Insurance carriers talk about market basket analysis constantly — in actuarial memoranda, rate filings, board presentations, and vendor pitches. Yet if you search for a clear explanation of what it actually is, how it works, and what it does (and doesn’t) tell you, you’ll find almost nothing useful.
That’s a problem, because market basket analysis is the primary way P&C carriers benchmark their competitive position. It shapes rate-level decisions, informs market entry strategy, and surfaces pricing blind spots that loss data alone cannot reveal. In our analysis of public rate filings in SERFF, over 900 filings across 55+ carriers explicitly reference market basket analysis — making it one of the most widely cited analytical techniques in insurance pricing.
This post explains what a market basket analysis is, how the methodology works, what it tells you, what it doesn’t, and how the approach is evolving.
What Is a Market Basket Analysis?
A market basket analysis in insurance is a structured comparison of premium rates across multiple carriers for the same set of risks. The carrier selects a “basket” of representative policyholders — typically thousands of synthetic or real risk profiles — and rates each one through its own algorithm and those of selected competitors. The output is a side-by-side view of how each carrier prices the same risk.
The term borrows from retail economics, where a “market basket” tracks the price of a fixed set of goods over time. In insurance, the “goods” are coverage for specific risk profiles, and the comparison reveals where a carrier is priced above, below, or in line with its competitive set.
In practice, a market basket analysis answers a simple but critical question: For the risks we want to write, are we priced to win them?
Where It Shows Up in Regulatory Filings
Market basket analysis isn’t just an internal exercise. Carriers routinely cite it in rate filings submitted to state Departments of Insurance (DOIs). It appears in actuarial memoranda, explanatory exhibits, and objection responses. A few examples from public SERFF filings:
- OnStar National Insurance Company (now GM National Insurance) referenced market basket analysis in rate filings across Illinois, Pennsylvania, Indiana, Arkansas, and Michigan as part of its multi-state PPA launch — using the analysis to calibrate initial rates in each new market.
- American Hallmark Insurance Company of Texas cited a market basket analysis in a South Carolina filing (PHXN-134489288) when entering the state, disclosing approximately 21,000 quotes generated through ITC to establish competitive positioning.
- Root Property & Casualty Insurance Company used LexisNexis InsurQuote market basket data in its Kansas and Wisconsin personal auto filings to support base rate selections.
- Midvale Indemnity Company referenced market basket results in Michigan PPA objection responses to justify rate-level adequacy to regulators.
These aren’t outliers. Across SERFF, we identified over 930 filings referencing market basket analysis, spanning personal auto, homeowners, and commercial lines.
How a Market Basket Analysis Works
The methodology has four core components: population design, competitor selection, rating execution, and output analysis.
1. Population Design
The foundation of any market basket is the risk population — the set of policyholder profiles that will be rated through each carrier’s algorithm. Population design determines whether the analysis reflects your actual book of business, your target market, or the broader market.
There are three common approaches:
| Approach | Description | Best For |
|---|---|---|
| Customer-based | Use actual submissions or in-force policies from your book | Understanding competitive position for risks you already see |
| Synthetic | Generate profiles by varying rating factors (age, territory, coverage, credit, vehicle) across a defined grid | Systematic factor-by-factor comparison; new market entry |
| Market-representative | Sample from industry data (e.g., statistical agent data, agency comparative rater quotes) | Estimating overall market position |
Population sizes typically range from 5,000 to 50,000 risk profiles, though some analyses run larger. The American Hallmark filing referenced above used approximately 21,000 quotes; OnStar’s Illinois filing used approximately 91,500 quotes over a one-year period to support a full rating plan development.
Larger populations provide more granular segmentation, but the value plateaus. For most carrier-level competitive analyses, 10,000 well-designed profiles cover the critical segments.
2. Competitor Selection
Carriers typically select 3 to 8 competitors for a market basket analysis. The selection matters more than most practitioners realize — it defines the competitive frame of reference and directly influences strategic conclusions.
Common selection criteria:
- Direct competitors — Carriers writing the same risks in the same territories
- Market share leaders — The top writers in the state/line, regardless of overlap
- Aspirational targets — Carriers whose book of business you want to capture
- New entrants — Competitors who recently entered your market
In public filings, we see competitor counts ranging from 3 (Unique Insurance in Tennessee, comparing against 3 non-standard auto carriers) to 10 (Equity/Aspen MGA in Mississippi, benchmarking against 10 competitors for GLM refinement) to OnStar’s 8-carrier comparison in Illinois.
3. Rating Execution
Each risk profile in the population is run through the rating algorithms of the selected carriers. This is the most operationally intensive step and the primary reason carriers use third-party vendors.
There are two fundamentally different ways to generate competitor rates:
Live quoting — Submitting risk profiles through carrier websites, agency comparative raters, or aggregator platforms to capture actual quoted premiums. This reflects real-world pricing including any proprietary factors, real-time adjustments, or algorithmic pricing not visible in filed rates. Vendors like EZLynx (Applied Systems), ITC (Insurance Technologies Corporation), and LexisNexis InsurQuote provide quote data from their agency and aggregator networks.
Filed rate reconstruction — Rebuilding competitor rating algorithms from public rate filings (rate manuals, factor tables, and rules filed with state DOIs). This captures the filed base rates, all published rating factors, and their interactions. Quadrant Information Services maintains a library of over 20,000 reconstructed rate sets from filed manuals. Zywave provides raw competitive datasets derived from filed rates.
Each approach has tradeoffs:
| Dimension | Live Quoting | Filed Rate Reconstruction |
|---|---|---|
| Captures proprietary/unfiled factors | Yes | No |
| Reflects real-time pricing | Yes | Lagged (filing effective dates) |
| Covers commercial lines | Limited (few comparative raters) | Yes (filed manuals are public) |
| Transparency into factor structure | Low (black box) | High (factor tables visible) |
| Consistency | Variable (quotes can change daily) | Stable (filed rates are fixed until amended) |
Most carriers use some combination of both. Personal lines carriers often rely on live quoting because comparative rater infrastructure exists. Commercial lines carriers — where comparative raters are rare — have historically had limited options, making filed rate reconstruction the primary path.
4. Output Analysis
The raw output of a market basket is a matrix: every risk profile × every carrier × the resulting premium. The analysis layer transforms this into actionable intelligence.
Overall position metrics:
- Win rate — What percentage of risks would the carrier price most competitively?
- Average premium differential — How far above or below market average?
- Position rank — Where does the carrier sit relative to competitors on average?
Segmented comparisons:
- By territory/geography — Where are you most and least competitive?
- By driver/risk profile — Which customer segments are you winning or losing?
- By coverage level — Are your minimum limits competitive but full coverage overpriced?
- By rating factor — Which specific factors (age, credit, vehicle type) are driving your relative position?
Displacement analysis:
- Premium at risk — How much premium could you lose to a competitor with a specific rate change?
- Elasticity modeling — How would a 5% rate decrease in Territory X change your win rate?
The segmented view is where most of the strategic value lives. A carrier may be competitive overall but dramatically overpriced for young drivers in urban territories — a blind spot that aggregate metrics would mask.
What Market Basket Analysis Tells You
When executed well, a market basket analysis provides:
-
Relative pricing position — Where you stand versus named competitors for specific risk segments. This is the core output and the primary input to rate-level strategy.
-
Factor-level benchmarking — How your rating factors (territory, age, credit, vehicle) compare to competitors. If your territory relativities diverge significantly from the market, that’s a signal worth investigating.
-
Market entry calibration — For carriers entering a new state, a market basket provides the competitive landscape needed to set initial rates. This is why new entrants like OnStar/GM National and American Hallmark cite market basket data extensively in their initial filings.
-
Rate change impact modeling — Before filing a rate change, carriers can model how the proposed rates would shift their competitive position. This is often the “so what” that connects actuarial analysis to business strategy.
-
Win/loss pattern diagnosis — If new business acquisition is declining in a segment, a market basket can confirm whether pricing is the cause or whether the issue lies elsewhere (underwriting appetite, distribution, service).
What It Doesn’t Tell You
Market basket analysis is a powerful tool, but it has clear boundaries. Misinterpreting what it shows is one of the most common pricing strategy errors.
It doesn’t tell you whether rates are adequate. A carrier can be the cheapest in the market and still be inadequately priced. Competitive position and profitability are separate questions. A market basket shows where you are relative to others — not whether anyone in the comparison is actually making money.
It doesn’t tell you actual market share. Pricing is one input to market share, but distribution, brand, underwriting appetite, claims experience, and service quality all play significant roles. Being the lowest price doesn’t guarantee volume.
It doesn’t capture the full pricing picture. Filed rates don’t include underwriting tier placement, schedule rating credits/debits, or consent-to-rate adjustments. Live quotes capture some of this but may not reflect the final premium a policyholder actually pays after agent/broker negotiation.
It doesn’t explain why competitors price the way they do. A competitor’s territory factor may differ from yours because they have different loss experience, different reinsurance costs, or different strategic priorities. The market basket shows the difference but not the reason.
It is a point-in-time snapshot. Competitive landscapes shift as carriers file rate changes. A market basket conducted in January may not reflect rates effective in June. Monitoring cadence matters.
The Vendor Landscape
Five established vendors dominate the market basket analysis ecosystem for P&C carriers, with newer entrants taking AI-native approaches. Each tackles the problem differently:
| Vendor | Primary Data Source | Strength | Coverage |
|---|---|---|---|
| Quadrant Information Services | Reconstructed filed rate sets (20,000+ rate set library) | Deepest filed-rate library; factor-level transparency | Broad — PPA, HO, commercial |
| Zywave | Raw competitive datasets from filed rates | Flexible data delivery for carriers with internal analytics capability | PPA, HO |
| LexisNexis InsurQuote | Aggregator live quote data | Real-world pricing including unfiled factors | Primarily personal lines |
| EZLynx (Applied Systems) | Agency comparative rater quotes | Reflects actual agent-quoted premiums | Personal lines |
| ITC (Insurance Technologies Corporation) | Comparative rater quotes | Strong in non-standard auto | Personal lines |
| Effective AI | AI-reconstructed rating algorithms from public SERFF filings | Executable raters with full source traceability; conversational follow-up analysis | PPA, HO, commercial — including lines without comparative rater infrastructure |
In public filings, the established vendors each appear across 38 to 57 unique carrier users, suggesting broad adoption. Quadrant claims virtually every major U.S. carrier as a client. Disclosed annual spend for competitive data services ranges from approximately $21,000 to $32,000 per year for large carrier groups (based on Illinois Supplier Diversity filings), though pricing varies significantly by scope, number of states, and level of service.
Effective AI takes a fundamentally different approach from the traditional vendor model. Rather than maintaining a static library of pre-built rate sets, Effective AI’s agents read filed rate manuals directly from SERFF, extract every table and rule, reconcile the full filing chain, and produce executable rating algorithms on demand. The output isn’t a dataset — it’s a working rater that can quote any risk profile, with every factor traceable to its source filing, page, and table. This means carriers can run competitive analyses against any filed program, including recently approved filings that haven’t yet been added to traditional vendor libraries, and across commercial lines where comparative rater infrastructure doesn’t exist.
The Data Problem in Commercial Lines
For personal auto and homeowners, the market basket ecosystem is mature. Agency comparative raters generate millions of quotes, and vendors maintain extensive rate set libraries. But for commercial lines — commercial auto, general liability, workers’ compensation, BOP — the picture is fundamentally different.
Commercial lines competitive analysis has been nearly impossible for three reasons:
-
No comparative rater infrastructure. Commercial risks are too heterogeneous for standardized quote comparison platforms. There is no EZLynx equivalent for commercial auto fleets.
-
Manual complexity. Commercial rate manuals can run thousands of pages, with layered classification systems, schedule rating, experience modification, and exception rules that make reconstruction labor-intensive.
-
Filing opacity. While commercial rate filings are technically public, extracting and reconstructing the complete rating algorithm from filed PDFs requires deep domain expertise and significant manual effort.
This gap has left commercial lines carriers largely flying blind on competitive positioning — relying on anecdotal broker feedback, loss of business reports, and occasional consulting engagements rather than systematic analysis.
How Public Filing Data Is Changing the Game
The competitive analysis landscape is shifting. Every rate filing submitted through SERFF is a public document. The base rates, rating factors, territory definitions, class plans, and rule sets that define a carrier’s pricing are all available in filed manuals and exhibits.
The challenge was never access — it was extraction. Filed rate manuals are PDFs, often hundreds or thousands of pages, with complex table structures, footnotes, cross-references, and amendment layers that accumulate over years of filing history. Turning that into a usable, comparable rating algorithm required armies of analysts and months of work.
AI is collapsing that timeline. At Effective AI, we’ve built agents that read filed rate manuals, identify and extract rate tables, parse classification rules, and reconstruct executable rating algorithms — with full traceability back to the source filing. This makes it possible to:
- Build competitor raters directly from public filings, including commercial lines where no comparative rater exists
- Compare factor structures at a granularity that live quoting cannot provide (because you can see the actual filed factors, not just the quoted premium)
- Monitor rate changes in near real time as new filings hit SERFF, rather than waiting for quarterly vendor updates
- Trace every number to its source — every factor, base rate, and rule maps back to a specific page in a specific filing
This doesn’t replace traditional market basket analysis — it extends it into lines of business and levels of granularity that were previously impractical.
A Simplified Example
To make this concrete, consider a carrier entering a new state for personal auto. They want to understand how their proposed rates compare to three established competitors.
They design a population of 10,000 synthetic risk profiles spanning the state’s territories, key age bands, coverage levels, and vehicle types. Each profile is rated through:
- Their own proposed rates (from the filing under development)
- Competitor A’s rates (reconstructed from the most recent filed manual)
- Competitor B’s rates (from live comparative rater quotes)
- Competitor C’s rates (from a vendor’s rate set library)
The output might show:
| Segment | Carrier Position vs. Market Avg |
|---|---|
| Young drivers, urban | +12% above market |
| Middle-aged, suburban | -3% below market |
| Senior drivers, rural | +8% above market |
| Full coverage, high-value vehicle | -1% below market |
| Minimum limits, economy vehicle | +15% above market |
This tells the carrier they are broadly competitive for their target segment (middle-aged suburban) but significantly overpriced for minimum-limits risks — which may or may not matter depending on their underwriting strategy. If they don’t want minimum-limits business, being 15% high is intentional. If they do, it’s a problem.
The factor-level view might further reveal that the carrier’s territory factors are well-aligned with the market, but their age curve is steeper than competitors — charging young drivers proportionally more and senior drivers proportionally more relative to the middle-aged base. That’s a specific, actionable finding that feeds directly into rate plan refinement.
Key Takeaways
-
Market basket analysis is the standard methodology for competitive rate benchmarking in P&C insurance. It appears in over 900 public rate filings and is used by carriers of all sizes for rate-level strategy, market entry, and regulatory support.
-
The methodology has four components: population design, competitor selection, rating execution, and output analysis. Each choice shapes what the analysis reveals. Population design and competitor selection deserve more attention than they typically receive.
-
It tells you where you’re priced relative to competitors — not whether your rates are adequate. Conflating competitive position with profitability is a common and costly mistake.
-
The vendor landscape is mature for personal lines but thin for commercial lines. Five vendors dominate the personal lines market basket ecosystem. Commercial lines carriers have historically lacked systematic competitive intelligence tools.
-
AI-powered filing extraction is opening commercial lines to competitive analysis for the first time. By reconstructing rating algorithms directly from public rate filings, carriers can now benchmark against competitors in lines where comparative raters don’t exist — with full factor-level transparency and source traceability.
Data in this post is drawn from analysis of public rate filings in SERFF. Filing references: PHXN-134489288 (American Hallmark, SC), ONST-133756928 (OnStar, IL), MISF-132875726 (Root, KS), HMSS-133064030 (Midvale, MI). Vendor market presence based on filing content analysis across 930+ filings referencing market basket analysis.
Effective AI offers a platform of specialized AI agents that reconstruct competitor rating algorithms directly from public filings — including commercial lines where no comparative rater exists. Define a population, pick your competitors, and get a factor-by-factor breakdown of where you win and where you lose. Every number traces to its source filing. See how it works →