Underwriting Engine at Ethos
Ankur Mishra, Director of Engineering, Underwriting Engine
Thank you for your interest in learning about the Underwriting Engine at Ethos. Before we delve any further, let’s answer the fundamental question:
“What is Underwriting?”
“Underwriting is the process through which an individual or institution takes on financial risk for a fee. This risk most typically involves loans, insurance, or investments. The term underwriter originated from the practice of having each risk-taker write their name under the total amount of risk they were willing to accept for a specified premium.” — Investopedia
Underwriting (UW) at Ethos
To further specify the above definition in the Ethos context, underwriting is a process to understand the mortality risk of the applicant and use this information to arrive at an approval and pricing decision.
We want to solve this problem with the following two tenets:
- Speed: Provide most of our users with an instant and accurate underwriting decision.
- Accessibility: Serve the broadest user base including underserved risk categories.
The Engine
Before we start diving into the nuts and bolts of the services, let’s look at the high-level view of the underwriting process. In order to make an Underwriting Decision, we need to complete the following steps:
- Data collection: This data consists of both the user-disclosed information as well as that collected through third parties. For the rest of the post, we will refer to this as evidence.
- Evidence evaluation: Transform and evaluate the evidence data to arrive at an Underwriting Decision. This phase provides intermediate results that need to be further summarized.
- Decision: Summarize the evaluations to a decision and communicate it to the user.
There are two broad categories of evidence:
- User-provided disclosures as they are applying for the insurance policy.
- Third-party information that we securely pull from data providers. The key consideration to note is that this information costs dollars and affects our cost and subsequent pricing.
The Underwriting Engine at Ethos is a decision engine which operates as a collection of several core services and components. In the next section, we will cover the high-level responsibilities of each of these services.
Application Engine — This service powers disclosures. The key entities in this service are:
- Interview: A configuration describing all possible questions that we can ask a user. User interview is a tree and we use the previous answer to dynamically determine the next question for the user.
- Questions: Nodes describing a question.
- Answers: User response.
- Journey: A collection of all the answers that the user provides.
Underwriting Engine — A service that interfaces with application funnels and drives the underwriting process for a policy. This service coordinates with other UW micro-services to arrive at a UW decision.
Evidence Integration — This service is used by Maestro to order third-party evidence. It is almost stateless and stores only the metadata for request and response. The service has pluggable integrations to securely call third-party APIs, store the raw response (and a minimally processed response to convert XML into JSON) and notify the status to Maestro.
Parsing Service — This service transforms evidence into a standard format. It provides the following functions:
- Transforms diverse evidence structure into a standard proprietary format. Evidence is converted into a canonical flattened array of objects.
- Post-processing (e.g., convert dates to month_since)
- Aggregation (e.g., convert traffic violations to number of severe violations)
- Grouping by taxonomy (helps fold the features)
Rules Engine (Janus) — A configurable system that uses our proprietary risk taxonomy to run thousands of computations to evaluate conditions. These conditions, in turn, lead to outcomes that are post-processed and summarized to an underwriting decision. This highly configurable system runs near real-time to provide our users with instant decisions.
Workbench — An underwriting workflow that allows underwriters and internal operational teams to have an overview of the underwriting process. This tool also exposes evidence data in a structured UI and allows underwriters to have access to all the information about the specific application and to use the tools to record their observations, order more follow-ups as well as take decisions.
Analytics — Anonymized data sets and models that power dashboards used by risk and actuarial science teams to get exposures to decision quality. These systems also provide insights into underwriting costs and latencies.
Authoring and Backtesting — Authoring system allows the team to release new configurations (interviews and rules) with minimal engineering support. These systems are built to provide validation, backtesting, and test automation for the changes in the configuration.
Knowledge Base — The master record of all the configurations. All the data is versioned with a snapshot ID and any changes to configurations are first deployed to KB as a draft. Once the changes are validated, there are publishing workflows that shovel changes from KB to underwriting services.
Architecture
Impact
Our UW Engine has been a driver of innovation at Ethos. Since its launch in Q3 2020, we have leveraged it to achieve several big impacts:
- Enabled consumers to check out online without needing to go through any manual workflows.
- Rapid integration with several carriers to expand Ethos offerings and provide our customers Simplified Issue, Whole Life, Elevated risk and Final Expense products — we can launch products faster than it takes for them to be filed and approved by governing bodies.
- Grew the number of automated decisions delivered to customers 133x year over year.
Next steps
- Predictive and personalized underwriting
- Decision explainer for users
- Better quotes
And many more exciting things. The UW team is always looking to hire great engineers in Ethos’ Singapore and India Engineering offices. We are hiring backend, machine learning and full stack engineers. If you’re interested in applying or want to learn more about working at Ethos, visit our careers page.
Ankur Mishra joined Ethos in March 2019 to lead Insurance Platform as Director of Software Engineering. When Ankur isn’t building and leading Platform Engineering teams, he enjoys spending time with family, reading and playing video games. Interested in joining Ankur’s team? Learn more about our career opportunities here.
