SettleCase.ai does not guess. Every demand recommendation is traceable back to a specific formula applied to your firm's own settled case history, adjusted for the specific factors of the case in front of you.
There are four models, applied in a hierarchy. The first model that has enough data to produce a reliable recommendation is used. Every output shows you which model was used and why, so you can evaluate the number with full context.
The philosophy is simple: your own data is the most accurate predictor of what your next case will settle for — more accurate than industry averages, more accurate than any AI that has never seen your cases. We build and personalize these models from your settlement history. The more you log, the more accurate they become.
Model 01
Comp Set Model
Used when: 5+ comparable settled cases in your history
The primary model. Builds a matched comparison set from your closed cases by filtering on injury type, insurer, state, special damages bracket, and liability clarity. Cases from the last 24 months are weighted 2x. From this comp set, the engine derives your personal multiplier for this case type and produces a demand range.
Formula
base_multiplier = avg(settlement ÷ specials) across comp set
low = specials × (base_multiplier × 0.90)
high = specials × (base_multiplier × 1.10)
Adjustments applied on top:
Contested liability → lo × 0.80, hi × 0.85
Shared/comparative fault → lo × 0.70, hi × 0.75
Treatment gap detected → lo × 0.85, hi × 0.90
low = specials × (base_multiplier × 0.90)
high = specials × (base_multiplier × 1.10)
Adjustments applied on top:
Contested liability → lo × 0.80, hi × 0.85
Shared/comparative fault → lo × 0.70, hi × 0.75
Treatment gap detected → lo × 0.85, hi × 0.90
Comp Set Filters
5 criteria matched
Injury type · Insurer · State · Specials bracket · Liability tier
Recency Weighting
2x for last 24 months
Recent settlements count double — market conditions change
Specials Brackets
4 tiers
Under $25K · $25K–$75K · $75K–$150K · Over $150K
Minimum Comps
n ≥ 5
Falls back to Model 02 if fewer than 5 matched cases
Example
Lumbar herniation · State Farm · TX · $68,000 specials · Clear liability
Comp set: 12 cases · Avg multiplier: 3.8x · No treatment gap
low = $68,000 × (3.8 × 0.90) = $232,560 → $232,500
high = $68,000 × (3.8 × 1.10) = $284,240 → $284,000
Comp set: 12 cases · Avg multiplier: 3.8x · No treatment gap
low = $68,000 × (3.8 × 0.90) = $232,560 → $232,500
high = $68,000 × (3.8 × 1.10) = $284,240 → $284,000
Recommendation: $232,500 – $284,000
Model 02
Broadened Comp Model
Used when: fewer than 5 injury+insurer matches, but 5+ same-injury cases exist
When the tight comp set is too small, the engine broadens the match — dropping the insurer filter and pulling all cases of the same injury type in the same state. This produces a less personalized but still data-backed range calibrated to your firm's actual outcomes for this injury class.
Formula — Same as Model 01, broader comp set
comp_set = all same injury type cases in your history (state optional)
base_multiplier = avg(settlement ÷ specials) across broadened set
low = specials × (base_multiplier × 0.90) + adjustments
high = specials × (base_multiplier × 1.10) + adjustments
base_multiplier = avg(settlement ÷ specials) across broadened set
low = specials × (base_multiplier × 0.90) + adjustments
high = specials × (base_multiplier × 1.10) + adjustments
⚠ Model 02 output is less precise than Model 01 because it crosses insurer lines. State Farm and USAA behave very differently. As your settlement history grows, more cases will qualify for the tighter Model 01 match.
Model 03
Standard Multiplier Model
Used when: fewer than 5 cases of any kind in your history
When your comp set is too small for a reliable personal multiplier, the engine falls back to industry-standard multiplier ranges by injury type. These are the same multipliers experienced PI attorneys apply manually — soft tissue at 2.5–4x, surgical cases at 5–8x, TBI at 6–10x. Adjusted for liability and treatment gaps.
Formula
multiplier_range = standard table lookup by injury type (see below)
low = specials × multiplier_range[lo] + adjustments
high = specials × multiplier_range[hi] + adjustments
low = specials × multiplier_range[lo] + adjustments
high = specials × multiplier_range[hi] + adjustments
Soft Tissue
2.5x – 4.0x
Neck/back strains, whiplash, sprains
Herniated Disc
3.5x – 5.0x
Lumbar/cervical herniation, disc injury
Fracture
4.0x – 6.5x
Broken bones, complex fractures
Surgical
5.0x – 8.5x
Any case requiring surgical intervention
TBI / Concussion
6.0x – 10.0x
Traumatic brain injury, severe concussion
Wrongful Death
8.0x – 15.0x
Fatal cases — wide range by jurisdiction
⚠ Model 03 is the industry baseline — the same mental math most PI attorneys do by feel. It is not personalized to your firm, your insurers, or your jurisdiction. Import 20+ settled cases to unlock Model 01.
Model 04
Yield Probability Score
Applied on top of Models 01–03 as a confidence layer
This is not a separate demand number — it is a second-order analysis applied to whatever range Models 01–03 produce. It answers a different question: given this demand number, what is the statistical probability it settles within 60 days at or above target yield? The score is built from 7 weighted factors and displayed as a 0–100 Demand Strength Index alongside every recommendation.
7 Scoring Factors
+ Comp set depth: n≥30 → +12pts, n≥15 → +8pts, n≥5 → +4pts
+ Settlement rate: ≥80% → +15pts, ≥65% → +8pts, ≥50% → +2pts
+ Demand alignment: in sweet spot vs. your median → +10pts
Overdemand risk: demand >2.5x median settlement → −10pts
+ Liability clarity: clear → +8pts, contested → −12pts, shared → −6pts
+ Treatment gap: none → +4pts, detected → −8pts
+ Insurer cooperation: your history with this carrier → ±8pts
+ Recency: 5+ comps in last 18mo → +5pts, all stale → −4pts
Score clamped 10–95. Label: Very Strong (≥80) · Strong (≥65) · Moderate (≥50) · Cautious (≥35) · High Risk
+ Settlement rate: ≥80% → +15pts, ≥65% → +8pts, ≥50% → +2pts
+ Demand alignment: in sweet spot vs. your median → +10pts
Overdemand risk: demand >2.5x median settlement → −10pts
+ Liability clarity: clear → +8pts, contested → −12pts, shared → −6pts
+ Treatment gap: none → +4pts, detected → −8pts
+ Insurer cooperation: your history with this carrier → ±8pts
+ Recency: 5+ comps in last 18mo → +5pts, all stale → −4pts
Score clamped 10–95. Label: Very Strong (≥80) · Strong (≥65) · Moderate (≥50) · Cautious (≥35) · High Risk
Example Output
Demand: $275,000 · Comp set: 18 cases · Settlement rate: 78% · Clear liability · No treatment gap · State Farm (72% yield with you historically)
Demand Strength Index: 74% — Strong
Model Comparison
Feature
Model 01
Model 02
Model 03
Model 04
Uses your own data
✓✓✗✓
Insurer-specific
✓✗✗✓
Jurisdiction-specific
✓✓✗✓
Injury-type specific
✓✓✓✓
Recency weighted
✓✓✗✓
Requires comp history
5+ matched5+ anyNoneAny
Outputs demand range
✓✓✓✗
Outputs probability score
✗✗✗✓
Accuracy vs. your outcomes
HighestHighBaselineN/A
Frequently Asked Questions
How is this different from what I do manually?
Most experienced PI attorneys already use multipliers — they do it from memory and spreadsheets. SettleCase.ai is the same methodology, made explicit, personalized to your actual case outcomes, and auditable. When a partner asks "why did you open at $275K?" you have a data-backed answer.
Can I override the recommendation?
Yes, always. You enter your actual demand amount and if it falls outside the recommended range, you select an override reason (local jury culture, strong liability facts, policy limits suspected, prior adjuster relationship, or other). Every override and its outcome is tracked. Over time, if your overrides consistently outperform the model, the system learns from that.
What is the Overdemand Cliff?
This is the most important insight the engine surfaces. Demanding too high does not just get rejected — it statistically extends the time to settlement even when it eventually resolves, and it lowers the final yield. For each injury type and insurer in your history, the engine identifies the demand band above which settlement rate drops significantly. It then tells you your current demand relative to that cliff.
Does the model use data from other firms?
No. Every recommendation is built exclusively from your own settled case history. Your data is never shared with other firms and no cross-firm aggregation is used in your recommendations. This is by design — your firm's specific insurer relationships, jurisdiction, and case mix are the most accurate predictors of your outcomes.
How many settled cases do I need before Model 01 activates?
The engine requires 5 matched cases (same injury type, same insurer, same state) to use Model 01. 20+ cases gives you reliable injury-type profiles. 50+ gives you insurer-specific behavioral patterns that meaningfully separate carriers like USAA from Allstate. The more you log, the sharper the recommendations.