Leadership
Failing by Designby Rita McGrath
From the Magazine (April 2011)
Summary. Reprint: R1104E It’s hardly news that business leaders work in
increasingly uncertain environments, where failures are bound to be more common
than successes. Yet if you ask executives how well, on a scale of one to 10, their
organizations learn from failure,…
It’s hardly news that business leaders work in increasingly
uncertain environments. Nor will it surprise anyone that under
uncertain conditions, failures are more common than successes.
And yet, strangely, we don’t design organizations to manage,
mitigate, and learn from failures. When I ask executives how
effective their organizations are at learning from failure, on a
scale of one to 10, I often get a sheepish “Two—or maybe three” in
response. As this suggests, most organizations are profoundly
more
5/29/23 下午 11:07⻚码：1/19
biased against failure and make no systematic effort to study it.
Executives hide mistakes or pretend they were always part of the
master plan. Failures become undiscussable, and people grow so
afraid of hurting their career prospects that they eventually stop
taking risks.
I’m not going to argue that failure per se is a good thing. Far from
it: It can waste money, destroy morale, infuriate customers,
damage reputations, harm careers, and sometimes lead to
tragedy. But failure is inevitable in uncertain environments, and,
if managed well, it can be a very useful thing. Indeed,
organizations can’t possibly undertake the risks necessary for
innovation and growth if they’re not comfortable with the idea of
failing.
What Are Your No-Fail Zones?
Leaders need to be clear about where failure will be
tolerated—and where it won’t. Mike Eskew, the former
CEO of UPS, …
An alternative to ignoring failure is to foster “intelligent failure,” a
phrase coined by Duke University’s Sim Sitkin in a terrific 1992
Research in Organizational Behavior article titled “Learning
Through Failure: The Strategy of Small Losses.” If your
!!
5/29/23 下午 11:07⻚码：2/19
organization can adopt the concept of intelligent failure, it will
become more agile, better at risk taking, and more adept at
organizational learning.
How Failure Can Be Useful
Some of the failures I’m about to describe were the results of
intentional experiments. Others were completely unplanned and
unexpected. But all of them provide valuable takeaways. A certain
amount of failure can help you:
Keep your options open.
As the range of possible outcomes for a course of action expands,
the chances of that action’s succeeding diminish. You’ll improve
your odds if you make more tries. This is the logic driving
businesses that operate in highly uncertain environments, such
as venture capital firms (whose success rates range from about
10% to about 20%), pharmaceutical companies (which typically
create hundreds of new molecular entities before coming up with
one marketable drug), and the movie business (where, according
to one study, 1.3% of all films earn 80% of the box office).
Learn what doesn’t work.
Many successful ventures are built on failed projects. Apple’s
Macintosh computers emerged in part from the ashes of a now-
forgotten product called Lisa, which introduced a number of the
graphical user interfaces and mouse operations in today’s
computers.
5/29/23 下午 11:07⻚码：3/19
In truly uncertain situations, conventional market research is of
little use. If you had asked people in 1990 what they would be
willing to pay for an internet search, no one would have known
what you were talking about. A massive amount of
experimentation was needed before workable search engines
emerged. Early entrants sought to be paid for doing the searches
themselves. Later, companies explored business models based on
advertising. Later still, Google developed a system to maximize
the profitability of the ad-based model. Without all that trial and
error, it’s highly unlikely that Google could have built the
algorithm-based juggernaut so familiar today.
Create the conditions to attract resources and attention.
Organizations tend to move on to new projects rather than fix
systemic problems with existing ones. Let something big go
wrong, though, and it’s all hands on deck!
I was personally introduced to how failure can be used
strategically years ago, when I worked for the City of New York. I
ran an IT group charged with installing an automated
procurement system. I was blissfully unaware of how challenging
it would be to gain political support and financial resources for
the project. Luckily, my boss was a political genius. One
afternoon, while I was running some analytics, I learned that the
data in the old system had become corrupted. I leaped into action,
determined to save the day. But after I ran my plan past my boss,
he quietly said, “Don’t do any of that. Sometimes things have to
fall apart before anybody musters the will to fix them.” He was
5/29/23 下午 11:07⻚码：4/19
absolutely right. The failure of the old system created a
compelling argument for the new one and was a turning point in
gaining support.
Make room for new leaders.
Sad but true: Even today many leadership positions are held by
people very much like those who selected them. Entire industries
have suffered the consequences of “lifers” who don’t challenge
unspoken assumptions and taken-for-granted rules. Only when
those assumptions and rules are proven ineffective—often,
unfortunately, in the course of great trauma—do boards recruit
fresh leaders. The change can be surprisingly beneficial. The U.S.
auto industry provides a case in point. Who would have thought
that Alan Mulally, a former senior executive at Boeing, would be
an inspirational turnaround CEO for Ford?
Develop intuition and skill.
Researchers say that what people think of as intuition is, at its
heart, highly developed pattern recognition. Those who have
never faced a negative outcome have a critical gap in the body of
experience that intuition is based on. Many venture capitalists
won’t invest in a new enterprise if the founder has never
undergone failure.
Many venture capitalists won’t investin a new enterprise if the founder hasnever undergone failure.
5/29/23 下午 11:07⻚码：5/19
Microsoft’s successful entrant in the game business, the Xbox
360, was developed by a team that had worked on 3DO’s failed
game console, the unsuccessful WebTV, Apple’s problematic
video card business, and Microsoft’s own short-lived UltimateTV.
Having been through so many disappointments, the team
members were able to spot warning signs and make smart course
corrections. For example, the earlier Xbox had used expensive
chips from outside manufacturers, and it reportedly lost about $4
billion from 2001 to 2005. The Xbox 360 team chose different
manufacturers, worked in close partnership with them to develop
the chips, and retained intellectual property rights to the chips,
allowing the system to generate profits very early on.
Putting Intelligent Failure to Work
Obviously, not all failures are useful, and even some that we could
learn from should be avoided at all costs. But if you accept that
failures will sometimes occur in uncertain environments, it
makes sense to plan for, manage, and learn from them—and in
many cases to consider them experiments rather than failures.
Here are seven principles that can help your organization
leverage learning from failure.
Principle 1: Decide what success and failure would look
likebefore you launch an initiative.
It never ceases to amaze me how often people working on the
same project have entirely different views of what would
constitute success. In one case I studied, an organization that
made environmental remediation equipment was hoping to
introduce a new product line. The marketing group thought the
5/29/23 下午 11:07⻚码：6/19
equipment’s selling point was that it met a tough new regulatory
standard. The engineering group thought the point was cost-
effectiveness—and to keep costs down, it was designing out the
very features the marketing group wanted to sell. This gap in
understanding could easily have led to a failure of the
unintelligent variety. But the company found out about it in time
to get everyone on the same page and prevent what could have
been a marketplace disaster.
Principle 2: Convert assumptions into knowledge.
When you’re tackling a fundamentally uncertain task, your initial
assumptions are almost certain to be incorrect. Often the only
way to arrive at better ones is to try things out. But you shouldn’t
start experimenting until you’ve made your assumptions explicit.
Write them down and share them with your team. Then make
sure that you and your team are open to revising them as new
information comes in. The risk is that we all have a tendency to
gravitate toward information that confirms what we already
believe—it’s called confirmation bias. A practical way to address
this bias is to empower one of your team members to seek out
information that suggests your course of action is flawed. You
want to find disconfirming information early, before you’ve made
extensive commitments and become resistant to changing your
mind.
Organizations that don’t record their assumptions tend to run
into two big problems. First, assumptions become converted into
facts in people’s minds. During a meeting, a manager might
venture a guess that a given market could generate $5 million in
sales—and before the meeting ends, the $5 million is baked into
5/29/23 下午 11:07⻚码：7/19
next year’s budget! This sort of leap causes all kinds of
dysfunctional behavior when the guess, almost inevitably, turns
out to be wrong. Second, such organizations don’t learn as much
as they could. They may right their course as they proceed,
learning as they go, but if they’re not rigorous about comparing
results with expectations, the lessons won’t be explicit and
shared, and future projects won’t benefit from them.
Having spelled out and revised your assumptions, you should
then design the organizational equivalent of an experiment to test
them. As with a scientific experiment, the idea is that whether or
not the outcome is what you’d hoped for, at least you will have
learned something.
Principle 3: Be quick about it—fail fast.
Quick, decisive failures have a number of important benefits.
First, they can save you from throwing additional resources at a
losing proposition. Second, it’s much easier to establish cause and
effect when actions and outcomes are close together in time.
Third, the sooner you can rule out a given course of action, the
faster you will move toward your goal. And finally, an early failure
lessens the pressure to continue with the project regardless,
because your investment in it is not large.
A practical way to help ensure that any failure happens quickly is
to test elements of your project early on. This is the main reason
that “agile software development” often produces better results
than the more conventional sequential process of systems design.
In an agile environment, small chunks of code are written and
shared in a quick, iterative fashion with other programmers and
5/29/23 下午 11:07⻚码：8/19
users before the team moves on. This is in sharp contrast to the
approach in which analysts spend months documenting user
requirements before submitting those requirements to
programmers, who only then begin coding. By the time a problem
is discovered, a project could have been heading in the wrong
direction for years.
Speed may require changing how you allocate resources. Instead
of going for maximum NPV over a project’s lifetime, for example,
you may want to break the financial evaluation into smaller
chunks in terms of both money and time. You may also want to
invest in more-flexible assets and people until you have learned
enough to confidently build a significant operation.
And the human benefits of failing fast should not be overlooked.
If people feel that a project’s failure will doom them to months of
waiting for another project, or to losing their jobs, then failure is
demoralizing. But if lots is going on and the conclusion of one
effort means that they’ll immediately get put on another (possibly
more interesting) project, then endings can be positive. At the
technical consultancy Sagentia, for example, employees are quick
to move from project to project. The finance director, Neil Elton,
told me, “They’ll proactively send around e-mails with a mini CV,
saying, ‘I was going to be busy, now I’m not. Can you use my
skills?’” This attitude is symptomatic of an organization that
knows how to experiment intelligently.
5/29/23 下午 11:07⻚码：9/19
Principle 4: Contain the downside risk—fail cheaply.
This is an important corollary to failing fast. Initiatives should be
designed to make the consequences of failure modest. Sometimes
it’s valuable to test a small-scale prototype before making a
significant investment. When the Japanese cosmetics firm Kao
was considering going into the manufacture of floppy disks, a big
question was whether or not customers would buy Kao-branded
disks. So the company went to another manufacturer and bought
disks that met its quality standards, put the Kao label on them,
and offered them to customers. The response was positive, so the
plan moved forward. Had the response been negative, Kao could
have stopped the project without incurring substantial costs.
This approach may require breaking ingrained habits. The chief
innovation officer of a highly technical company I worked with
observed that the company would typically get “some guy in a
white lab coat” to do a technical feasibility study before deciding
whether to enter a new product area. Such studies are not only
5/29/23 下午 11:07⻚码：10/19
expensive—upward of $200,000—but also relatively unindicative
of business feasibility. So the innovation officer started making
mock-ups of potential new products and showing them to
prospective customers. In many instances the company learned
that nontechnical issues, such as form factor, usability, and fit
with existing systems, would have prevented customers from
adopting a product. The difference in cost between the
approaches was an order of magnitude: A typical mock-up cost
around $20,000. The difference in speed was also considerable: a
few weeks rather than nine to 12 months.
3M’s reputation for being failure tolerant took a beating under
former CEO Jim McNerney, a GE-trained leader who sought to
utilize Six Sigma quality practices throughout the company, even
in its research labs. Although these worked wonders in 3M’s
factories, the emphasis on generating predictable results
hampered employees’ willingness to take risks on unproven ideas.
When George Buckley took the reins as CEO, in 2005, part of his
challenge was to restore the culture of risk taking. He
discontinued the use of Six Sigma in the labs and spurred
scientists and researchers to pursue new ideas—provided that the
downside was small. During the recession, 3M’s historical
philosophy of “make a little, sell a little” when introducing a new
product was successfully coupled with Buckley’s emphasis on
bottom-of-the-pyramid innovations—inexpensive items that
could appeal to very broad markets.
Principle 5: Limit the uncertainty.
5/29/23 下午 11:07⻚码：11/19
There isn’t much point to encouraging failure in an arena your
organization is already familiar with. But experiencing it in an
arena completely divorced from your current capabilities won’t
do you much good either: You probably won’t be able to use what
you find out, because you won’t understand the context and you
won’t know how to connect what you’ve learned to your existing
knowledge base.
Google, which is ordinarily very good at experimentation, went
too far afield when it tried to launch a non-internet radio venture.
The company wanted to automate the pricing of radio ads, as it
had with internet ads. Radio stations would give Google a portion
(ideally all) of their ad inventory, and Google would pit
advertisers against one another to bid for the spots. Problems
emerged, however, because stations were reluctant to give over
control. Worse, the Google ads went for less than those sold
directly by the stations, and although Google argued that
increased demand would eventually drive up the auction prices,
stations were unwilling to take the chance. Media buyers, for their
part, were reluctant to engage with Google, which refused to
continue the conventional practices of negotiating prices ahead of
time and bundling ads together. After shuttering the business, in
2009, CEO Eric Schmidt attributed its failure to the company’s
inability to measure an ad’s performance on the radio—
something it could do on the web by tracking views and clicks.
The venture cost the company well over $100 million. That’s not a
lot of money in Google’s world; the more important point is that
relatively little useful learning seems to have occurred. The
chasm between Google’s core business and the radio business
proved just too great.
5/29/23 下午 11:07⻚码：12/19
It’s wise to minimize the number of uncertainties that need to be
resolved at any particular decision point. One way to do this is
through what Chris Zook, of Bain, calls adjacencies. For example,
you can introduce an existing product in a new market: IKEA sells
essentially the same furniture in many different countries. You
can offer your customers a new but related product: Wells Fargo
has had a lot of success cross-selling. Or you can build a new
business on the foundation of an existing capability: Air Products
and Chemicals has done this with its plant management
capabilities. The point is to learn from failure (and leverage
success) in areas that are fairly close to your established activities.
Zook says that the number of major uncertainties should be
exactly one. That’s a little extreme. I suggest limiting major
uncertainties to those that relate either to the market (pricing,
acceptance, form factor, and so forth) or to technology and
capability issues (standards, scalability, availability of talent, and
so forth)—not taking on uncertainties in both dimensions at once.
5/29/23 下午 11:07⻚码：13/19
Another way to experiment without going too far afield is to break
a long-term project up into smaller pieces. Consider the
commercialization of nanotechnology: Eventually we’ll be able to
construct objects at the level of individual molecules, which will
be a truly revolutionary change. But that future is likely to be a
long time coming. So for the time being, how are we using
nanotechnology? Think wrinkle-free Dockers pants. Think cell
phone displays that don’t show fingerprints. Those more modest
projects make a lot of sense: They apply brand-new technology to
familiar products, which fosters learning.
Principle 6: Build a culture that celebrates intelligent failure.
People often fear that their career prospects will be in trouble if
something goes wrong on their watch. (And, of course, they’re
often right!) Senior managers need to create a climate that
5/29/23 下午 11:07⻚码：14/19
encourages intelligent risk taking and doesn’t punish any failures
that result. Some companies have found it useful to codify this
principle. (See the sidebar “We Won’t Punish Failure.”)
We Won’t Punish Failure
Here are one UK-based global retailer’s formalized rules
for when failure is acceptable. The effort involves
genuine …
This is an area where CEOs can show strong leadership. A.G.
Lafley made fearlessness in the face of failure a core tenet of his
time at Procter & Gamble. He said repeatedly that a very high
success rate is a sign of incremental innovation, and that he was
looking for breakthroughs instead. In his book The Game-
Changer, published while he was still CEO, he lists and even
celebrates his 11 most expensive product failures, focusing on
what the company learned from each. The reasons he gives for the
failures range from “required significant consumer habit change”
(an at-home dry-cleaning kit) to “small idea” (several new laundry
detergents).
During performance reviews onedivision leader would say, “Show meyour scrap heap.” All high achievers
!!
5/29/23 下午 11:07⻚码：15/19
try some things that don’t work out.
That kind of culture building should happen at all levels of the
organization. One senior division head I worked with would say to
his team members during their performance reviews, “Show me
your scrap heap.” The request perfectly conveys the idea that high
achievers will, of necessity, try some things that don’t work out.
Principle 7: Codify and share what you learn.
An intelligent failure whose lessons are not shared is worth far
less than one that teaches something to the group or, ideally, the
whole organization. There are many ways to capture and transfer
learning. Among the most popular are mini postmortems as a
project proceeds, checkpoint reviews as key thresholds are
reached, and after-action review meetings at the project’s
conclusion. In each case the point is to identify what the
assumptions were going in, what happened, what that implies for
those assumptions, and what should be done next. It is critical to
avoid finger-pointing—restraint that is easier to exercise when
the underlying ideas are labeled “assumptions” rather than
“projections” or “data.”
I recently facilitated a postmortem for a large organization
struggling with an IT implementation that had gone dreadfully
wrong. Before we convened, I interviewed key decision makers
and developed a time line showing when critical decisions had
been made. We kicked off the meeting with some general
observations about why IT systems often go awry; the message
was “You are not alone.” Next we discussed the core assumptions
5/29/23 下午 11:07⻚码：16/19
that had been in place when the project was authorized, some
four years earlier; these came as a surprise to the newer members
of the team. We then walked through five decisions that had made
a big impact on the project’s evolution, discussing the
assumptions that were held at the time, what we would have done
differently, and what had been learned. The day ended with two
breakout sessions: one to determine what to do about the current
situation, and one to crystallize lessons that could be valuable in
other projects and help avoid similar problems in the future. To
make sure the learning was transferred, we charged specific
individuals with documenting and communicating those lessons.
Let’s come back to the point I made at the outset: In an uncertain
and volatile world, avoiding failure is not an option. If you accept
this premise, the choice before you is simple: Continue to use
practices that limit what you can gain from failures—or embrace
the concept of intelligent failure, in which learning can create
substantial value.
How to Walk Away from a Project Intelligently
Even companies with a highly disciplined process for
beginning new projects seldom have a good one for
getting out. …
!!
5/29/23 下午 11:07⻚码：17/19
The example set by senior management is crucial. Leaders must
be willing to talk about failures and what was learned from them.
I’ve seen organizations use symbolic rituals to celebrate a failure
that taught important lessons; this can create an environment in
which failures are discussable. Making the ground rules for risk
taking explicit, whether in a contract or by other means, can be
useful as well. Telling stories about failures past can make people
more comfortable talking about failures in progress. And having
graceful ways to shut down initiatives and move on makes the
inevitable failures much more palatable. Fumbling toward
success by learning from failure will differentiate firms that can
thrive during uncertainty from those that cannot.
A version of this article appeared in the April 2011 issue of Harvard BusinessReview.
Rita McGrath is a Professor at ColumbiaBusiness School and a globally recognizedexpert on strategy in uncertain and volatileenvironments. She is the author of The End ofCompetitive Advantage (Harvard BusinessReview Press), and most recently, SeeingAround Corners (Houghton Mifflin Harcourt).
Recommended For You
Why the Fail-Fast Approach Isn't Right for Breakthrough Ventures
5/29/23 下午 11:07⻚码：18/19
Increase Your Return on Failure
Design Thinking Is Fundamentally Conservative and Preserves theStatus Quo
PODCASTThe “Jobs to be Done” Theory of Innovation
5/29/23 下午 11:07⻚码：19/19
