Modeling customer lifetime value for policyholders

Finance will approve a larger acquisition budget on one condition: that you can say what a policy is worth over its whole life and show the arithmetic. That number is not the premium. It is premium net of claims and servicing, repeated for as many renewals as the customer actually gives you, plus whatever second product they buy along the way. Get the duration assumption wrong by two years and the case falls apart.

Why LTVLTVLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → works differently in insurance

Retail and SaaS models usually assume a smooth churn curve and one product. Insurance breaks both, and adds a third complication those models never meet.

Lapse is lumpy. Policyholders decide at fixed intervals, normally once a year, not continuously like a subscription. An 8% rate increase at renewal produces a cliff, not gentle decay.

Cross-sell changes duration, not only revenue. Multi-policy households lapse less, so the second product pays twice. Ping An organised its retail strategy around this, reporting roughly three contracts per retail customer and arguing consistently that customers holding more contracts, and using its health and motor services, stay longer.

Third, and this is where insurance stops resembling any other LTV model: revenue is not margin, and claims decide the gap. Contribution is premium × (1 - loss ratio - expense ratio). Trupanion, the North American pet insurer, publishes this structure openly: it aims to pay roughly 70 cents of every premium dollar back out in veterinary claims, which leaves a thin band to cover variable costs, acquisition and profit. A marketing team that models on premium alone will cheerfully buy customers who destroy value.

Building the formula

LTV = Annual premium × (1 - loss ratio - expense ratio) × Expected duration × (1 + cross-sell probability × cross-sell premium ratio)

• Loss ratio: claims incurred over earned premium, for the segment you are buying, not the book average. Segment loss ratios inside a single line routinely differ by 15 points or more.
• Expense ratio: servicing, commission and admin. Acquisition marketing stays out of it; that sits on the other side of the comparison.
• Expected duration: 1 / lapse rate if the hazard is flat, which it never quite is (see below).
• Cross-sell probability and premium ratio: the chance of a second product, and its premium relative to the first.

Worked example

A home policyholder:

• Annual premium $1,400; loss ratio 62%; expense ratio 8%, so margin 30% = $420 a year
• Lapse 9% → expected duration ≈ 11.1 years
• Cross-sell probability 40%, cross-sell premium ratio 0.9 (auto at about 90% of the home premium)

1. $420 × 11.1 ≈ $4,662
2. Uplift factor: 1 + (0.40 × 0.9) = 1.36
3. LTV ≈ $6,340

Now a customer with a bigger premium from a price-comparison source: $2,200, but a 72% loss ratio and the same 8% expenses, so margin 20% = $440 a year. Lapse 35% gives 2.9 years, and no cross-sell. LTV ≈ $1,260, about a fifth of the first customer despite a 57% larger premium.

One correction before you circulate the number: eleven years of margin is not eleven years of cash today. Discounted at 8%, an eleven-year stream is worth roughly two thirds of its undiscounted total, so the defensible figure for the home customer is nearer $4,200.

Where the duration assumption breaks

The 1 / lapse shortcut assumes every renewal carries the same risk. It does not. First-renewal lapse runs well above the book average in most personal lines, so applying a mature book's blended rate to a freshly acquired cohort flatters it. If a fifth of new policies go at the first renewal and 9% a year after that, expected duration is 1 + (0.8 × 11.1) ≈ 9.9 years, not 11.1. LTV falls about 11% and payback moves out by months. Persistency curves by line are the benchmarking lesson's material; the point here is that the model needs a curve, not a constant.

A flat margin across the whole duration is the second weak assumption, and it is worst where claims rise with age: pet, health, life. Trupanion prices for this by raising premium as the animal gets older, which protects the loss ratio but raises lapse risk at precisely the point where the customer has become expensive to replace. Model margin and lapse as functions of policy year, or accept a systematic overstatement in exactly the lines with the longest tails.

The third failure mode is regulatory. In the UK, Financial Conduct Authority rules in force since January 2022 require renewal prices for home and motor to match the equivalent new-business price. An LTV model funded by walking the rate up on inert renewers is not available there, and the duration term has to be earned through service, bundling or genuine loyalty value instead.

Connecting LTV to 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 →

Set LTV against the channel-level acquisition cost the CAC lesson builds. The usual health check is 3:1 or better; below that, growth eats overhead, and well above 5:1 you are probably underspending. Our home customer at $600 acquisition cost gives about 10.6:1 undiscounted, 7:1 discounted. At $2,500 through expensive paid search or broker commission, the discounted ratio drops near 1.7:1, and the choice is to cut spend or fix lapse.

Ratios also hide the cash problem. At $420 of annual margin, a $600 acquisition cost takes 17 months to repay, which means fast growth burns cash even when every cohort is profitable and the reported loss looks alarming to anyone reading only the income statement. Trupanion frames the same decision as an internal rate of returninternal rate of returnThe Internal Rate of Return is the discount rate that makes a project's net present value equal zero. It expresses an investment's expected annualized return.View full definition → on pet acquisition spend rather than a ratio, which is the sturdier framing once duration runs past five years.

Practical levers marketing teams actually control

• Early engagement moves two terms at once. Discovery reports that highly engaged Vitality members lapse less and claim less than inactive members, which lifts duration and margin together; the early-signal proxies the engagement lesson identifies are the ones to wire into the model. Read the published uplift with care, though: Discovery licenses Vitality to partner insurers around the world, so those figures are also sales material for the programme itself.
• Cross-sell sequencing: second-product probability peaks in the first year. A bundling offer timed at month three does more work than the same offer at month twenty, and it compounds, because the bundled household then renews better.
• Discount arithmetic, done properly. A 5% multi-policy discount on the $1,400 policy costs $70 of a $420 margin, 17% of contribution. To break even it must pull lapse from 9% to about 7.5%. Treat that as a hurdle the discount has to clear, with measurement attached.
• Channel mix is a loss-ratio decision, not only a cost decision. Aggregator-sourced motor commonly carries a worse loss ratio and shorter tenure than direct or bundled business, so two channels with identical acquisition costs can produce lifetime values that differ by half.

🎬 [VIDEO: "Customer Lifetime Value Explained" - https://www.youtube.com/results?search_query=customer+lifetime+value+explained+insurance - a primer on LTV mechanics applicable across subscription and insurance-style renewal businesses, useful for building intuition before applying insurance-specific lapse and cross-sell adjustments]

Key takeaways

• LTV = premium × (1 - loss ratio - expense ratio) × expected duration × (1 + cross-sell probability × cross-sell ratio). Leaving the loss ratio out turns the model into a revenue forecast.
• A $2,200 premium with a 72% loss ratio and 35% lapse is worth about a fifth of a $1,400 premium that stays eleven years and bundles. Premium size ranks third behind duration and claims.
• Discount long streams. At 8%, eleven years of margin is worth roughly two thirds of the raw total, and that is the number to put in front of finance.
• Use a lapse curve, not a single rate: high first-renewal lapse alone can cut modelled duration by more than a year.
• Payback period matters as much as the ratio. A 17-month payback means growth consumes cash even when every cohort clears its hurdle.