AI Forecasting — Pressure-Test the Forecast
Adversarial review of an AI-generated Q2 forecast.
About this simulation
A four-round, advanced decision-sciences simulation set inside the investment committee of Lutèce Mobility SA, a Lyon-based mid-cap industrial-tech firm (€620M revenue, 9% EBIT, €140M net debt). A German rail OEM — 31% of revenue — will dual-source unless Lutèce closes a silicon-carbide (SiC) technology gap before a binding 30 June 2026 supplier decision.
The committee controls a €90M discretionary envelope and three mutually exclusive paths with disclosed odds: escalate the in-house SiC program (€60M; 55% +€130M NPV / 30% +€25M / 15% −€70M write-off), license a proven US design (€45M; 85% +€70M / 15% −€10M), or walk away and defend the legacy silicon line (€0; certain +€15M but the OEM dual-sources, putting 31% of revenue at structural risk). €22M is already sunk into the in-house program, and a −€70M write-off would breach a 3.0x net-debt/EBITDA covenant.
Operationalizing Enrico Diecidue's research on aspiration levels, anticipated regret, and risk, the simulation forces you to (1) set an explicit aspiration line and compute expected value per path, (2) build a regret matrix and size each path's worst-case covenant exposure, (3) commit under the live deadline and react to a leaked signal that cuts Path A's on-time probability from 55% to 45% — hold, switch, or hedge (€75M combined) — and (4) escalate or cut losses on an interim result drawn from your own committed odds, then defend the decision trail.
The math rewards decision hygiene — a consistent aspiration, EV-and-regret-aware choice, a pre-set switch trigger, and covenant discipline — and punishes the five classic errors: computing EV with no aspiration, sunk-cost escalation, ignoring maximum regret, hedging into a covenant breach, and flinching on the signal instead of updating.
Final KPIs track Decision Quality (process consistency), Expected Value captured, Anticipated Regret, and Covenant Headroom.
Who it is for
An advanced simulation for participants used to working with the main frameworks and trade-offs of the subject, designed for executive education and experienced professionals.
How a session runs
- The instructor creates a session from the Eureka dashboard and invites the participants.
- Participants play 4 rounds. In each one they submit their decisions and the simulation calculates the results.
- The instructor follows each participant's progress and results from the dashboard, and uses the class results for the debrief.