When you or your team try new things — and not all of them succeed — that’s called experimentation. Learning from experiments is essential for your company’s growth. On the other hand, when you deviate from proven practice because of inattention or lack of training, that’s probably a mistake. It’s critical to know the difference, and to create a workspace where individuals feel psychologically safe to take smart risks.
For this episode of our video series “The New World of Work”, HBR editor in chief Adi Ignatius sat down with Harvard Business School professor Amy Edmondson, an expert in psychological safety and author of the upcoming book Right Kind of Wrong: The Science of Failing Well, to discuss:
- Productive, intelligent ways to fail
- Dangers of not experimenting enough
- Balancing individual employees’ needs with those of the team and organization
Edmondson says that leaders should do a thorough post-mortem after every failure, whether it was productive or not, to ensure that it does not repeat itself. “A failure, even an intelligent failure, in new territory, new discovery, is no longer intelligent the second time it happens.”
“The New World of Work” explores how top-tier executives see the future and how their companies are trying to set themselves up for success. Each week, Ignatius talks to a top leader on LinkedIn Live — previous interviews included Microsoft CEO Satya Nadella and former PepsiCo CEO Indra Nooyi. He also shares an inside look at these conversations —and solicits questions for future discussions — in a newsletter just for HBR subscribers. If you’re a subscriber, you can sign up here.
Amy, welcome to The New World of Work.
Great to be here, Adi. Thanks for having me.
Your book is primarily about failure. I was under the impression that we all understood that failure is noble and not shameful, and provides useful learning lessons. But you’re writing a book that seems to be saying that we need to think hard and maybe differently about failure. What are you trying to accomplish with this book?
I was with you and then I poked around and realized that the truth is many people were still confused about failure. There is a lot of happy talk about failure out there. There’s the virtual mantra of Silicon Valley to fail fast, fail often, failure’s good, let’s learn from failure, let’s have failure parties, let’s have failure resumes and so forth. And the truth is, the future of work will be riddled with failure. We can’t just wish it away, even if we wanted to, we have to work with it.
But I think no one can really take to heart the happy talk about failure unless they have a coherent framework. You could think of it as two camps: the Silicon Valley fail fast, fail often camp. And then the other camp, which is, “I live in the real world, failure’s not an option.” And they’re both right or they’re both partially right, but neither is terribly helpful nor context specific.
So I think the happy talk, when it’s not qualified with a coherent way of making distinctions between the good kind of failure and the not-so-good kind, is possibly more destructive than helpful. It drives the honest conversation underground. It’s important to talk about the kinds of failure for which we really should be welcoming it with open arms and the kinds where we maybe shouldn’t.
I think the best thing you can say about failure is if you have a culture that permits failure, that tolerates failure, it means you’re stretching, you’re pushing, you’re trying to innovate, you’re trying to do things that are difficult. That’s part of the definition of what a digital company is. A digital company experiments frequently and tries and fails, and is able to tolerate failure. I would guess if you talk to most companies, they’d say, “Yeah, we do that. That’s the culture we have. We didn’t used to, but we do that.” So I want to push you a little bit more: the rhetoric, that’s the happy talk, but in reality that’s not really how the world works?
First of all, it’s not how most incentives are set up. I’m not saying uniformly that’s the case, but most of the time, failure is not rewarded in organizations, and people would rather do anything but fail.
And you’re right, maybe a better way to talk about this is not as failure, but as experimentation. We have to be very pro-experimentation. But we have to be pro-smart-experiments. And I think smart failures are the result of smart experiments.
Smart experiments are ones that happen in new territory — honestly, if you can look up the answer, find the recipe, find the blueprint, please do, no need to experiment — new territory in pursuit of a goal that’s consistent with the value proposition of the organization, with a hypothesis you’ve done your homework on and importantly, as small as possible.
Those are the kinds of both experiments and failures we must welcome with open arms. They are discoveries and they allow us to figure out quickly what to try next. But a portion of the book is devoted to best practices for failure-proofing that which can be failure-proofed. The activities, the operations in your company that are in known territory, are ones that should be well set up to make failure extremely rare.
Are there industries that do not tolerate failure? Airline pilots? You don’t really want them to fail. This isn’t a rhetorical question. Are there industries that really don’t tolerate failure? And can you look at them and say, “You actually can get interesting results if you have that kind of policy?”
Let’s start with airlines because clearly none of us want them to be comfortable with failure. Yet I think the reason why airlines have an extraordinary record of success and safety is because they’re willing and able to talk about failure. The failures that they do tolerate happen in the simulator. There’s training, there’s a lot of emphasis on speaking up early to prevent something worse from happening. So their safety record does not come from being intolerant of failure, but rather being intolerant of major accidents.
Therefore, we have to be very tolerant of the reality of human error so that we can catch and correct, we can train, we can allow people to take the kinds of risks and experiments we were just talking about in safe settings like the simulator, not in the execution of the real duties.
But I don’t think it’s possible to describe industries in the way your question implies. I think there’s variation across companies. Pick an industry like fast-moving consumer goods. It’s going to be not that hard to find differences in cultural failure tolerance within those industries across companies. So a more sensible way to put that is that some companies are doing better than others in having a healthy tolerance of intelligent failure.
What does a productive failure look like? You did mention that there are good and bad failures. What’s the difference and how does one try to make sure their failures are the good kind?
In known territory where we have a process or a formula for getting the result we want, it’s best practice to use that process, use that formula and get the result we want. So when a Citibank employee a number of years ago accidentally made a small human error and accidentally wired $800 million to a client that he shouldn’t have, that was a basic unproductive failure. Turns out they were not even able to get the money back. So, not celebrating that kind of failure.
A productive failure is one where we get new and useful knowledge, new knowledge that helps us go forward in creating the kind of value we’re trying to create in our market, for our customers. So we discovered something that we could not have discovered without trying it, without the experiment.
Would you recommend that there be an elaborate postmortem? I think the military is very focused on doing detailed postmortems: what happened, what went wrong, why? Presumably to learn from that and not have it happen again.
It is not the case that a postmortem has to take inordinate amounts of time, but it should be thorough. It should be analytical, and look carefully at the different facets of the failure, to understand accurately what happened and why, for the express purpose of preventing that exact failure from happening ever again. So a failure, even an intelligent failure, in new territory, new discovery, is no longer intelligent the second time it happens.
I want to shift gears a little bit to talk more generally about the workplace. The question really is: Are we OK? You wrote a recent piece in Harvard Business Review that suggested maybe things are not so great, with relatively low levels of engagement and productivity, high rates of burnout. We can speculate as to why that’s true, but is that accurate? It’s hard to generalize, but you know, are we suffering? And if so, how do we respond to that as managers?
Well, I don’t have a kind of systematic worldwide dataset from which I can make solid inferences about how people are doing. My impression comes from informal conversations, qualitative research, reading HBR and so many other outlets, to see how people are doing. So really, in a way, I’m commenting on the conversation in HBR and so many other business media contexts, maybe LinkedIn and elsewhere.
One thing I think I can say for sure is that the anxiety is real, and people are worried about the future. They’re worried about it on so many fronts. They’re worried about climate change. They’re worried about AI. They’re worried about burnout, as you mentioned. I’ll come back to burnout.
But that anxiety tends to push us toward a retreat to our individual corner, and people start to think, “Am I going to be OK?” They become more focused on their own wellbeing than on the health of the team or health of the organization. And that gives rise to a real potential for erosion, even vicious cycles, where organizations find themselves in the trap of responding to requests and issues in isolation, one by one.
We need a more holistic way of thinking about it. And I see limited evidence of companies being at least described as pausing to think about the larger picture, their value proposition, what it implies for how they must be structured and led, to get the necessary work done, and how to organize that work, with all its variety and variable needs, in a thoughtful way, and how to inspire and motivate people to do it well.
Let me just briefly go to the burnout issue, because there actually has been some recent data, some studies that have caught my eye, showing that burnout is systematically higher when psychological safety is lower, right? For instance, it seems to me that some portion of the burnout is associated with loneliness and isolation. I think it’s fair to say that we can endure many challenges when we feel genuinely that we’re in it together, that we’re connected and engaged with our colleagues in trying to sort of navigate these challenges.
One can’t help but think, “Okay, is some of this related to the pandemic?” Which for many of us, broke up teams, created new work environments with work from home, that in many ways is fantastic for people who are balancing their work and life. It must take a toll on maybe the teaming imperative that you’ve written about. Is that your hunch?
I do think the pandemic took a toll on us, on all of us. It created such an obvious uncertainty. It was such an obvious disruption. It wasn’t the gradual shifts that we’re normally used to. It was a very abrupt shift, and it gave rise to wonderful—and I think productive—experiments on different work arrangements.
Now it’s time for a very systematic assessment of what’s working and what isn’t? And it can’t be incremental, right? And it also can’t be based on what people say they want. Because oftentimes, what we say we want is not actually what we need or truly want in the longer term, bigger picture, to get where we need and want to go.
You talked a second ago about trying to have a comprehensive policy and approach that, if I heard you right, is not dealing with people always individually, but that’s sort of the nature of management now. Suddenly, managers are expected to be, in addition to everything else, almost like psychiatrists, that there’s an openness for people to share their personal situations, challenges, problems, and that it’s the role of the manager, increasingly, to engage with that in an intelligent way. So you end up where management becomes hyper-personalized, but I think maybe you’re already on to the risks in that, which is losing the sense of the teaming and the collective effort.
It’s almost as if we’ve lost sight of tensions and trade-offs. There will always be a tension between me and we, right? There will always be a tension between my desires in the moment and my aspirations over the long term. If you ask me what I want: pay me infinitely and don’t ask me to do anything, and let me eat ice cream all day, right? But that’s not going to get me where I really, really need to go, and want to go. I want to make a difference.
I think we’re in a moment of not helping people value the collective. As human beings, we’re social creatures. That’s part of it, but it’s also that we want to matter. We want to matter to others. We want to matter in some way that’s larger than ourselves and our hedonistic desires.
You can think of an old-fashioned management theory of the firm, right? If markets worked by themselves, we would just have only contractors doing tasks, and it would be efficient, it would be sensible, it would be logical. But it doesn’t work, because a lot of the work we have to do is inherently collaborative, and dynamically so. And it isn’t easily parceled out, dividing-and-conquering style. It requires us to really work together in meaningful ways. The good news is, that can be a very engaging, rewarding, exciting experience. The bad news is, it’s not easy to manage.
But I think we can go down that rabbit hole of each person has to be managed differently, each person, you’re almost a psychiatrist to that person, versus let’s step back and rethink, how do we design our activities, our operations, so that we create the most value for those we serve?
Yeah, I love that, and I have to say that I don’t think companies have figured that out yet. The disruption of Covid opened our eyes to some flexibility. But I think the things you’re putting your finger on, we’re trying to solve for that, and I think a lot of us haven’t yet and need to keep experimenting.
So we’re in this age of anxiety, where there’s burnout. And then, you throw on top of that generative AI, and a fear—possibly irrational, possibly not—that generative AI will be able to do all of our jobs at almost no cost. I assume you haven’t done quantitative research. But qualitatively, what’s your advice for people as generative AI enters the workplace at every level and the possibilities become clearer and clearer?
As you indicated, it’s a little outside my wheelhouse, except for the effects on people and culture. I speak from the perspective of someone listening at the margins to the many conversations in work and social gatherings alike, and I think you’re right. I think fear is the dominant emotion, that certainly some are excited, some are super optimistic about the amazing changes to come, but I think casually I hear more fear than optimism.
The truth is we need both. We need some positive, thoughtful, design-oriented approaches to experiment and figure out what’s going to work. But I don’t think they’re going to be simple solutions to the dramatic shakeup of what’s possible.
Here’s a question from Omar from Monterrey, Mexico. What kind of metrics can be used to measure smart failures?
My first response is that it’s a good idea to have metrics. One of the things that I’ve spent the most time studying is how many failures just don’t even get the chance to be measured, because people don’t speak up about them. This was how I got into this topic in the first place: the discovery of dramatic differences across groups, even within the same organization, and their willingness to speak up about things that go wrong rather than just things that go right.
Here’s the challenge more broadly than just people not necessarily speaking up: the category of intelligent failure covers vast territory. I think the metrics have to be tailored to the context — and let me illustrate vast territory. A well-run clinical trial on a new cancer drug is an intelligent failure when it turns out it doesn’t have the efficacy that we hoped. It was in new territory. There was no other way to find out but to do a clinical trial. It’s the right size, it’s no bigger than it has to be. It’s hypothesis-driven in pursuit of a goal.
But so is a really bad blind date. That’s intelligent failure. Maybe your friend thought you’d like each other. You are willing to go out and have a coffee. Smallest possible new territory in pursuit of a goal, all the rest. So a bad blind date and a failed clinical trial are clearly apples and oranges, yet they both qualify under the category.
I think the best way to answer the measurement question is, let’s make sure the criteria are adhered to. And then, let’s think about what the right frequency is, given the work we’re trying to do, of intelligent failures?
Another way to say that is: What’s the right frequency of experimentation? How often should we be trying new things to push the envelope, to discover new possibilities, even to discover efficiencies? And are we doing that often enough? The answer is usually no, because most of us would rather succeed than fail, and most of us would rather keep doing what we’re doing because we’re kind of good at it.
So here’s another question along those lines from Mohammed in Pakistan. Employees may be hesitant to offer feedback that could be perceived as negative, which can impede professional development, hinder organizational progress. How does one tackle this situation?
Such a good question because it’s true. We are very reluctant to do things, to speak up with negative or difficult information, because frankly it will always be easier not to. It will always be easier to hold back than to speak up candidly and forthrightly about something that you hope could be made better.
The way to make this very difficult thing easier is to set the stage by pointing out how valuable it is. Periodically, I would say even frequently, refer to the fact that “We need to do this hard thing. We need to do it well if we want to be as good as we can as a team.”
But even individuals who have the ambition to grow and develop in their roles and in their careers have to train themselves to be willing to do this and receive it because of its value. So we’ve got to call attention to its value. We’ve got to call attention to the fact that it’s hard and then do it anyway and support each other.
This question is from Don from Calgary in Canada. If it’s true that we learn acutely from mistakes, what are some ways to encourage permission from our leaders who may be risk averse?
We’re all risk averse, and maybe leaders even more than others, maybe not. But first of all, I make a distinction between mistakes and failures. I’m not anti-mistake because I’m a human being, and I make them, we all do.
But a mistake is not the same thing as a failure. A failure is something that went wrong that we wish were otherwise. A mistake is a deviation from a known practice. Now, that could happen because of inattention, because of lack of training, because of exhaustion, you name it.
But I think it’s helpful for leaders, and others for that matter, to talk about the reality that we will make mistakes because we’re human. The very best practice is not to never make a mistake. It’s to catch and correct them quickly, and then also to make that distinction between smart experiments in new territory that we also want to see more of because it’s the secret to future value creation. And we welcome those, too.
Furthering that, here’s a question from Benny from California. What’s the best way to speak to subordinates after a failure to boost morale and communicate that, “This was a good failure. It’s OK”?
I’m going to say “honestly.” You can be honest about, “Wow, this was disappointing for all of us, and let’s get everything we can out of it. Let’s learn as much as possible.” And in fact, given that something substantial that goes wrong nearly always has multifaceted aspects to it, it’s helpful to have a thoughtful and data-driven conversation about what happened. Not “Who did it?”, but, “What happened?”
We may go around the team and ask, “What did you see?” And we’re really looking for what happened, what contributed to that, and that’s both commission and omission, things that you did that may have contributed, things that you didn’t do that may have helped. It’s a thoughtful, deliberately learning-oriented conversation designed to help us be better next time.
How do you reenergize your team these days? How do we reenergize our team particularly now, in 2023, where it feels like there’s a lot of stuff swirling around?
It starts with personally taking the time to reconnect with your own sense of purpose for doing the job, the role that you are currently doing. And consider why it matters to you and why what you are doing or leading matters to the world.
Having done that, share it. Share it often and then just as quickly invite others in to help navigate the necessarily stormy waters that lie ahead. I think it starts with you and then it’s an honest sharing of why you care, why it’s challenging, why you very much need and are interdependent with others. Because all of us want to be needed. We want to be needed. We want to matter.
The last few years, with the pandemic, I’d say certainly in the U.S., there’s been increased attention to social issues, which on the one hand I think felt right to people in the workplace. On the other, it brought more challenges into the workplace. One imagines there’s a pendulum, and it might swing between leadership needing to be very empathetic to, I don’t know, the backlash if that’s the right word. Leaders need to achieve productivity. That’s what it’s all about. Do you believe in that pendulum or are we in a different place? And if you do, where are we right now on that swing?
I believe in the pendulum. I believe that the pendulum happens and I believe there may be a better way. It’s often thought of as empathy versus productivity. And I looked this up actually: Productivity is defined as the effectiveness of productive effort as measured in terms of the rate of output per unit of input.
The first problem is that not all work is easily measured for productivity. The second problem is, often it’s not the right way to measure excellence. Productivity is often a short-term measure, and it has limited predictive value for the future performance of the firm. For example, one way to be really productive is to just push people to their limits. But that has time constraints. Eventually they will burn out, leave, etc. It’s like Buckminster Fuller used to say that it was foolish to burn down the house to keep warm on a cold winter’s night. The excessive pressure can be the equivalent of that error.
And also innovation work in particular, we have case study after case study where the work actually suffers when productivity metrics are brought to bear.
In a way, I wish the pendulum were more about excellence than productivity, because I think productivity is really tricky and variable to measure.
I see the pendulum existing, but maybe it’s a false dichotomy. Maybe it’s not empathy versus productivity. Maybe we need smart, caring leaders who understand the importance of both. And given that that’s very challenging, they’re open about it being challenging. They’re asking for help. They’re sharing the burden of caring and excellence with their teams and considering, again, the fundamentals of what it is the organization must do well to stay alive in its market, to perform in its market. And talk about it honestly.
I sometimes think we don’t talk often enough about the fact that work is work. It’s supposed to be a little bit of work, but that doesn’t mean it can’t be fun, energizing, collaborative, and full of empathy.
I love that. Well, that’s a good point to end on. Amy Edmondson, thank you for being on the show.
Thanks for having me. All the best.