How Insurance Carriers Score Personal Injury Cases (and How Plaintiff Firms Can Counter)

Insurance carriers don't value cases the way plaintiff attorneys do. They run claims through structured scoring models — often automated, sometimes adjuster-driven — that produce a number long before any settlement conversation happens. That number anchors the negotiation.

Understanding how that scoring works is the difference between negotiating against a model you understand and negotiating against a black box.

The Colossus problem

The most well-known carrier scoring system is Colossus (Computer Sciences Corporation, used by Allstate, Aetna, and others historically). Colossus assigns severity points based on documented injuries, treatment, demographics, jurisdiction, and several dozen other factors. Different carriers use different systems — Liberty Mutual has Claim IQ, Farmers has the Injury Tracking System — but the core idea is the same: a structured model produces a settlement range, and the adjuster's authority is bounded by it.

These systems weren't designed maliciously. They produce consistency across thousands of claims and prevent rogue overpayments. But they also systematically undervalue claims with weak documentation — even when the underlying injury is real and significant.

What carrier models reward (and punish)

Reward signals — push value up:

• Specific diagnostic coding (ICD-10) — "lumbar disc herniation L4-L5 with radiculopathy" scores higher than "back pain"
• Continuous treatment — gaps over 30 days hurt; longer gaps hurt more
• Surgical intervention — single biggest multiplier in most models
• Documented permanency — disability ratings, MMI documentation
• Lost wages with W-2 backup — hard numbers from employer letters and pay stubs
• Future medical projection — life-care plans for catastrophic cases
• Plaintiff age and dependents — younger plaintiffs with dependents score higher
• Jurisdiction multiplier — Philadelphia, Manhattan, Cook County multipliers exist in some models

Punish signals — pull value down:

• Soft-tissue-only diagnoses without imaging
• Treatment gaps — flag of malingering or causation issues
• Pre-existing conditions to the same body region
• Inconsistent statements between intake form, deposition, and medical records
• Social media activity contradicting disability claims
• Letter of protection treatment only (in some states post-reform)
• Plaintiff demographics that statistically correlate with lower verdicts (controversial but documented in some models)

What plaintiff firms can do at intake

The core insight: most of what carriers reward is established or destroyed at the intake stage, not at settlement. By the time you're negotiating, the documentation either exists or doesn't.

1. Capture diagnostic specificity early. When the client describes the injury during intake, the AI receptionist (or human intake) should push for specifics: "Did the doctor mention disc herniation or just back pain?" "Did they recommend imaging?" Specific medical language gets coded into the case file from day one.

2. Map the treatment timeline. Continuous treatment is the single most controllable variable. If the client has gaps, document the reason (insurance issue, work conflict, pain tolerance) before the carrier flags it as inconsistency.

3. Flag pre-existing conditions immediately. Pretending they don't exist guarantees a worse outcome than disclosing and addressing them. The client's prior orthopedist visit for the same back two years ago needs to be in your case file before discovery, with a story.

4. Document lost wages aggressively. W-2 employees: pull pay stubs and an employer letter early. Self-employed: tax returns and a CPA letter. Generic "lost wages" claims get heavily discounted.

5. Build the demand letter against their model. Demand letters that anticipate carrier model factors — citing diagnostic specificity, treatment continuity, jurisdictional comparables — outperform demand letters that just narrate facts.

6. Use comparable verdicts as anchors. Carrier models reference jurisdictional verdict data. So should your demand. A demand letter that cites $1.8M and $2.4M verdicts from the same county for comparable cases anchors the conversation differently than a generic ask.

The asymmetry plaintiff firms can exploit

Carriers see thousands of cases through the model. Plaintiff firms see one case. That's an asymmetry that historically favored carriers — but AI is changing it.

When plaintiff firms can score their own cases against thousands of comparable verdicts (using the same kind of structured scoring carriers use), they enter negotiations with their own anchored number, not just a guess. Cases where the firm's model produces a much higher number than the carrier's first offer become obvious negotiation opportunities.

How Valryn handles it

Valryn scores every case 0–180 across five tiers using a three-layer model: global factors (liability, severity, damages), case-type factors (med-mal, auto, premises specifics), and modifier factors (carrier behavior, county tendencies, venue patterns). The third layer specifically encodes carrier signals — when your firm sees the same carrier handling cases similarly, that pattern gets baked into the scoring.

Comparable verdicts come from a database of 1,700+ indexed PI cases with county-level resolution, surfaced directly in the case file alongside the score.

See how case scoring works → County-level verdict benchmarks →

The bottom line

Carriers score cases with structured models. Plaintiff firms that don't have their own structured scoring negotiate at an information disadvantage — accepting first offers that the carrier knows are below model value. Building scoring on the plaintiff side levels the field, which is the whole point of AI case intelligence in PI practice.