TL;DR
A gate with authority to kill rarely stops failing work, because the continue-or-stop call goes to the people who approved it. If the program falls short of a target set at approval, someone who did not approve it should decide whether it continues.
What the paper develops
A finance-system consolidation is eighteen months into a three-year plan. Two of six business units have moved, each migration has cost well above its estimate, and the next round of funding covers the largest unit. The committee at the funding review can stop the program. The sponsor asks for one more quarter and names a new integration partner. The committee members, who approved the first two rounds of funding, request a revised schedule and release the funding. A quarter later the conversation repeats.
The standard remedy is a tougher gate. Stage-Gate's founders, Robert Cooper and Scott Edgett, write that many projects move too far without serious scrutiny and that once a project begins, “there is very little chance that it will ever be killed.” Their fix is tough Go/Kill decision points. This committee already had that authority.
The people deciding approved the work
Escalation research points at who holds the decision. In Barry Staw's classic experiment, people personally responsible for a bad outcome committed the most additional resources. A meta-analysis of more than 35 years of studies linked personal responsibility for the initial decision to escalation, and in four studies the effect was stronger when more than one person held authority over the decision or the decision maker was monitored. A funding review fits that description: the responsible sponsor argues the case before a committee that shared the approval.
Field evidence points the same way. Across 132 California banks over nine years, turnover among senior executives predicted later write-offs of bad loans, and the write-offs did not predict the turnover. In troubled software projects, de-escalation was most often triggered by senior managers, internal auditors, or external consultants. Across IS projects, 30 to 40 percent showed some degree of escalation, with significantly worse perceived outcomes.
The executive who made a commitment is the person least able to end it. Giving the committee that shared the approval more authority likely strengthens the people already pulled toward continuing.
Give the decision to someone who did not approve the work
A laboratory comparison of de-escalation techniques found three most effective: minimum targets set before feedback, evaluation on decision process rather than outcome, and making negative outcomes less threatening. Combined with the field evidence, they become three gate practices.
Record the minimum result at approval: the smallest result, by a date set early in the program, below which the original case no longer holds. Keep it separate from the forecast.
Tell the sponsor at approval that the evaluation will rest on the soundness of the case and how early a miss is raised. In the experiment, evaluation on outcomes did not reduce commitment, while evaluation on process did.
When the minimum is missed, the sponsor recommends continuing, narrowing, pausing, or stopping, and a reviewer who did not approve the funding decides. The EPMO tracks the date and convenes the reviewer. The approving committee stays informed and can ratify the outcome, but it does not make the call.
Where the design stops
The reviewer can still continue the work on a revised case, because some troubled projects are turned around. Reserve the split for commitments large enough that a wrong continuation matters. The technique evidence comes from one laboratory experiment, the shared-authority finding rests on four studies, and no study has tested the design as a whole.
At the next funding approval, choose one commitment and record the minimum result and its date, how the sponsor will be judged, and who decides if the minimum is missed. When a sponsor next asks for one more quarter, nobody deciding will have to overturn their own earlier judgment.
The operating move
At the next funding approval, record the minimum result and its date, how the sponsor will be judged, and the reviewer with no prior approval who decides if the minimum is missed.
Inside the white paper
- Why a gate with authority to kill still extends failing work, and why the people asked to stop it are the least able to
- What bank write-offs and troubled software projects show about who ends a losing commitment
- A three-part stop design: a minimum result set at approval, sponsor evaluation on the case, and a reviewer with no prior approval
Sources and notes
- Robert G. Cooper and Scott J. Edgett, "Stage-Gate and the Critical Success Factors for New Product Development," BPTrends — the method's originators prescribe tough Go/Kill gates and observe that begun projects are rarely killed.
- Mark Keil, Joan Mann, and Arun Rai, "Why Software Projects Escalate," MIS Quarterly 24(4), 2000 — 30 to 40 percent of IS projects showed some escalation, with significantly worse perceived outcomes.
- Barry M. Staw, "Knee-Deep in the Big Muddy," Organizational Behavior and Human Performance 16, 1976 — participants personally responsible for a bad outcome committed the most additional resources.
- Dustin J. Sleesman, Donald E. Conlon, Gerry McNamara, and Jonathan E. Miles, "Cleaning Up the Big Muddy," Academy of Management Journal 55(3), 2012 — meta-analysis linking personal responsibility to escalation, with a stronger effect when decision authority is shared.
- Barry M. Staw, Sigal G. Barsade, and Kenneth W. Koput, "Escalation at the Credit Window," Journal of Applied Psychology 82(1), 1997 — across 132 California banks over nine years, executive turnover predicted write-offs of problem loans; write-offs did not predict turnover.
- Mark Keil and Daniel Robey, "Turning Around Troubled Software Projects," Journal of Management Information Systems 15(4), 1999 — de-escalation was most often triggered by senior managers, internal auditors, or external consultants.
- Itamar Simonson and Barry M. Staw, "Deescalation Strategies," Journal of Applied Psychology 77(4), 1992 — pre-set minimum targets, process-based evaluation, and less threatening outcomes reduced commitment to a losing course.