Rationality Check — Catch the Bias Before the Board Votes

A four-round, advanced management-accounting simulation set inside Hellwig Antriebstechnik GmbH, a family-owned German Mittelstand maker of electric drive systems (EUR 540M revenue, 7.8% EBIT, 2,300 staff, three plants). On 12 March 2026 the COO — the founder's nephew and a board favourite — submits a EUR 24M proposal to fully automate the Plant 2 assembly line, claiming a 3.1-year payback and a positive NPV of +EUR 5.2M. The supervisory board votes in 9 working days. The new CFO hands the dossier to controlling with one instruction: 'Tell me whether this number is real.' Playing the Head of Controlling, you must perform Jurgen Weber's rationality assurance: not merely recompute, but expose the decision pathologies embedded in the case. (1) Diagnose the biases before touching the model — anchoring on the vendor's EUR 7.6M savings deck, optimism in the 6% volume growth (the order book shows 1.5%), the EUR 1.4M sunk-cost plea, and motivated forecasting — naming each and tying it to a line in the NPV model. (2) Demand the right analysis: replace the growth anchor with a scenario range, strip the EUR 1.4M sunk cost out of the decision, model a realistic 9-14 month ramp, and decide whether to commission an independent EUR 40K / 3-day savings estimate that eats the 9-day clock. (3) Confront the powerful sponsor in the business-partner conversation, choosing a stance and using evidence rather than authority without either capitulating or detonating the relationship. (4) Escalate or endorse: decide what goes to the supervisory board under controlling's name — sign the original (a EUR 24M write-down lands near a 3.0x net-debt/EBITDA covenant test), sign a corrected/phased EUR 9M pilot, or formally escalate a dissent — and design the rationality safeguards for all future capex sign-offs. The math rewards the textbook controlling sequence (detect before compute, independent verification, sunk-cost discipline, phased commitment, evidence-led escalation) and punishes the five classic errors: computing before critiquing, trusting the counterparty's numbers, importing the sunk-cost bias, capitulating to power, and escalating without evidence. Final KPIs track Rationality Score, Decision Quality, Controlling Credibility, the 9-day clock and the verification budget.