Insurance & fintech

Building insurance platforms in Southeast Asia: what to ask a dev partner

Insurance software punishes sloppy engineering. The seven questions that separate a dev partner who can build an insurance platform from one who will learn on your money.

Marco Pravato Published 4 min read

Insurance platforms punish sloppy engineering harder than almost any other software: premiums, taxes, commissions and claims must be correct to the smallest unit, reproducible months later, and auditable through every state change - while the customer experience competes with the slickest apps in the region. Choosing a development partner therefore comes down to seven questions about engineering discipline - calculation testing, rule versioning, audit trails, policy state, integrations, data protection, and references where correctness mattered. Fluent answers are cheap; ask for artifacts. Here is each question and what a real answer sounds like.

Why is insurance software unforgiving?

Because every defect is either money or trust. A rounding error in premium calculation is wrong revenue at scale; a broken policy state transition strands a customer at claim time - the exact moment trust is the product; a gap in the audit trail becomes a regulatory finding. Meanwhile the front of the platform must quote in seconds and sell on a phone, because that is the market standard set by the region’s digital insurers.

The tension is the point: consumer-grade experience on top of accounting-grade correctness. Teams fluent in one half routinely underestimate the other.

The seven questions, and what good answers contain

1. “How do you test calculations?” Good: a dedicated test suite for the money math - hundreds of input combinations per product, boundary cases, rounding rules, run on every build. Bad: “our QA tests the application” - screen testing samples a few paths; calculation testing enumerates them.

2. “How do you version rating rules?” Good: rule sets with effective dates; every quote and policy records the version that priced it; past calculations reproducible forever. Disqualifying: rates hard-coded in application logic, “updated on deploy”.

3. “How do you design audit trails?” Good: every state change and every human override recorded with who/when/what-changed, designed in from the schema up. Bad: “we have logs” - server logs are not an audit trail an auditor will accept.

4. “Show me how you model policy state.” Endorsements, mid-term adjustments, cancellations, reinstatements, renewals: the state machine is where insurance systems rot. A partner who asks you sharp questions here - what happens to commission on a backdated cancellation? - is demonstrating the domain scar tissue you are hiring.

5. “What have you integrated?” The platform lives in a mesh: payment gateways and installments, identity verification, LINE/WhatsApp messaging, legacy core systems, reinsurance and regulatory reporting. Ask for integration war stories - especially the reconciliation and silent-failure handling, because in insurance a silently dropped webhook is unbilled premium.

6. “What is your data-protection posture?” Health declarations, financial details, claims histories - sensitive data under Thailand’s PDPA and, for some products, stricter rules. Good answers talk about data minimization, access design and processing terms in the contract, unprompted.

7. “Where has correctness mattered in your references?” Not “have you built insurance?” - the honest pool is small - but where did your engineering discipline face money-math and audits? Billing systems, financial reporting, regulated-sector work all transfer. Then check the references properly.

How does this shape the engagement?

DecisionRecommendationWhy
Engagement modelDedicated team for the platform; fixed price only for bounded piecesInsurance products evolve continuously; the domain knowledge must accumulate, not reset
First stepPaid discovery: product rules, integrations map, architectureThe rating rules and state machine must be specified before anyone prices a build
Team shapeSenior-heavy core, testers from day oneCorrectness is designed in; it cannot be QA’d in later
Location economicsOffshore works - with the discipline verified firstCorrectness work is labor-intensive; Bangkok rates buy the testing depth onshore budgets cut. Verify via pilot

In short

  • Insurance software = consumer-grade UX on accounting-grade correctness - partners fluent in both are rare, so interview for the second half.
  • Seven questions expose the discipline: calculation tests, versioned rules, audit trails, policy state, integrations, data protection, correctness references.
  • Demand artifacts, not fluency: a test suite sample, a versioning design, an audit schema.
  • Build with a senior dedicated team after a paid discovery - and let offshore economics fund the testing depth the domain demands.

AD System Asia builds for the insurance sector from Bangkok - our clients include Roojai, one of Thailand’s best-known digital insurance players. Bring us your product and your seven-questions interview; we enjoy that conversation.

A few more answers

Frequently asked questions

What makes insurance software different from other business software?

Money math and auditability. Premiums, taxes, commissions and claims must be correct to the smallest unit, reproducible months later, and traceable through every state change - under regulation. The engineering disciplines this demands (testing depth, audit trails, versioned calculation rules) are exactly what general web-dev teams cut first.

What should an insurer ask a potential software vendor?

Seven things: how they test calculations, how they version rating rules, how they design audit trails, how they handle policy state (endorsements, cancellations, reinstatements), their integration experience with payment and core systems, their data-protection posture, and references from work where correctness mattered.

Can development for an insurance platform be outsourced offshore?

Yes - several of the region's insurance players build offshore. What matters is the partner's engineering discipline, not geography: calculation test suites, documented releases, audit-ready traceability. Offshore rates actually help, because correctness work is labor-intensive and onshore budgets routinely cut it.

How do rating and pricing rules stay maintainable in an insurance system?

By versioning them like contracts: every rule set carries effective dates, every quote and policy records which version priced it, and past calculations remain reproducible forever. A vendor who proposes hard-coding rates into application code is disqualifying themselves.

What integrations does an insurance platform in Southeast Asia typically need?

Payment gateways and installment providers, identity verification, messaging channels customers actually use (LINE, WhatsApp), core or legacy policy systems, reinsurance and regulatory reporting, and increasingly embedded-insurance APIs for partner distribution - each one a place where silent failure costs real money.