Friday, December 1, 2017

Nuclear Safety Culture: Focus on Decision Making



McKinsey Five Fifty cover
We have long held that decision making (DM) is a key artifact reflecting the state of a nuclear organization’s safety culture.

The McKinsey Quarterly (MQ) has packaged a trio of articles* on DM.  Their first purpose is identifying and countering the different biases that lead to sub-optimal, even disastrous decisions.  (When specific biases are widely spread in an organization, they are part of its culture.)  A second purpose is to describe the attributes of more fair, robust and effective DM processes.  The articles’ specific topics are (1) the behavioral science that underlies DM, (2) a method for categorizing and processing decisions and (3) a case study of a major utility that changed its decision culture. 

“The case for behavioral strategy” (MQ, March 2010)

This article covers the insights from psychology that can be used to fashion a robust DM process.  The authors evidence the need for process improvement by reporting their survey research results showing over 50 percent of the variability in decisional results (i.e., performance) was determined by the quality of the DM process while less than 10 percent was caused by the quality of the underlying analysis. 

There are plenty of cognitive biases that can affect human DM.  The authors discuss several of them and strategies for counteracting them, as summarized in the table below.


Type of bias
How to counteract
False pattern recognition (e.g., saliency (overweight recent or memorable events), confirmation, inaccurate analogies)
Require alternative explanations for the data in the analysis, articulate participants’ relevant experiences (which can reveal the basis for their biases), identify similar situations for comparative analysis.
Bias for action
Explicitly consider uncertainty in the input data and the possible outcomes.
Stability (anchoring to an initial value, loss aversion)
Establish stretch targets that can’t be achieved by business as usual.
Silo thinking
Involve a diverse group in the DM process and define specific decision criteria before discussions begin.
Social (conformance to group views)
Create genuine debate through a diverse set of decision makers, a climate of trust and depersonalized discussions.


The greatest problem arises from biases that create repeatable patterns that become undesirable cultural traits.  DM process designers must identify the types of biases that arise in their organization’s DM, and specify debiasing techniques that will work in their organization and embed them in formal DM procedures.

An attachment to the article identifies and defines 17 specific biases.  Much of the seminal research on DM biases was performed by Daniel Kahneman who received a Nobel prize for his efforts.  We have reviewed Prof. Kahneman’s work on Safetymatters; see our Nov. 4, 2011 and Dec. 18, 2013 posts or click on the Kahneman label. 

“Untangling your organization’s decision making” (MQ, June 2017)

While this article is aimed at complex, global organizations, there are lessons here for nuclear organizations (typically large bureaucracies) because all organizations have become victims of over-abundant communications, with too many meetings and low value e-mail threads distracting members from paying attention to making good decisions.

The authors posit four types of decisions an organization faces, plotted on a 2x2 matrix (the consultant’s best friend) with scope and impact (broad or narrow) on one axis and level of familiarity (infrequent or frequent) on the other.  A different DM approach is proposed for each quadrant. 

Big-bet decisions are infrequent and have broad impact.  Recommendations include (1) ensure there’s an executive sponsor, (2) break down the mega-decision into manageable parts for analysis (and reassemble them later), (3) use a standard DM approach for all the parts and (4) establish a mechanism to track effectiveness during decision implementation.

The authors observe that some decisions turn out to be “bet the company” ones without being recognized as such.  There are examples of this in the nuclear industry.  For details, see our June 18,2013 post on Kewaunee (had only an 8 year PPA), Crystal River (tried to cut through the containment using in-house expertise) and SONGs (installed replacement steam generators with an unproven design). 

Cross-cutting decisions are more frequent and have broad impact.  Some decisions at a nuclear power plant fall into this category.  They need to have the concurrence and support of the Big 3 stakeholders (Operations, Engineering and Maintenance).  Silo attitudes are an omnipresent threat to success in making these kinds of decisions.  The key is to get the stakeholders to agree on the main process steps and define them in a plain-English procedure that defines the calendar, handoffs and decisions.  Governing policy should establish the DM bodies and their authority, and define shared performance metrics to measure success. 

Delegated decisions are frequent and low-risk.  They can be effectively handled by an individual or working team, with limited input from others.  The authors note “The role-modeling of senior leaders is invaluable, but they may be reluctant” to delegate.  We agree.  In our experience, many nuclear managers were hesitant to delegate as many decisions as they could have to subordinates.  Their fear of being held accountable for a screw-up was just too great.  However, their goal should have been to delegate all decisions except those for which they alone had the capabilities and accountability.  Subordinates need appropriate training and explicit authority to make their decisions and they need to be held accountable by higher-level managers.  The organization needs to establish a clear policy defining when and how a decision should be elevated to a more senior decision maker. 

Ad hoc decisions are infrequent and low-risk; they were deliberately omitted from the article. 

“A case study in combating bias” (MQ, May 2017)

This is an interview with a senior executive of a German utility that invested €10 billion in conventional power projects, investments that failed when the political-economic environment evolved in a direction opposite to their assumptions.  In their postmortem, they realized they had succumbed to several cognitive biases, including status quo, confirmation, champion and sunflower.  The sunflower bias (groups aligning with their leaders) stretched far down the organizational hierarchy so lower-level analysts didn’t dare to suggest contrary assumptions or outcomes.

The article describes how the utility made changes to their DM practices to promote awareness of biases and implement debiasing techniques, e.g, one key element is officially designated “devil’s advocates” in DM groups.  Importantly, training emphasizes that biases are not some personal defect but “just there,” i.e., part of the culture.  The interviewee noted that the revised process is very time-intensive so it is utilized only for the most important decisions facing each user group. 

Our Perspective 

The McKinsey content describes executive level, strategic DM but many of the takeaways are equally applicable to decisions made at the individual, department and inter-department level, where a consistent approach is perhaps even more important in maintaining or improving organizational performance.

The McKinsey articles come in one of their Five Fifty packages, with a summary you can review in five minutes and the complete articles that may take fifty minutes total.  You should invest at least the smaller amount.


*  “Better Decisions,” McKinsey Quarterly Five Fifty.  Retrieved Nov. 28, 2017.