Strategic Thinking, Decision Making, and Innovation-1

ILLUSTRATION BY GUSTAVO BRIGANTE

FEATURE STOP DOUBLING DOWN ON YOUR FAILING STRATEGY

110  HARVARD BUSINESS REVIEW NOVEMBER–DECEMBER 2017

y the end of the 1990s the British music com- pany HMV was on top of the world. Its busi- ness model—operating Main Street stores in

which customers could browse through a wide col- lection and listen to tracks with an in-store headset before they decided whether to buy a CD—had de- livered the company an enviable 40% market share in Britain.

HMV’s rise started with the pop music revolution of the 1960s, when the company began expanding its retail operations in London. It doubled in size in the 1970s and had established itself as the country’s leading specialist music retailer by the early 1980s. It opened stores in Ireland and Canada in 1986 and in the United States, France, Germany, and Japan soon afterward. By the 1990s it had more than 320 stores,

STOP DOUBLING DOWN ON YOUR FAILING STRATEGY HOW TO SPOT (AND ESCAPE) ONE BEFORE IT’S TOO LATE BY FREEK VERMEULEN AND NIRO SIVANATHAN

B about 100 of them in the United Kingdom. In 2002 HMV floated on the London Stock Exchange, valued at about £1 billion. By then, however, some employees and analysts

had started to express doubts about the long-term sustainability of HMV’s business model. Although the arrival of DVDs and computer games initially boosted store profits, supermarket chains had begun selling popular CDs at a discount, and in early 1998 Amazon had started selling CDs online. A few years later down- loadable music appeared on the internet, culminating in the launch of Apple’s iTunes store in 2003.

But HMV’s top management doggedly stuck to its strategy. In 2004 the company opened its 200th store in the UK and began acquiring rival chain stores, sometimes out of bankruptcy. By 2008 the company

NOVEMBER–DECEMBER 2017 HARVARD BUSINESS REVIEW 111 

was running a global network of more than 600 out- lets. As early as 2002 its advertising agency had tried to alert the board to pending dangers—online retail- ers, downloadable music, and supermarket discount- ing—but HMV’s managing director, Steve Knott, had angrily rejected the warning: “I have never heard such rubbish. I accept that supermarkets are a thorn in our side, but not for the serious music…buyer, and as for the other two, I don’t ever see them being a real threat; downloadable music is just a fad.”

Not until 2010 did HMV open a digital music store. By then, of course, the company was far too late to the party, and in January 2013 it went into receivership.

HMV’s story is a classic example of what is known in the management literature as an escalation of com- mitment: holding on too long to a strategy that was once successful. Of course, many factors can contrib- ute to the failure of a specific company, but in nearly every academic case study on the demise of a former leader in its industry, escalation was shown to play a major role. Nokia’s failure, for example, which has been well documented, was to a large extent caused by the company’s continued investment in its propri- etary operating system even as Android and iOS were dominating the market.

Once escalation takes hold, it can be difficult to re- verse, but you can reduce the chances of falling into that trap. The psychological and sociological dynam- ics underlying escalation have been researched by one of us (Sivanathan) and countless other scholars from many academic perspectives; in the following pages we draw on this rich body of work to offer tried and proven organizational rules to help managers de- sign their decision-making processes. But first we’ll look at the causes of escalation.

WHY IT HAPPENS Escalation of commitment is deeply rooted in the hu- man brain. In a classic experiment, two groups of par- ticipants were asked whether they would be willing to invest $1 million to develop a stealth bomber. The first group was asked to assume that the project had not yet been launched and that a rival company had already developed a successful (and superior) product. Unsurprisingly, only 16.7% of those participants opted to commit to the funding.

The second group was asked to assume that the project was already 90% complete. Its members, too, were told that a competitor had developed a superior product. This time 85% opted to commit the resources to complete the project.

These results underscore the fact that people tend to stick to an existing course of action, no matter how irrational. The project’s likely outcome was identical

IN BRIEF

THE PROBLEM Companies often stick too long to a once successful but failing strategy. The British music company HMV did so, and it went from commanding a 40% share of Britain’s music market to receivership in just over a decade.

WHY IT HAPPENS Research has identified many biases that explain why decision makers may escalate a prior commitment, including the sunk cost fallacy, loss aversion, the illusion of control, preference for completion, pluralistic ignorance, and personal identification.

THE SOLUTION Companies can reduce their exposure to escalation by adopting six practices: Set decision rules; pay attention to voting rules; protect dissenters; expressly consider alternatives; separate advocacy and decision making; and reinforce the anticipation of regret.

FEATURE STOP DOUBLING DOWN ON YOUR FAILING STRATEGY

ESCALATION OF COMMITMENT— CLINGING TO A ONCE SUCCESSFUL STRATEGY—IS DEEPLY ROOTED IN THE HUMAN BRAIN.

112  HARVARD BUSINESS REVIEW NOVEMBER–DECEMBER 2017

for both groups. Because a competitor had beaten the company to the market with a superior product, the new product was almost bound to fail. The only differ- ence between the two situations was the timing of the question: before commitment to the project versus when it was nearing completion.

What exactly is going on? Research has identified a number of mutually reinforcing biases that collec- tively explain why people’s judgment may be swayed by a prior commitment to a course of action. The six most important are: • The sunk cost fallacy. This bias is well known in

management literature. When making investment decisions, people often factor in costs they have already incurred. If they abandon a project, those costs won’t be recovered. Their hope is that if the project continues, the costs can be recouped, vin- dicating earlier decisions to invest. But a rational decision maker will look only at future costs, not at past ones.

• Loss aversion. This bias, too, is well established. If withdrawing from a course of action implies certain and immediate losses, decision makers often prefer to allocate more resources to continue with it—de- spite low expected returns—if they see any chance of turning the situation around.

• The illusion of control. This bias clearly rein- forces the previous two: People habitually over- estimate their ability to control the future. In one experiment two groups of participants bought lottery tickets for $1. One group was assigned ran- dom lottery numbers and asked at what price they would be prepared to sell their tickets. The average answer was $1.96. The second group, whose mem- bers were allowed to pick their numbers, wanted at least $8.67. Prior success—as in HMV’s case— tends to amplify the illusion; people are quick to take credit for the outcomes of decisions and also confuse having correctly predicted the future with having made it happen.

• Preference for completion. A wealth of psy- chological experimentation suggests that people have an inherent bias toward completing tasks— whether that means finishing a plate of food or seeing a project through.

• Pluralistic ignorance. Dissenters often believe that they alone have reservations about a course of action; as a consequence, they remain silent. Others, meanwhile, interpret their silence as agreement. In extreme cases this can result in ev- eryone’s agreeing to a decision that no one believes in. Jerry Harvey, of George Washington University, called this the Abilene paradox. He described a trip that he and his wife and parents made one 104° July afternoon in his parents’ unairconditioned

1958 Buick from Coleman, Texas, to Abilene. They had all tacitly agreed to the trip, but as it turned out, none of them had wanted to take it.

• Personal identification. Research in both psy- chology and sociology suggests that people’s identities and social status are tied to their com- mitments. Thus withdrawing from a commitment may result in a perceived loss of status or a threat to one’s identity. At the same time, no executive likes to admit that a decision was wrong, because the ability to make smart decisions is part of what defines a good executive. In combination, these biases lead a company’s

decision makers to ignore signals that their strategy is no longer working. It is what Karl Weick, of the University of Michigan, calls consensual neglect: the tendency of organizational decision makers to tacitly ignore events that undermine their current strategy and double down on the initial decision in order to justify their prior actions.

Powerful as these biases are, the research also shows that it is possible to counteract them by apply- ing certain processes and practices in decision mak- ing. In the remainder of this article we’ll describe the six of them that have proved most effective in a busi- ness context. A company that applies all six practices will significantly reduce its likelihood of falling into the escalation trap.

SET DECISION RULES One way to stimulate more-objective decision making is to agree to decision rules in advance. Intel, for ex- ample, when it was still focused on producing DRAM memory chips rather than microprocessors, made a rule that production capacity would be allocated to products according to several criteria, particularly margin per wafer. This objective formula was de- signed when no concrete decisions were yet at stake.

Some time later, when production capacity had to be allocated between the new technology of micro- processors and the old one of DRAMs (to which sev- eral top managers at the time were still firmly commit- ted), managers helped sway the company toward the new technology by pointing to the objective formula, which favored microprocessors.

When hard figures aren’t available and judgment must be applied, non-numerical rules can serve. A large television production group, for example, which owns companies across the globe, created a decision rule to guide investments in new series, which were always proposed by local companies rather than de- veloped centrally. After a series had been prototyped, it would be shown to the other production companies. If some of them signed up to license it for their home

01

NOVEMBER–DECEMBER 2017 HARVARD BUSINESS REVIEW 113 

EXECUTIVES CAN MAKE DISSENT SAFER FOR SUBORDINATES BY VOICING THEIR OWN DOUBTS.

markets, the series would automatically get funded. But if no other company was interested in the li- cense, the project would cease to exist. Thus, instead of leaving the decision to a small number of top man- agers, this decision rule tapped into the collective wisdom of the company’s highly knowledgeable on- the-ground executives.

When we asked the company’s CEO why he didn’t just make these investment decisions himself, he replied, “Why would I know any better than all the other very experienced television executives in my firm? It is not my job to make the decision; it is my job to make sure the best decision gets made.”

PAY ATTENTION TO VOTING RULES Creating a decision rule requires careful reflection, because quite subtle differences can lead to oppo- site outcomes. Consider the following situation: The three members of a top management team are debat- ing whether to continue investing in the company’s current technology or switch to a new one. They agree that two criteria are relevant: (1) whether the current technology is likely to require substantial ad- ditional investment; (2) whether the new technology is likely to improve significantly over time. They also agree that they should switch only if it appears that both criteria are met.

Let’s suppose that Team Member 1 thinks that both criteria are met, Member 2 thinks that only the first is met, and Member 3 thinks that only the sec- ond is. The team’s recommendation will depend on how those opinions are aggregated. As shown in the exhibit “Rethink How You Count Votes,” if you tally by team member (which academics describe as con- junctively), the team will continue investing in the existing technology, because it’s clear that two out of three members don’t believe both criteria have been met. But if you tally by criterion (disjunctively, in academic jargon), each garners two votes for and only one against, meaning that the company should switch to the new technology.

Note that in both situations, the criteria are ex- actly the same and the team members hold exactly the same opinions. It’s the procedure that makes the difference.

Most companies follow a conjunctive procedure (simply tallying people’s overall judgments). But as the example above suggests, this procedure is likely to lead to escalating commitment, because it tends to overwhelm reservations about the status quo. We argue that when a company is evaluating whether to switch to an alternative strategy, a disjunctive proce- dure will better reflect any growing unease with the current course of action.

PROTECT DISSENTERS Companies that have doubled down on a failing strat- egy are usually not without dissenters. The trouble is that dissenters can be ruthlessly suppressed—and the knowledge that this might happen itself acts as a sup- pressant. We also know from various studies in social psychology that people are reluctant to speak up if they think they are alone in their disagreement.

That’s because they’re engaging in what scholars call a tacit calculus: balancing the immediate risk of speaking up against a course of action (and potentially being dismissed by the group) against the longer- term consequences of not speaking up (and possibly witnessing the failure of their organization). When the probability of being dismissed appears high, they will opt to remain silent. Chances are, moreover, that loss aversion bias will cause them to overweight the probability of being dismissed.

To prevent escalation, it is essential that leaders create an environment in which people do speak up, share dissenting information, and challenge the or- ganization’s course of action. Amy Edmondson, of Harvard Business School, refers to this as psycholog- ical safety: a belief that one will not be punished or humiliated for sharing ideas, questions, or concerns. Organizations can create this safety by:

Providing anonymous feedback channels. Creating safe channels that lower-level executives can use to share opinions is one way to surface dissent. These channels can take multiple forms, such as an

FEATURE STOP DOUBLING DOWN ON YOUR FAILING STRATEGY

02

03

114  HARVARD BUSINESS REVIEW NOVEMBER–DECEMBER 2017

online system or a third party. Research indicates that management consultants, for example, can play this role effectively—provided they are explicitly hired for that purpose.

Deploying larger teams. CEOs often rely heavily on a kitchen cabinet or an executive committee con- sisting of just three or four trusted colleagues. But in a small team, a dissenter may well be a lonely voice. A review of 97 studies in social psychology showed that single-person minorities consistently had min- imal influence on majority opinions, because they were easily discounted as reflecting an idiosyncratic perspective. In a team of four, therefore, three people who agree are inclined to dismiss the differing opin- ion of the fourth person, even though she represents 25% of the team. The good news is that it takes only two to get a hearing: Research shows that in a team of 12, people will pay attention if only two members disagree, even though they represent less than 17% of the team.

RETHINK HOW YOU COUNT VOTES A team of three must recommend whether its company should change its core technology. Managers agree that this should happen only if both of two criteria are met. What this team recommends will depend on how the members’ votes are counted.

CRITERION #1 FURTHER

INVESTMENT IN THE OLD TECHNOLOGY

IS NEEDED

CRITERION #2 THE NEW

TECHNOLOGY IS LIKELY TO IMPROVE

CONJUNCTIVE PROCEDURE

(TALLY BY MEMBER): SWITCH

TECHNOLOGIES?

TEAM MEMBER 1 YES YES YES

TEAM MEMBER 2 YES NO NO

TEAM MEMBER 3 NO YES NO

DISJUNCTIVE PROCEDURE (TALLY

BY CRITERION): SWITCH

TECHNOLOGIES?

YES YES

In general, therefore, we suggest that CEOs avoid delegating input on strategic decision making to groups of only four or five people. To be sure, smaller teams reduce coordination and communication costs and reach consensus faster. But larger teams have more information-processing capacity and a greater diversity of perspectives. We recommend enlisting 10 to 14 executives when it comes to debating the company’s long-term strategy. (More than 14 is inad- visable, because members of very large teams tend to disengage.)

Calibrating diversity. In addition to enlarging the strategy-making team, companies should increase its diversity. More than two decades’ worth of research demonstrates that diverse groups produce more in- novative and creative solutions, are better at solving complex problems, and are more capable of incorpo- rating novel information. But diversity must be care- fully calibrated. Consider the two teams in the exhibit “Make Sure Your Teams Have Subgroups.” On Team 1, every member is demographically unique. Team 2, however, has two distinct subgroups. Research by one of us (Vermeulen) shows that teams with sub- groups are more likely to develop alternative courses of action, because the probability is greater that no dissenter will be alone.

Modeling doubt. Executives can make dissent safer for subordinates by expressing their own doubts about a current strategy. To be sure, leaders are not used to doubting themselves—a situation reinforced by the fact that followers expect them to be decisive and con- fident. But the payoff for occasionally admitting some fallibility can be significant.

Consider this example from a large European air- line. The top management team had been planning a major new investment for one of its key divisions. During the final meeting with the three senior exec- utives involved in the plan, the CEO decided to make sure that everybody was really on board. He stood up and declared that he was willing to proceed, but he thought they should know that he felt unsure about it. After a short silence, another executive spoke up, admitting that he, too, had been having doubts. He was swiftly followed by a third person, who carefully explained his reasons for lacking confidence in the venture’s chances of success. It appeared that of the four people in the room, only one really wanted the project to go ahead.

Yet until then, none of them had openly opposed the investment. Not until the CEO’s public admission of doubt did the other executives feel psychologi- cally safe enough to admit reservations and surface arguments to end the course of action. The team abandoned the project, and the division concerned remained one of the corporation’s most profitable.

NOVEMBER–DECEMBER 2017 HARVARD BUSINESS REVIEW 115 

EXPRESSLY CONSIDER ALTERNATIVES For a study published in 2009, Shane Frederick, a pro- fessor at Yale, ran a revealing experiment with two groups of participants. Both groups were asked to as- sume that they had a sum of money available to buy themselves a present. They were told to imagine that on a trip to a video store, they came across a DVD on sale for $14.99 that included their favorite actor or actress and was their favorite type of movie.

The first group was given a simple binary choice: (1) buy the video; (2) don’t buy the video. In this group 75% bought the video. The second group, however, was given a slightly different choice: (1) buy the video; (2) don’t buy the video and keep the $14.99 for some- thing else. Only 55% of this group chose to buy the video. The simple reframing of options to include do- ing something else with the money was sufficient to significantly shift people’s decisions.

This experiment suggests that framing strategic questions to include the possibility of alternatives is an effective way to avoid an escalation of commitment to one course of action. Of course, it also means that you must have alternatives available (and research shows that spending time and money on consider- ing them is generally well worth it). Paul Nutt, of the Ohio State University, analyzed 137 key decisions in as many North American companies and found that when only one course of action had been considered, 52% of the decisions resulted in failure. By contrast, when just one alternative had been considered, the failure rate dropped to 32%.

SEPARATE ADVOCACY AND DECISION MAKING Managers who initiate a course of action are more likely to continue funding it (even in the face of fail- ure) than managers who assume leadership after a project is started. You can reduce the likelihood of es- calation if you give responsibility for a strategic move to people who did not advocate or initiate that move.

Research in banking, for example, shows that loan officers who have approved a loan to a particular cli- ent often escalate their commitment to the borrower by assigning further loans, even if the borrower is rel- atively likely to default. Banks that make a practice of separating initial credit decisions from subsequent requests outperform banks that place those deci- sions in the same hands. Similarly, other research has found that new managers tend to rate underperform- ing employees less favorably than the managers who hired them; likewise, entrepreneurs who buy existing businesses invest less capital than the entrepreneurs who established them.

FEATURE STOP DOUBLING DOWN ON YOUR FAILING STRATEGY

MAKE SURE YOUR TEAMS HAVE SUBGROUPS Diversity helps creativity, but if everyone is different, there’s a risk that no one will speak up. In building a team, therefore, make sure each member can identify a potential fellow dissenter. Both teams shown below include men, women, whites, and Asians of various ages, functions, and tenures. But in Team 1 no two members are alike, whereas Team 2 has two distinct subgroups. Team 2 is therefore more likely to have a debate around decisions.

TEAM 1

Age 28 29 52 54

Sex male female male female

Ethnicity Asian white white Asian

Function finance sales production finance

Tenure 2 11 3 13

TEAM 2

Age 28 29 52 54

Sex male male female female

Ethnicity Asian Asian white white

Function finance finance sales production

Tenure 2 3 11 13

05

04

116  HARVARD BUSINESS REVIEW NOVEMBER–DECEMBER 2017

The British bank Barclays offers a good example of the wisdom of separating decision making from strategy advocacy. In 2007, after much preparation and internal negotiation, Barclays decided to make a £43 billion bid for the Dutch bank ABN AMRO. Unexpectedly, the Royal Bank of Scotland (RBS) made an unsolicited rival bid of £48 billion. A takeover battle was in the cards, and the Barclays executive team was gearing up to raise its bid. The Barclays board, how- ever, was persuaded by independent directors to vote against the move, and the bank withdrew its offer.

RBS ended up acquiring ABN AMRO, taking on a lot of debt in the process. Barclays’s decision proved smart: When the financial crisis struck, RBS was among the hardest hit of the big UK clearing banks because of its high leverage.

REINFORCE THE ANTICIPATION OF REGRET The social psychologist Marcel Zeelenberg has defined regret as an “emotion that we experience when real- izing or imagining that our present situation would have been better had we decided differently.” A good way to prevent doubling down on a failing strategy is to get managers to anticipate the regret they may feel at not having taken a different road. This can be done in two ways:

By taking a temporal perspective. The first approach is to get people to explicitly consider what might go wrong with the current strategy. Of course, companies claim to routinely undertake this sort of exercise, but in most cases they simply ask managers to look forward in time. That’s unlikely to be help- ful. Ample research in social psychology, including our own, has shown that people—especially those in leadership positions—are inherently overoptimis- tic about the future and their ability to affect it (the illusion of control).

A far better exercise is to get people to imagine a concrete scenario and then work backward, using what is called prospective hindsight. For example, instead of asking people to imagine why a strategy might fail, try telling them, “It is January 2025, and the unexpected has occurred: Our strategy has failed to deliver even a respectable market share. Think about the reasons why.” J. Edward Russo, of Cornell, conducted several experiments along these lines with various colleagues. They found that partici- pants who were prompted to apply prospective hind- sight to a course of action came up with about 25% more ways it could fail than those presented with an exercise in forecasting—and the reasons surfaced through prospective hindsight tended to be more specific and relevant to the situation.

One form of this, introduced by the research psy- chologist Gary Klein, is the “premortem.” At a point when a management team had almost come to an im- portant decision but was not yet formally committed, he would say, “Imagine that we are a year into the fu- ture. We implemented the plan as it now exists. The outcome was a disaster. Please write a brief history of that disaster.”

By taking an interpersonal perspective. You can also persuade managers to question commit- ment and consider alternatives by getting them to step into different roles. If they end up imagining a compelling new strategy as a result, the potential for regret will increase.

Intel again provides a classic example. CEO Gordon Moore was initially reluctant to withdraw from DRAM, because it was “the product that had made Intel.” He changed his mind only after the company’s cofounder Andy Grove famously asked him, “If we got kicked out and the board brought in a new CEO, what do you think he would do?”

We recommend a similar exercise: Create three groups of no more than five members of your top man- agement team and ask them to prepare answers to the following questions for presentation to the full team:

Group 1: Imagine that an entirely new executive team enters the company. What would it change?

Group 2: A hedge fund has shorted our stock. Please explain its reasoning.

Group 3: A small group of middle managers have produced a memo urging us to change course. Please write down their arguments.

Variants of this exercise can be developed accord- ing to the strategic issue at hand. Whatever its pre- cise form, purposeful perspective taking can enable decision makers to imagine dissent.

BY ITS NATURE, an escalation of commitment is difficult to detect. Rather like the apocryphal frog that doesn’t know until too late that it’s being boiled alive, over- committed executives are prone to ignore signs of their company’s imminent collapse. That is precisely why companies need to establish organizational processes and practices of the kind we’ve laid out—to encourage managers at all levels to make decisions more objec- tively and explicitly consider alternative strategies and perspectives. HBR Reprint R1706H

FREEK VERMEULEN is an associate professor of strategy and entrepreneurship at London Business School and the

author of Breaking Bad Habits: Defy Industry Norms and Reinvigorate Your Business (Harvard Business Review Press, 2017). Twitter: @Freek_Vermeulen. NIRO SIVANATHAN is an associate professor of organizational behavior at London Business School whose research has explored strategies for de-escalating commitment. Twitter: @Niro_Sivanathan.

06

NOVEMBER–DECEMBER 2017 HARVARD BUSINESS REVIEW 117 

Copyright 2017 Harvard Business Publishing. All Rights Reserved. Additional restrictions may apply including the use of this content as assigned course material. Please consult your institution's librarian about any restrictions that might apply under the license with your institution. For more information and teaching resources from Harvard Business Publishing including Harvard Business School Cases, eLearning products, and business simulations please visit hbsp.harvard.edu.