Outcome Lab

Outcome Lab — compare legal options with simulation, not guesswork

Lay out the realistic options on a dispute as stage-by-stage paths with ranges for cost, time and chance. A seeded simulation runs thousands of futures per option and shows where each one tends to end up — and what would change the answer.

What it does

What you can do

  • Two to five options, each a sequence of one to six stages with a chance, a cost, a duration and what you gain or pay.
  • Ranges as fixed values, uniform, triangular, PERT or lognormal distributions.
  • 200 to 20,000 simulated runs per option (4,000 by default), with the same random draws across options so the comparison is fair.
  • Average and p10/p50/p90 net value, chance it works, chance of a loss, time, endings, regret, five stress tests, kill lines and a chart of what moves the answer most.
  • Every assumption is tagged observed, inherited, inferred or defaulted. A model may propose the options; the numbers are always computed by the engine.
  • Each run records a forecast; resolve it later and the lab is scored (Brier score, meaningful after five resolved forecasts).

Good to know

Limits, stated plainly

  • Results are labelled “Model estimates — uncalibrated”. They are not a prediction of any court's decision.
  • Default ranges are wide and illustrative, not court statistics. Replace them with what you know.
  • In the Judiciary workspace the lab is refused for cases before the court.

Related

Works well with

See it on a real workspace.

Open the demo, pick a persona and click around a workspace filled with synthetic matters, deadlines and documents.

The live demo opens a workspace filled with synthetic data — pick a demo persona on the sign-in page. Self sign-up is closed while India hosting is being arranged; request early access to be told when it opens.