Pricing Models

The insurance industry is primarily an endeavor of hedging against uncertain future loss, in which the insured trade risks with the insurers by premiums via insurance contracts. Therefore, insurance products' pricing lies at the core of any insurance business, and has its unique offerings here as well.

Most pricing models in current blockchain-based insurance communities heavily rely on the value staked on individual protocols: the higher value staked for the specific protocol, the lower premium will be priced. This staking-driven pricing structure fails to assess each protocol's real risk and is very likely to significantly over-estimate the premium of those less staked protocols. will adopt new actuary-based pricing models to substantially mitigate this issue to assess the expected loss of insurance products fairly, reduce costs, and enhance capability.

The loss assessment is conducted on the portfolio level, which will consolidate portfolio level actuarial pricing and constituents' risk scores for each protocol involved in the portfolio.

Mechanism of Portfolio-based Pricing

We will follow the Aggregate Loss Distribution model's key ideas in actuarial science to estimate the portfolio level's expected loss. The modeling workflow is illustrated in the figure below.

Work flow of Aggregate Loss Assessment

The modeling's main inputs are the number/amount of claims and number/amount of exposures in a given time window, which will be used for selecting and training two separate models - frequency model and severity model. Frequency modeling produces a model that calibrates the probability of a given number of losses occurring during a specific period, while severity modeling produces the distribution of loss amounts and sets the level of deductible and limit of the coverage amount. When both models have been well estimated, we will combine them to solve aggregate loss.

We will incorporate the decided aggregate loss into the risk factors of protocols and formulate the final premium calculations.

The models' parameters will rely on historical data to devise and validate. We carefully select, define the parameters, and constantly refine them with new data at the platform's initiation. We will adopt new Machine Learning methodologies to fine-tune and optimize the models and parameters.