The Regulated Outcome Lifecycle is Criterica Group's account of the architecture running underneath every product built on the platform: five stages — Predict, Underwrite, Deploy, Monitor, Resolve — organized as a closed loop rather than a linear pipeline. A prediction states a probability distribution over how a matter resolves and how long it takes. Underwriting takes that distribution and prices capital against it. Deployment commits the capital. Monitoring re-scores the position as new facts arrive rather than holding it on a stale estimate. Resolution captures what actually happened as a structured, comparable label — and that label feeds back into the population the next prediction draws on, which is what makes the five stages a lifecycle instead of five disconnected steps.
Publishing the Lifecycle as a standalone reference is a governance choice as much as an explanatory one. It gives institutional counterparties — funds, insurers, enterprise legal teams, and prospective partners — a fixed description of how prediction and capital allocation are meant to relate to each other across every Criterica product, regardless of which specific product or division they are engaging with. An institution can hold any part of the Criterica platform against this architecture and ask where a given function actually sits.
Alongside the Lifecycle, Criterica Group's Insights series opens with three essays aimed at the same institutional audience. The first develops the Lifecycle argument in full: why treating a prediction and an underwriting decision as the same act collapses two distinct sources of error into one number, and why the loop only compounds in value if resolved outcomes actually flow back into the model rather than stopping at the point capital is deployed.
The second essay addresses calibration directly — the property that separates a probability an institution can price capital against from a confident-sounding number with no verifiable track record. It draws a sharp line between a model whose stated probabilities are checked against realized frequency out-of-sample and out-of-time, and a language-model inference that produces a percentage-formatted answer with no population it was ever validated against. The essay sets out, concretely, what an institutional counterparty should demand from any vendor claiming a calibrated probability: the reliability curve itself, the sample size behind it, confirmation the test was out-of-sample and out-of-time, results broken out by the buyer's own exposure segments, and a stated re-validation cadence.
The third essay makes the case that litigation, regulatory enforcement, and insured claims are not three separate domains but one underwritable asset class, because each shares the same three structural properties: a defined, enumerable set of outcomes; a historical record that is verifiable rather than self-reported; and a bounded time window that makes it possible to check a forecast against what actually happened. The essay argues that this shared structure, not the subject matter of any one domain, is what makes a category tractable for outcomes intelligence in the first place.
Both the Lifecycle page and the Insights series are published without institutional statistics attached — the argument in each case is architectural and methodological, meant to stand regardless of the specific scale of any given model fleet or corpus at the time of reading. Further essays in the Insights series will publish on an ongoing basis at criterica.com/insights.