Designing a parametric product looks deceptively simple from the outside: pick a trigger, set a payout, collect premium. In practice, getting it right across multiple perils is one of the more demanding pieces of actuarial work I've done. Let me walk you through how I approached building and pricing a multi-peril parametric cover for Southeast Asia — spanning heat, rain, storm and thunder, flood, and pollution — and the decisions that actually matter.

Step 1: Define the trigger

Everything starts with the trigger. For each peril you need an index that is objective, independently measurable, and — critically — well correlated with the actual loss the client suffers: rainfall in millimeters over a window, wind speed, a temperature threshold, an air-quality index for pollution.

The tension is always between correlation and availability. The index that best matches losses may rely on data that isn't reliable or granular enough; the index with the cleanest data may not track losses well. You resolve this by testing candidate indices against historical loss experience and picking the structure that minimizes basis risk without becoming unmeasurable. As Milliman puts it, the better the correlation between trigger and actual loss, the lower the likelihood of basis risk — which is why careful trigger selection and location-specific data are the whole game. With multiple perils you also have to decide whether triggers operate independently, in tiers, or in combination. A storm that brings both extreme wind and extreme rain shouldn't double-count or, worse, fall through a definitional gap.

Step 2: Structure the payout

Once the trigger is defined, you design the payout. The classic choices are a binary payout (trigger breached, fixed sum paid), a stepped/tiered structure (progressively larger payouts as the index climbs through bands), or a linear/continuous payout scaled to the index value.

Tiered and linear structures generally reduce basis risk because the payout tracks severity more closely, but they're more complex to communicate and price. Binary triggers are simple and transparent but blunt. For a multi-peril product I lean toward stepped payouts, with attachment points and limits calibrated to the return-period losses of each peril.

Step 3: Burn cost and AAL

Now the pricing. I anchor on two complementary views:

  • ● Burn cost: Run the trigger definition and payout structure over the full historical data set and calculate what the product would have paid, year by year. Average that, and you have the empirical expected annual payout. It's intuitive and grounded in real events.
  • ● Average Annual Loss (AAL): The modeled long-run expected cost of the risk, annualized across all possible scenarios rather than just those in the historical record. Return-period losses (the 1-in-20, 1-in-100 event) feed directly into where you set attachment points and payout limits.

Burn cost tells you what history says; AAL tells you what the full distribution says. When they diverge, that divergence is itself information — often a sign that the historical window is too short to capture tail events, which matters enormously for climate perils where the tail is getting fatter. The financial translation has to be auditable: an insurer setting a premium needs to see the whole chain from physical threshold to damage percentage to monetary loss.

Step 4: Stress testing and loadings

Historical burn cost is necessary but not sufficient, because climate risk is non-stationary — the past systematically understates the future. So I build Python-based prototypes to test product performance under both historical and stressed climate scenarios. Shift the rainfall distribution, increase storm frequency, warm the temperature baseline, and watch what happens to projected payouts.

This is what supports a genuine go-to-market decision. It tells you whether the product is solvent under plausible adverse futures, not just in an average year — a live concern when WTW's 2026 Natural Catastrophe Review found that flood-related losses in Southeast Asia may grow as much as tenfold in the coming years. On top of the pure risk cost, you then layer loadings — for expenses, for the cost of capital, for uncertainty and basis risk, and for profit — to arrive at a commercial premium. The heavier the tail and the thinner the data, the larger the uncertainty loading needs to be.

Step 5: Document it properly

The final, underrated step is drafting the policy documents and payout terms and conditions. In parametric insurance the wording is the product — the contract defines the exact index source, the measurement methodology, the trigger thresholds, and the settlement mechanism. Ambiguity here is where disputes are born. Getting the T&Cs precise and unambiguous is as important as getting the price right.

The through-line

Every one of these steps circles back to basis risk. Trigger choice, payout shape, data granularity, stress testing — they're all levers to make the payout track the real loss more closely. Research on weather parametric portfolios shows that diversifying across independent contracts materially reduces aggregate basis risk, which is another reason a well-constructed multi-peril book can be more robust than the sum of its parts. A parametric product that pays reliably when the client hurts earns renewals; one that misses builds resentment and churns.

That end-to-end view — from trigger design through pricing, stress testing, and contract wording — is what I bring to clients building parametric programs. If you're designing a multi-peril product and want a practitioner's hands on the pricing and structure, let's talk.