The ERP That Won't Quit
Midstate Distributors, a $60 million regional wholesaler with 400 employees, is two years into a company-wide ERP implementation. The original budget was $3.2 million; $4.7 million has been spent and the system is 14 months late. The vendor's latest recovery plan says $1.8 million more and eight months. An independent reviewer hired by the board estimates $3.5 million and 18 months, and warns that two critical inventory modules may never work as promised. The legacy system still processes orders but vendor support ends in 12 months. The CFO argues, 'We cannot walk away from $4.7 million.' The COO wants to scrap the project and rebuild internally. The CEO, who sponsored the ERP, has stopped attending steering meetings. The IT project lead resigned last month, and two analysts are interviewing elsewhere. The board has asked for a go/no-go decision within 30 days and wants a documented rationale.
1. What decision error does the CFO's statement illustrate, and how should the board frame the go/no-go analysis?
Read the worked answer
The CFO is expressing the sunk-cost fallacy and, more broadly, escalation of commitment. The $4.7 million is irrevocable; it should not influence a go/no-go decision except insofar as it signals that forecasts were unreliable. The board should compare the incremental future costs and benefits of the realistic alternatives: continue with the vendor under a renegotiated, milestone-based plan; replace the vendor or rebuild internally; or abandon and run the legacy system while planning a later replacement. Each option must include the real deadline created by legacy support ending in 12 months. The vendor's $1.8 million/eight-month estimate and the independent reviewer's $3.5 million/18-month estimate should not be averaged; the board should ask what assumptions drive the gap and assign probabilities or a contingency reserve. A defensible frame is to choose the option with the best risk-adjusted net benefit, not the option that recovers past spending. That may still be continuing, but only because remaining benefits exceed remaining costs-not because money has already been spent.
2. The board has 30 days and incomplete information. Apply bounded rationality and satisficing to recommend a decision process.
Read the worked answer
Bounded rationality means the board cannot optimize; it lacks complete information, time, and computational capacity. Satisficing means setting minimum acceptable criteria and choosing the first available option that meets them. In 30 days the board should define must-haves-for example, functional inventory, vendor support for 24 months, total remaining cost under a stated ceiling, and implementation before legacy support ends. It should commission a short independent verification of the two disputed inventory modules. It should require the vendor to accept a stage-gate contract with a 60-day proof-of-capability milestone and financial penalties, and prepare a fallback with the legacy vendor or an internal rebuild. This preserves options instead of forcing a premature yes or no. A satisficing rule might be: continue only if the vendor meets the milestone and the revised total stays below a board-approved threshold; otherwise switch. The process should document criteria before seeing the vendor's revised pitch, to reduce anchoring.
3. Identify two group biases likely in the board's discussion and one structural safeguard for each.
Read the worked answer
Two likely biases are groupthink and confirmation bias, reinforced by the CEO's sponsorship and the COO's public preference to scrap the project. Groupthink may suppress doubts because the CEO sponsored the ERP and has been absent from steering meetings; confirmation bias may lead members to notice only evidence supporting their prior position. Structural safeguards: assign a formal devil's advocate who argues the strongest case against the leading option, and run a pre-mortem in which the board imagines the chosen option failed two years later and writes the reasons. Also useful: collect anonymous written votes before discussion, invite the independent reviewer to present without the vendor present, and separate the decision chair from the project sponsor. A decision log recording criteria, assumptions, and dissents makes the rationale auditable. These safeguards do not guarantee a good outcome, but they reduce the chance that the board mistakes consensus or sunk cost for evidence.