The Mechanism
When a matter concludes, its actual outcome and actual duration become known facts for the first time. Resolution captures those facts as structured labels, matched back to the original prediction the position was priced against, so the gap between predicted and realized outcome is measured rather than assumed. This is not always a single event: a matter can settle in stages, resolve on some claims while proceeding on others, or reach a judgment later modified on appeal, and each partial resolution is captured as its own data point rather than collapsed into a single pass-fail grade recorded only when the file finally closes.
The mechanism also includes the settlement and payoff reconciliation that resolution requires in practice: confirming the final proceeds match the passport's economic rights record, confirming lien and subrogation obligations were satisfied in the correct priority, and confirming the capital position was repaid on the terms deployment established.
Resolution closes the loop rather than simply closing the file. The structured label produced here does not sit in an archive; it becomes part of the same corpus the predict stage draws its comparable population from, which means every fully resolved matter makes the next prediction for a similar matter marginally better informed than it would otherwise have been.
Inputs and Outputs
Inputs are the matter's final disposition, its actual duration from filing to resolution, and the reconciled settlement or judgment stack. Outputs are a structured resolution label matched to the original prediction, an updated repayment and recovery record for the funded position, and, critically, a data point that re-enters the training population the predict stage draws on for every subsequent matter with similar characteristics. Skipping this last output does not make the original model's accuracy irrelevant; it makes it permanently unknowable for this class of matter.
The repayment and recovery record produced here is also what closes the loop back to underwriting: it is the realized counterpart to the priced terms underwriting established, and the comparison between the two is the raw material for every subsequent calibration review of that pricing function.
What Can Go Wrong
The most common failure is treating resolution as a closing formality rather than a measurement event, recording that a file closed without recording how its actual outcome compared to what was predicted. A second failure is failing to capture partial and staged resolutions as they occur, which discards the information in exactly the cases, mixed verdicts, phased settlements, appealed judgments, that carry the most nuance about where a model's calibration held and where it did not. An institution that cannot produce, for its own book, a comparison of predicted outcome distributions against realized resolutions has no basis for claiming its underwriting is model-driven, regardless of what its materials assert.
A third failure is closing the loop only partially: capturing the outcome label but not the reconciled settlement stack, or reconciling proceeds without recording the resolution label needed for calibration. Both halves are required, because a resolution record that only measures repayment tells the institution whether it got paid, not whether the model that priced the position was actually right.
A fourth failure is allowing recency bias to distort calibration review: weighting the most recent resolutions more heavily than an unbiased sample would justify, simply because they are freshest in reviewers' minds, can mask a longer-running drift in model accuracy that only shows up when the full resolved population is reviewed together rather than a recent slice of it.
How Criterica Implements This Stage
Final disposition, duration, and settlement-stack reconciliation are captured for every position as a required close-out step, not an optional one. Criterica Intelligence's model calibration process consumes those structured resolution labels as the training population's next update, which is what makes the lifecycle a loop rather than a pipeline: a pipeline processes each matter once and moves on, a loop uses every completed matter to sharpen what the next matter is compared against. Criterica Group's standard requires published calibration reporting, predicted-versus-actual error tracked and disclosed by model family, so that resolution's discipline is auditable rather than merely claimed.
Resolution is also where the lifecycle's governing premise is tested most directly: integrity, evidence, and governance were established before economics were optimized at every prior stage, and resolution is the stage that either confirms that discipline held or reveals precisely where it did not, which is the entire reason the loop exists.
Discuss the Resolve Stage
Institutional buyers and capital partners evaluating the Regulated Outcome Lifecycle.