When Thomas Edison was asked if he regretted all his failed tests on light bulbs, he famously responded, “I have not failed. I’ve just found 10,000 ways that won’t work.” Most people find it hard to be quite so sanguine as Mr. Edison about failure. It goes against human nature.
Ed Catmull, the co-founder of Pixar, explains the psychology: “One of the things about failure is that it’s asymmetrical with respect to time. When you look back and see failure, you say, ‘It made me what I am!’ But looking forward, you think, ‘I don’t know what is going to happen, and I don’t want to fail.’ The difficulty is that when you’re running an experiment, it’s forward-looking. We have to try extra hard to make it safe to fail.”1
Making it safe to fail is crucial because learning happens through experimentation, and experimentation often results in failure. Recent McKinsey research shows, in fact, that respondents at successful organizations are more than twice as likely as their peers elsewhere to strongly agree that employees are rewarded for taking risks of an appropriate level.2 So a willingness to fail has to be embedded in a company’s culture. Many companies have gotten good at saying that it’s safe to fail, but for most people that’s not enough. Human nature being what it is, risk aversion (even fear) will rule people’s actions if they don’t actually believe it’s safe to fail.
At one telecom company, for example, a team working with an outside vendor passed what it felt was a tricky decision to its manager, who referred it all the way to the CEO’s desk. The CEO then called the CEO of the vendor to try to resolve the issue. This game of “pass the buck” happened four times as lower-level teams and managers were unwilling to make decisions, fearful they might be the wrong ones.
Companies have to put in place various “safety nets” to allow a culture of experimentation to take hold. Here are a few ways we’ve seen companies successfully combat this kind of fear:
Tech designed to make it safe to try
Technology can minimize the consequences of failing, which can often inhibit people from trying something new. Advances in tech have, in fact, made failure cheap, fast, and reversible. Automated software testing, for example, means code failures can be spotted in seconds, allowing fixes to be made without putting significant parts of the business at risk. The most sophisticated companies can even roll back website changes to fix issues with a single system command. Performance tests can measure the system under load or stress, and security tests measure its resilience against malicious attacks.
One international consumer-goods company migrated its software to the cloud so it could give developers self-service access to production-like environments for testing and delegate the execution of thousands of automated tests to “virtual servers” that take seconds to run.
Pharma companies, meanwhile, are turning to advanced analytics and machine learning to scale the number of tests on new drugs and the quality of the results, reducing the chances of costly wrong turns and speeding drug development. Some high-volume testing (for microbial detection and water sterility, for example) can be performed online rather than in physical labs, through automation. Moving to instantaneous microbial detection for environmental monitoring can reduce overall lab lead time by 40–75 percent.3
Like so many things in a digital transformation, change starts at the top. You won’t make much progress encouraging people to take risks if the head of the business doesn’t show it’s really safe to do so. Sam Yagan, ShopRunner CEO, for example, has asked executives in their reviews to describe recent failures. If the failures hadn’t cost the company money, the executives didn’t get their bonuses.4
You need to constantly talk about executive failures, even CEO failures. When I was CEO at Match, we acquired a company, and I think I overpaid for it by $50 million. I got up in front of everyone and talked about that—and people were able to see that I was still there. So people could say, ‘If he made a $50 million mistake, maybe that’s not so bad. Maybe I can make a $50,000 mistake.’
Microsoft’s CEO Satya Nadella famously didn’t rant following a high-profile PR debacle when Tay, a bot that had been programmed to interact with Twitter users, was taught by some of those users to spew racist, misogynistic, and pornographic text. Besides grounding the bot and expressing regret for the unintended offense, Nadella also emailed the team that developed Tay, eager to make sure it didn’t regret taking the risks they took. “You’ve got to make sure that if you make mistakes, you learn from [them],” Nadella said.5 The CEO at a leading retailer took a more active approach to protect his employees so they knew they’d be safe. He made sure that the lead on any project got the credit when an experiment succeeded, but took the blame himself when it failed.
Accepting failure is important as well in encouraging fast decision-making, a key capability in digital companies. The mentality is that a bad decision is better than no decision. Company leaders can lend a helping hand here too. The head of one European retail company, for example, mandated that decisions be made within 30 minutes of the start of any meeting. Another CEO wanted teams to make more of their own decisions rather than escalating them higher up in the organization and delaying progress. His solution? Hold escalation meetings at 11:00 on Saturday mornings. Not surprisingly, teams started making more decisions for themselves.
“Most decisions,” Amazon’s CEO Jeff Bezos wrote in a 2017 letter to shareholders, “should probably be made with somewhere around 70 percent of the information you wish you had. If you wait for 90 percent, in most cases you’re probably being slow.”6
Clear process and information access
There’s another culprit that can undermine efforts to create an environment where it’s safe to fail: uncertainty. Putting in place guidelines and processes—and clearly communicating them—helps reduce uncertainty, which in turn provides people with greater confidence to act. Ensuring there is clarity over roles and responsibilities, for example, is absolutely crucial to avoid painful, time-wasting uncertainties.
We aim to make mistakes faster than anyone else.
Access to information also has an important effect in reducing uncertainty. Equipping call-center employees with real-time analysis on account profiles, for example, or data on usage and profitability, empowers them to modify offers in real time as they have the information to gauge their decisions. In the retail and hospitality industries, companies give frontline employees information such as purchase history so they have the confidence to resolve customer issues on the spot, without having to escalate it to management.7
One oil and gas company automated the evaluation of several business cases so that it had continuous hard data on how the experiments were going, which reduced anxiety among executives about the firm’s digital investment decisions. Indeed, the insights gave them confidence to push ahead with more cutting-edge solutions because they knew they could see in near real time what worked, while failures could be detected quickly and do so relatively cheaply.8
Google made a conscious effort to destigmatize failure after it discovered that when employees felt they could take risks without being shamed or criticized for failure, they performed better.9 So it introduced a process for when any incident is recorded, the team asks questions such as “What went well?” and “Where did we get lucky?” to understand how processes could be improved rather than to find fault. Team leads admit their own mistakes and failures and are willing to document them. Teams are discouraged from blaming anyone and reminded that no one will be punished.10
Food for thought
- The CEO has the key role in creating an environment where people understand they won’t be punished for failing.
- A corporate culture that accepts failure encourages fast decision-making.
- Reducing uncertainties and providing people with access to good information give employees the confidence to make more of their own decisions.
- Automation and analytics have advanced to the point where “failing fast” is a real option for specific cases, especially testing.
- Failure is only acceptable if you learn from it and act on what you learned.