Claims or Outcome Experience Rating¶

Feedback pricing method — instantiates Adverse Selection Filtering

Feeds each participant's own realized losses back into their price and the pool's base rates, so a pool quietly drifting toward bad risks gets caught and re-rated instead of silently subsidized.

Every filter in this archetype acts before exposure, on evidence that can be gamed or that simply isn't there yet. Claims or Outcome Experience Rating is the one that acts after: it watches what actually happened — losses, claims, defects, failures — and feeds that realized experience back into two things, the participant's own future price and the class base rates everyone is judged against. Its distinctive move is treating the pool's own history as the most trustworthy signal of hidden type, because outcomes can't be coached the way an application can. Where an upfront screen guesses who will cost the pool, experience rating measures it and closes the loop, so a pool that is quietly filling with bad risks is surfaced as a moving loss ratio rather than discovered at insolvency.

Example¶

A workers' compensation insurer covers hundreds of small employers in one class. Upfront underwriting can only see so much; two roofing firms look identical on paper but injure workers at very different rates. Experience rating is how the pool learns the difference. Each employer carries an experience modification factor — an industry-real multiplier built from its own recent claim history relative to the class average — so a firm running well below its expected losses earns a credit and one running above pays a surcharge. The insurer also watches the class in aggregate: if the whole roofing book's loss ratio creeps up year over year, that is the pool telling it the base rate itself is stale and the upfront filter is letting worse risks in than the pricing assumes.

The output isn't a single verdict but a continuously updated calibration: individual mods that reward the genuinely safe, and a class base rate that gets pulled back toward reality every cycle. That is what stops a slow, invisible drift from becoming a solvency problem.

How it works¶

Blend individual experience with the class base rate. A single participant's record is noisy, so it is combined with the class average under a credibility weight — more history earns more weight — rather than pricing each participant purely on its own thin sample.
Track the pool, not just the person. Aggregate loss ratios, exit patterns, and claim frequency are monitored at the class level to detect drift that no individual record would reveal.
Recalibrate the base rates. When realized experience diverges from the assumed base rate, the base rate is updated, and that corrected rate is what the pricing and underwriting mechanisms then build on.
Re-rate on a cadence. Prices and credits move as new experience lands, so the loop keeps closing rather than freezing at the first estimate.

Tuning parameters¶

Credibility weight — how much a participant's own record counts versus the class average. More weight rewards true type faster but overreacts to a run of luck; less weight is stable but slow.
Experience window — how many years of history feed the rate. A long window is stable but lags a changing risk; a short one is responsive but noisy.
Re-rating cadence and lag — how often, and how long after the outcome, prices move. Faster closes the loop but destabilizes budgets.
Aggregation granularity — individual, sub-segment, or whole-class rating; finer surfaces real differences but shrinks each sample.
Drift trigger — how large a swing in the pool's loss ratio counts as "the base rate is stale" and forces recalibration.

When it helps, and when it misleads¶

Its strength is that it catches exactly what upfront screens miss: the hidden type that no document revealed but that the losses eventually expose, and the slow compositional drift that only shows up in aggregate. Because it is grounded in realized outcomes, it is the hardest term for an entrant to fake.

Its central weakness is that it is retrospective — you pay real losses to learn the lesson, so it complements an entry screen rather than replacing it. On thin samples it confuses bad luck with bad type, which is precisely why credibility weighting exists as the standard corrective.^[1] It can also be gamed from the other side, by suppressing or under-reporting claims to protect a mod, and a tidy factor can lend false precision to what is still a small, noisy sample. The discipline that keeps it honest is to weight individual experience by its credibility, watch pool aggregates rather than reacting to single events, and treat the base rate as something to recalibrate, not a constant.

How it implements the components¶

Claims or Outcome Experience Rating realizes the archetype's feedback-and-monitoring arm — the components that keep the filter honest over time, not the ones that decide entry:

base_rate_and_feedback_calibration — its core loop: realized outcomes are fed back to update the class base rates so the whole system doesn't overfit stale data.
pool_composition_monitor — it tracks loss ratios, claim frequency, and exits at the pool level to detect compositional drift the upfront filter let through.

It does not set the forward price or terms — Risk-Adjusted Pricing consumes its recalibrated base rates for that — nor does it screen entrants at the gate (Prequalification Process, Underwriting Review) or build the per-applicant hidden-type estimate (Underwriting Review).

Instantiates: Adverse Selection Filtering — it is the feedback loop that keeps the pool's filter calibrated after exposure begins.
Sibling mechanisms: Risk-Adjusted Pricing · Underwriting Review · Deductible or Copay Schedule · Minimum Eligibility Standard · Prequalification Process · Probationary Entry · Quality Certification Requirement · Risk Tier Assignment · Seller Rating or Quality Grading · Waiting Period · Warranty or Guarantee Requirement

Notes¶

Experience Rating is the only backward-looking member of this set. It cannot protect a pool at the moment of entry — by the time it speaks, exposure has already happened — so it is a partner to the upfront screens, not a substitute. Its real product is a corrected base rate; keeping that separate from the price lets the pricing and underwriting mechanisms improve as the calibration improves, without re-deriving their own machinery.

References¶

[1] In actuarial practice, credibility theory is the formal method for blending an individual's own limited experience with the broader class average, weighted by how statistically credible the individual sample is. It is the standard guard against re-rating a participant on a run of good or bad luck. ↩