Insurance · Policy administration
Policy administration system development for insurers and MGAs
We build policy administration systems and the parts around them: product configuration, rating engines, quote-bind-issue flows, endorsements, renewals, billing, document generation and agent portals, plus the migration of in-force policies off a legacy core. We have no published insurance case study yet; the adjacent projects this work draws on are named on this page.
Short answer
A policy administration system (PAS) holds the insurance product, rates the risk, issues the policy and manages every change until it lapses or renews. Innovation Insight builds custom PAS platforms for $80k to $200k+, migrates policies off legacy cores for $40k to $150k+, and starts either with a $5k to $15k assessment. We also extend Guidewire, Duck Creek, Socotra or similar platforms where buying is the better choice. We have no published insurance case study yet.
Reviewed by Zain Khalid Malik, CTO & Co-founder · Updated
What a policy administration system does
The PAS is the system of record for the insurance contract. It defines what can be sold, prices it, issues it, and keeps a versioned history of every change for the life of the policy. Claims, billing, reinsurance and regulatory reporting all read from it, so its data model decides how hard everything else will be. These are the modules we scope.
| Module | What it does | Design decision that matters |
|---|---|---|
| Product configuration | Coverages, limits, deductibles, eligibility rules, forms and endorsements defined per product, state and effective date | Products as versioned data, not code, so a new state filing does not need a release |
| Rating engine | Base rates, factors, tables and algorithms that price a risk; rate versions by state and date | Every premium must be reproducible later: store the rate version and every input with the quote |
| Quote, bind, issue | Submission, underwriting questions and referrals, quote options, bind with payment, policy number, issued documents | Referral rules and underwriting authority belong in configuration, with a full decision log |
| Endorsements | Mid-term changes to drivers, vehicles, locations, limits; pro-rata or short-rate premium adjustments | Out-of-sequence endorsements: a backdated change after a later one must re-rate the chain correctly |
| Renewals | Renewal offers generated ahead of expiry with re-rating, re-underwriting and notice periods per state | Notice timing differs by state and line; drive it from rules, not hard-coded days |
| Cancellation and reinstatement | Insured or carrier cancellations, non-payment flows, return premium and reinstatement | Statutory notice periods and earned premium calculations need tests per state |
| Billing | Payment plans, instalments, invoices, card and bank payments, commissions, refunds | Idempotent payments and a ledger that reconciles to the bank every day |
| Document generation | Declarations, policy forms, endorsements, certificates and notices assembled per policy version | Templates under version control, and the exact generated PDF stored, never regenerated later |
| Agent and broker portal | Quoting, book of business, commissions, document access and service requests for distribution partners | Agency hierarchy and permissions modelled early; retrofitting them is expensive |
Buy, build or extend
Established platforms such as Guidewire PolicyCenter, Duck Creek Policy, Majesco, EIS and Socotra solve problems that took their vendors years: product versioning, out-of-sequence endorsements, state forms and statutory reports. For a multi-line carrier, one of them is usually the right core. A custom PAS earns its cost when the product does not fit their model, when the team is small enough that a suite implementation would cost more than the business earns in its first years, or when the platform is itself the product, such as an embedded insurance API.
| Option | Fits when | Our usual role | Trade-off |
|---|---|---|---|
| Buy a carrier suite (Guidewire PolicyCenter, Duck Creek Policy, Majesco, EIS) | Multi-line carrier, many states, large in-force book | Integrations, portals, data migration, document and reporting work around the suite | Highest licence and implementation cost; long programmes; you configure within the vendor's model |
| Buy an API-first core (for example Socotra) | MGA or new carrier launching products quickly, with engineers on staff | Front ends, distribution portals, rating and underwriting services that call the core | Lower entry cost; still a vendor model, and the experience layer is yours to build |
| Build a custom PAS | Embedded, parametric or niche products; a single line in a few states; a platform you will sell to others | The whole system, designed around your product and filings | You own product versioning, out-of-sequence logic and statutory changes for good |
| Extend a legacy core | The core is stable but quoting, portals or documents are slow | New quote and portal layers writing to the core through an adapter | Two systems to run; the adapter must be the only way in |
Architecture we use for a custom PAS
- Policy as a versioned aggregate: every transaction (new business, endorsement, renewal, cancellation) creates a new immutable policy version with effective and transaction dates, so the policy as of any date can be rebuilt.
- Rating as a separate, versioned service: rate tables and algorithms loaded per state and effective date, with a test suite of known quotes that must keep their premiums when a new version ships.
- Product definitions as data: coverages, forms and rules held in configuration that product owners can edit in a test environment and promote, with an approval step.
- Events out, not shared tables: policy issued, endorsed and cancelled events published to claims, billing, data warehouse and reinsurance, so they never read the PAS database directly.
- Documents rendered once: generated PDFs stored with their template version and data snapshot, because the insured's copy must never change.
- Multi-tenant when the PAS serves several MGAs or programmes, with each tenant's products, rates and data isolated. Our multi-tenant SaaS development page covers the isolation options.
- ACORD standards where partners need them: ACORD XML or AL3 for agency and carrier data exchange, mapped to and from the internal model at the edge.
Migrating policies off a legacy core
Policy migration is harder than claims migration because in-force policies have a future: renewals, endorsements and billing that the new system must handle exactly as the old one would have. We migrate the way our legacy modernization service describes, adapted to insurance:
- Assessment ($5k to $15k): products, rate versions, forms, data quality and interfaces inventoried; a plan that says which books move first.
- New business first: new policies for one product are written on the new system while the old one keeps the in-force book, so the new system proves itself on low-risk volume.
- Renewal conversion: in-force policies move at renewal, re-rated on the new system, which avoids converting mid-term history that nobody needs to re-rate.
- Parallel rating: for every migrated policy, the old and new systems rate the same risk and differences are investigated before the renewal offer goes out.
- Reconciliation: policy counts, written premium and outstanding balances compared every night until the old core is retired.
The insurance legacy system modernization post goes further into mainframe and green-screen cores.
Mistakes we see in policy system projects
| Mistake | What it costs later | What we do instead |
|---|---|---|
| Rates hard-coded in application code | Every rate filing becomes a release, and old quotes cannot be reproduced when a regulator or customer asks | Rate tables and algorithms versioned as data, with a regression suite of known quotes per state |
| Overwriting the policy record on each change | No way to answer what the policy looked like on the date of loss; claims and audits suffer | Immutable policy versions with effective and transaction dates |
| Regenerating documents on demand | A template fix silently changes what the insured was sent two years ago | Generated PDFs stored once with their template version |
| Billing built as a set of invoices without a ledger | Refunds, instalment changes and cancellations stop reconciling within months | A double-entry ledger behind billing, reconciled with the bank daily |
| Agency hierarchy added after launch | Commissions, permissions and book transfers need a data migration on a live system | Agencies, sub-agencies and producers modelled in the first release |
| Copying every legacy quirk into the new system | The migration takes twice as long and keeps behaviour nobody wanted | Parity only where it affects premium, coverage or documents; the rest is listed and decided case by case |
Most of these are invisible in a demo and expensive a year after go-live, which is why we test them first. The rating regression suite and the policy-as-of-date query are the two things we ask to see working before any portal screen is designed in detail.
Security and compliance
A PAS holds personal and financial data for every insured, so the controls on the claims side apply here too. In US states that have enacted the NAIC Insurance Data Security Model Law (Model #668), licensees need a risk-based information security programme, encryption of nonpublic information, multi-factor authentication, third-party service provider oversight and notice to the commissioner within 72 hours of determining a cybersecurity event. For EU and UK insureds, GDPR adds lawful basis, retention and data subject rights. We build role-based access by agency hierarchy, field-level masking, append-only audit logs and retention rules into the data model. Rate and form filings with state regulators remain your filing team's work; the system stores which filed version every policy used.
What a policy administration system costs
| Engagement | Price | What you get | Timeline |
|---|---|---|---|
| Discovery | $3k to $8k | Product and transaction map, integration inventory, buy versus build recommendation, costed plan | 1 to 2 weeks |
| Legacy core assessment | $5k to $15k | Review of the existing PAS: products, rates, forms, data quality, interfaces, migration plan | 2 to 4 weeks |
| Policy migration and modernization | $40k to $150k+ | New business and renewal conversion onto a new core or suite, parallel rating, reconciliation, retirement of the old system | 4 to 12 months |
| Custom policy administration system | $80k to $200k+ | Product configuration, rating, quote-bind-issue, endorsements, renewals, billing, documents, agent portal, audit trail | 6 to 12 months for a first line |
Suite licences, payment processing and hosting are paid to those providers. The policy administration system cost post shows how scope moves the figure, payment details are on payment gateway integration, and pricing lists every published rate.
Our relevant work
We have no published insurance case study yet. These projects contain the hardest PAS patterns, and we would show you their code structure in a call:
- Solas: a multi-tenant operations platform where each record moves through Draft, Upcoming, In Progress and Completed states, estimates with expiry dates convert to invoices with discounts and tax, and documents are exported as PDFs and emailed to clients. Quote to bind to issue is the same lifecycle with stricter rules.
- Paint Nite: a platform rebuilt beside a legacy system with sync services keeping both aligned while features migrated in slices, and a checkout pipeline with idempotent payments and per-state tax rules. This is the migration and billing pattern a PAS replacement needs.
- TamTracker: a multi-tenant SaaS with separate customer, agency and admin front ends, agency partners with client impersonation, and Stripe Connect payouts. An agent and broker portal with hierarchy and commissions has the same shape.
Claims is the other half of the core: see claims management software development, or go back to the insurance industry page.
Sources and further reading
Next step
Tell us what you're building and get a written estimate.
A senior engineer replies within one business day. NDA on request.
Products we've shipped, and what happened next.
Case studies written from the technical documentation of each project: the stack, the scale and the outcome.
Questions we get asked a lot.
How much does a policy administration system cost?
A custom PAS for a first line of business costs $80k to $200k+. Migrating policies off a legacy core onto a new system costs $40k to $150k+, and an assessment of the existing core costs $5k to $15k. Suite licences are extra and paid to the vendor.
Should we build a PAS or buy Guidewire, Duck Creek or Socotra?
Buy if you are a multi-line carrier in many states, or an MGA whose products fit an API-first core. Build when the product does not fit their model, when the platform is your product, or when a suite implementation would cost more than the line earns early on.
How do you handle out-of-sequence endorsements?
Each transaction creates an immutable policy version with effective and transaction dates. A backdated endorsement re-applies the later versions on top of it and re-rates the chain, with tests that cover these sequences per product.
How do you migrate in-force policies from a legacy system?
New business moves first, then in-force policies convert at renewal with parallel rating on both systems. Counts, written premium and balances are reconciled every night until the old core is retired.
Have you built insurance software before?
We have no published insurance case study yet. The relevant work is Solas (lifecycle states, estimates to invoices, PDF documents, multi-tenant), Paint Nite (migration beside a legacy system, idempotent payments) and TamTracker (agency portal and payouts).
Do you support ACORD standards?
Yes, at the edges. We map ACORD XML or AL3 to and from the internal policy model for agency and carrier exchange, and keep the internal model designed around your products.