams discounting at 15% will never fund a watch client at all.

Two clients, one comparable number

Illustrative assumptions, not published figures. Hermès (which sells the accessories side of this comparison) reports gross margins around 70%, so use margin rather than revenue throughout.

*Frequent accessories client:* $2,000 average order, three purchases a year, 15-year horizon. Revenue $6,000 a year, contribution $4,200 a year. Discounted at 8%, present value ≈ $36,000 (undiscounted: $63,000).

*Episodic watch client:* $45,000 today, $70,000 in year 12, 70% margin. Contribution of $31,500 now plus $49,000 later. Present value ≈ $51,000 (undiscounted: $80,500).

The episodic client wins on both measures, which is the opposite of what a frequency-ranked list shows. But note what happened to the second watch: $49,000 of contribution collapses to about $19,500 of present value. That single number is the discipline. The far purchase is real and it counts, and it counts for roughly 40 cents on the dollar.

The household is the unit, not the buyer

Rolls-Royce delivered about 6,000 cars in 2023. At that volume, no analyst counts buyers; they count families and offices. The wife's Hermès account, the husband's car, the son's first watch and the family office that pays for all three routinely sit under four client IDs, in three systems, in two countries.

Merge them on a household key before calculating anything. Two rules follow. Count a transaction once, under the household, not under both the buyer and the beneficiary: in gifting-heavy categories the purchaser and the wearer differ often enough that double 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 → visibly inflates the base. And treat the household's non-purchase revenue as revenue. A watch client who buys twice a decade still comes in for a service every five to seven years, at several hundred to a few thousand francs, plus strap changes, restoration and resizing. Patek Philippe will service any watch it has ever made, which turns a two-transaction decade into eight or nine touchpoints and a steady, discountable cash line between the peaks.

Generational value, and the haircut it deserves

Patek Philippe has run the line "you never actually own a Patek Philippe, you merely look after it for the next generation" since the Generations campaign launched in 1996. As positioningpositioningThe mental space you want your brand to occupy in your target customer's mind relative to alternatives.View full definition → it is exact. As a metric it hides a trap: an inherited watch is a transferred asset, not a sale, and it can suppress the heir's first purchase rather than trigger it.

So model it explicitly and modestly:

Generational value = P(next generation buys) × first purchase contribution / (1 + r)^g

At 8%, the divisor over 25 years is 6.85. A $60,000 first purchase by the daughter in 2051, at 70% margin, is worth about $6,100 today. Multiply by a continuation probability you cannot yet observe, say 0.2 to 0.3, and you have $1,200 to $1,800. That justifies keeping the archive, servicing the piece for free and inviting the family to the manufacture. It does not justify a media line. Houses that let generational value carry a business case are usually using it to rescue an acquisition maths that failed on its own terms, and the acquisition-cost lesson's payback threshold is where that argument belongs. For the general discounting framework, see Harvard Business Review's overview of customer lifetime value.

Failure modes worth naming

• Discount rate shopping: a 4% rate in the boutique business case, 9% in the group model.
• Horizon inflation: stretching to 30 years to make this quarter's spend clear the bar.
• Counting the inherited object as next-generation revenue.
• Survivorship: lifespan fitted only on clients who returned.
• Second-order effect on behaviour. Whatever LTV you publish, advisors will optimise against it. Rank on undiscounted totals and they will chase the client with the biggest promised future; rank on present value alone and nobody nurtures the 40-year-old who buys at 52. Publish both columns and let them see the spread.

🎬 [VIDEO: "Customer Lifetime Value Explained" - youtube.com/@HarvardBusinessReview - a concise walkthrough of LTV mechanics applicable across high-touch and low-frequency business models]

A working model

Household LTV (present value, horizon H)
= Σ_t  (expected purchase value_t × gross margin) / (1 + r)^t
+ Σ_t  (service & aftersales_t × margin)         / (1 + r)^t
+ P(next generation buys) × (first purchase × margin) / (1 + r)^g

r = company cost of capital, set centrally
t = years to each expected purchase, capped at H
H = the life of the decision being funded, stated on the page

Report it as two figures per household: present value inside the horizon, and the generational line beneath it, unmixed. The persistent one-to-one profile its own lesson describes is what feeds t; you are not guessing intervals, you are reading them off a relationship record.

Key takeaways

• Horizon and discount rate decide the number before client behaviour does. State both, cap the horizon at the life of the decision, and set r once at the cost of capital.
• A $70,000 purchase expected in year 12 is worth roughly $28,000 today at 8%. Far revenue counts, at about 40 cents on the dollar.
• Calculate on 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 →, not revenue, and on the household key, not the buyer ID. Count gifted transactions once.
• Service, restoration and resizing turn a two-purchase decade into a continuous cash line; include it or you understate every hard luxury client.
• Generational value is real and small once discounted over 25 years and multiplied by an unobserved continuation rate. Report it separately, never inside the headline figure.
• Fit lifespan on category decay curves, not on the clients who happened to come back.