All Posts

Technology10 min read

From Spreadsheets to a Business Rules Engine: How Insurers Move Rating and Underwriting Logic Out of Excel

From Spreadsheets to a Business Rules Engine: How Insurers Move Rating and Underwriting Logic Out of Excel

Key Takeaways

  • Spreadsheets are a strong tool for developing rates and analyzing portfolios. They are a weak tool for running production decisions.

  • The risk starts when a spreadsheet becomes the source that someone re-keys into the policy administration system, or the system itself.

  • Field audits of operational spreadsheets found serious errors in most of them, and rating errors are expensive: one Texas consent order in 2025 required more than $18 million in restitution to over 80,000 auto policyholders.

  • The answer is not to take Excel away from actuaries. Keep rate development in Excel, and move production rating tables and underwriting rules into a rules engine, imported through CSV, tested and versioned.

  • A business rules engine gives every rating and underwriting decision an owner, a version, a test record and an audit trail.

Why Do Insurers Still Run Decisions in Spreadsheets?

Because spreadsheets work well for the people who design the rules. Actuaries build rating tables, relativities and what-if scenarios in Excel. Underwriters keep appetite grids and referral criteria in workbooks. Product teams draft new coverages in them. Excel is flexible, fast to change and familiar to everyone.

The trouble is what happens next. In many mid-market carriers, the spreadsheet becomes the specification that IT re-keys into the policy administration system, or it becomes the production tool itself: an underwriter opens the workbook, enters risk details and reads the result. At that point, a design tool is doing the job of a decision system, without the controls a decision system needs.

What Goes Wrong When Spreadsheets Run Production Decisions?

Errors are common and hard to find. In field audits of operational spreadsheets summarized by Raymond Panko of the University of Hawaii, the five most rigorous studies found important errors in 91% of the 55 spreadsheets examined, and in inspection experiments people caught only about 60% of errors (Panko, field audits).

Re-keying creates a second source of truth. When rating tables are copied from a workbook into a system, the two drift apart. Nobody can easily prove that production matches the filing.

Version control depends on file names. “Rates_final_v3_JS_approved.xlsx” is not an audit trail.

Testing is manual. Checking a change against hundreds of sample quotes means copying formulas and comparing by eye.

Knowledge sits with individuals. When the person who built the workbook leaves, the logic is hard to maintain.

There is no decision record. For a quote issued six months ago, you cannot easily show which table version produced it.

The cost of rating errors is concrete, wherever the error comes from. In February 2025, the Texas Department of Insurance issued a consent order after a personal auto carrier’s rating system assigned a premium consequence to zero-dollar claims. More than 80,000 policies were affected, and policyholders were to receive more than $18 million in restitution, including interest (TDI Consent Order 2025-9142, February 2025). That error was in a system, not a spreadsheet, but the lesson is the same: rating logic has to match the filing, and changes have to be tested before they reach policyholders.

Spreadsheets vs. A Business Rules Engine: How Do They Compare?

  Spreadsheet Business rules engine
Best use Rate development, analysis, what-if scenarios Production rating, eligibility, underwriting and claims rules
Who changes logic Anyone with the file Named owners, with approval rights by business area
Version history File copies and naming conventions Versioned rules with rollback
Testing Manual comparison Run changes against sample or historical cases before publishing
Audit trail Limited; depends on discipline Who changed what, when, and which version produced each decision
Integration Export, copy or re-key API calls from the policy, rating and claims systems
Consistency Each copy can differ One version serves every channel
Performance at scale Manual or batch Real-time decisions for quoting and claims

The table is not an argument against Excel. It is an argument for using each tool for the job it does well.

What Should Stay in Excel, and What Should Move?

Keep in Excel:

  • Rate development and indication analysis.
  • Scenario modeling and impact analysis before a filing.
  • Ad hoc portfolio analysis and reporting.
  • Drafting new rating structures before they are final.

Move into a rules engine:

  • Production rating tables and factors that apply to live quotes.
  • Eligibility and appetite rules by state, class and coverage.
  • Underwriting referral criteria and authority limits.
  • Validation rules for submissions and quotes.
  • Claims triage and straight-through processing rules.

The dividing line is simple: if the logic affects a live policy or claim, it belongs where it can be versioned, tested, approved and logged.

How Do You Move Rating Tables and Rules Out of Spreadsheets?

  1. Inventory the spreadsheets that drive production. List every workbook that someone uses to quote, underwrite, refer or settle, and every table that IT re-keys into a system. Note owner, frequency of change and which decisions depend on it.
  2. Separate analysis from production logic. In each workbook, mark which tabs are working calculations and which hold the final tables or rules.
  3. Design the target tables. One decision table per factor or rule set, with clear inputs and outputs. Workbook layouts rarely map one to one to well-structured tables, so plan the structure rather than copying cells.
  4. Import through CSV. Export the final tables from Excel to CSV and import them into the rules engine. Keep the CSV as the hand-off format between actuarial work and production.
  5. Test against known results. Run sample quotes, filed rating examples and recent production cases, and compare results with the current process. Explain every difference.
  6. Set owners and approvals. Each table gets an owner who can propose changes and an approver who publishes them.
  7. Retire the production workbooks. Once the engine is the source of truth, the workbook goes back to being an analysis tool.

What Does a Migrated Rating Table Look Like?

A typical actuarial workbook holds a territory factor grid like this, often with formulas, notes and helper columns around it:

State Territory Coverage Factor
TX 01 Bodily injury 1.12
TX 01 Collision 1.05
TX 02 Bodily injury 0.97
TX 02 Collision 0.94
… … … …

Values are illustrative. In a rules engine, the same data becomes a decision table with defined inputs (state, territory, coverage) and one output (factor). The formula that combines factors into a premium becomes a separate, versioned function. When the next filing is approved, the actuary exports the new grid to CSV, the team imports it as a new version, runs the test cases, and publishes on the effective date.

What Changes for Actuaries, Underwriters and IT?

Actuaries keep Excel for analysis and rate development. They stop sending workbooks for re-keying and instead publish tested table versions, with a record of what changed.

Underwriters stop maintaining personal copies of appetite grids. Rules apply the same way to every submission, and referrals reach the right authority level automatically.

IT stops re-keying tables and writing rate changes into application code. This removes backlog from the IT queue and gives IT more time for architecture, integrations and data quality.

What Should You Look for in a Rules Engine for This Move?

Spreadsheet-friendly import. Can actuaries move tables from Excel through a standard format such as CSV, without developer help?

Business-readable tables. Can underwriters and actuaries read and review the rules in production form?

Testing on real cases. Can you run a new table version against hundreds or thousands of quotes before publishing?

Versioning and effective dates. Can you see which version applied to any policy, and roll back if needed?

Permissions. Can drafting and approval be split by line of business or domain?

Performance. Can the engine rate in real time for agent and direct quoting?

Integration. Can your policy administration and rating systems call it through an API?

How Does Higson Support the Move from Spreadsheets?

Higson is a business rules engine built by Decerto for P&C insurers and financial services companies. It fits the hand-off between actuarial and underwriting work and production systems.

  • Decision tables and functions. Rating tables, eligibility rules, appetite grids and referral criteria live in decision tables in Higson Studio; formulas that combine factors are written as functions in Groovy or Python.

  • Excel → CSV → Higson. Higson’s native formats are XML and CSV. Tables prepared in Excel are exported to CSV and imported into Higson, where they become versioned decision tables.

  • Testing before production. The tester and mass tester run a new table version against sample or historical quotes, so actuaries and underwriters see what changes before it goes live.

  • Versioning, audit trail and granular permissions. Every change is versioned with rollback. Higson 4.3 LTS adds role-based permissions at the level of business domains, sessions and operations, with a full audit trail.

  • Visibility. The Diagnostic Panel shows flamegraphs of rule calls, heatmaps of decision table rows and P50/P95/P99 percentiles, so teams see which rows actually drive results.

  • Real-time performance. Rule execution takes 0.23 ms per decision, with throughput of 9,000 requests/second.

  • AI Assistant (beta). The AI Assistant in Higson Studio drafts functions and decision tables from natural-language descriptions.

Carriers use Higson exactly for this. At TUW, actuaries publish a new motor rate themselves, in 2 hours instead of 2 weeks. Allianz Poland uses Higson as a central product configuration layer and launches product variants in days instead of months.

How Do You Start?

  1. Pick one line and one table set. Personal auto rating factors or a small commercial appetite grid are common first choices.
  2. Agree on the hand-off. Actuaries keep the workbook; CSV becomes the format that moves final tables into production.
  3. Build the test set from filed rating examples and recent quotes.
  4. Run in parallel for one rate change cycle, then switch.
  5. Plan the timeline realistically. Platform go-live with Higson typically takes 3–6 months. After that, each new product takes weeks, and a single rule or table change takes hours.

See Your Rating Tables in Higson

Bring one workbook, such as a territory or vehicle factor grid, and we will show you how it would look in Higson, tested on your own sample quotes.

Book a 30-minute call

Download the Business Rules Engine Comparison Guide 2026

Take Full Control of Your Product Logic

We provide fee Proof Of Concept, so you can see how Higson can work with your individual business logic.