The Economics of a Premium Lock

Base cohort ultimate59%3-year payout ratio

Final-year blended49%calendar-year ratio

Payout frequency24%anchor: ~25%

Median severity$165public payouts: $219–$1,317

Unit contribution$140per policy, ultimate

P95 capital / 10kself-insured trough

Calibration. Defaults are tuned to reproduce three public stats: ~25% payout frequency (the “5× vs. home insurance” claim), payout severities in the published $219–$1,317 range, and a base case where the product is profitably priced.

FIG. 01

The Growth Mask

Fresh policies dilute mature-vintage losses.

↗ Growth masks the tail.

Blended calendar ratioCohort ultimate rangeActive policies

Acquisition years shown; the two runoff years remain in the seven-year capital path.

FIG. 02

Cohort development

Incremental payout ratio as each vintage ages.

FIG. 03

Loss-development triangle

Vintage	Year 1	Year 2	Year 3	Ultimate
C12,500 sold	44%	62%	75%	59%
C25,500 sold	44%	55%	64%	53%
C312,100 sold	43%	59%	68%	56%
C426,620 sold	43%	57%	67%	54%
C558,564 sold	43%	58%	66%	55%

FIG. 04

Positive payout distribution

24% frequency

Median $165P90 $770

RISK MEMO

Capital at risk

What the balance sheet absorbs without risk transfer.

P50 trough / 10k

P95 trough / 10k

200 replications · N₁ = 500 · normalized per 10,000 policies sold · $0 means accumulated fees cover every modeled year-end settlement

MODEL NOTES Methodology & limitations+

Premium path

Each property begins at a baseline premium dispersed uniformly from 70% to 130% of the selected baseline. Its annual renewal combines realized market trend, a constant adverse-selection drift, normally distributed property-level noise, and an occasional shock; annual declines are clamped at −10%.

Pricing rule

The forecaster is assumed unbiased on the market trend and blind to selection and shocks — that wedge is what this toy explores. The threshold compounds the initial premium by priced trend plus buffer for each of three coverage years.

Cash & development

Net fees are collected at the start of each active coverage year; payouts settle at year-end. Incremental ratio is that year's payout divided by that year's net fee. Ultimate ratio is all three years' payouts divided by all three years' net fees. Blended ratio pools the cohorts active in a calendar year.

Churn & risk

Policies may lapse before years 2 and 3. Churn truncates the worst tail, since year-3 payouts are the largest. Capital at risk is the worst cumulative cash position across seven calendar years, reported at P50 and P95 and normalized per 10,000 policies sold.

Metric definitions: Payout frequency is the share of active policy-years with payout > 0. Severity is the median and P90 of positive payouts. Unit contribution is total net fees less total payouts, divided by policies sold, per cohort on an ultimate basis.

Explicitly out of scope: Threshold resets on coverage upgrades (ACV→RC, deductible changes) · geographic segmentation · behavioral churn after a payout · fee/commission pass-through to the customer's price · reinsurance structuring · real-data upload.

v1.1 candidates: risk-transfer attach/detach layer · tornado sensitivity on cohort ultimate · yearly-varying market trend.

DILIGENCE / FIVE QUESTIONS

The model is simple. The questions are not.

What do realized payout ratios look like by cohort vintage versus priced thresholds — not blended across a growing book?
How, and how often, do payout thresholds recalibrate as realized payout data comes in?
How is the book's premium-growth drift monitored against the market-wide data the model trains on — i.e., how is adverse selection measured, not assumed?
What share of exposure is correlated (carrier- or state-level repricing), and what does the risk-transfer pipeline look like — reinsurance, capital markets, or balance sheet?
What's the fee's headroom: how price-elastic is demand relative to the expected-payout cost the model implies?