Rocket scientists. Billion-dollar founders. They still missed it.
Every team here was full of very smart people. Some failed outright. Others survived and had to rebuild how they worked. They’re listed in order of how directly each points to the systems a company should set up early.
Fast
Fast raised about $120 million, including a $102 million round led by Stripe. It announced in April 2022 that it was shutting down after failing to raise another round.
The company hired hundreds of people. The Information reported its main checkout product brought in about $600,000 in 2021 revenue, and TechCrunch reported cash burn as high as $10 million a month.
Tie hiring to milestones the business has actually hit, and hold a regular review that puts spending next to results.
Source: The Information, TechCrunch and NPR coverage, April 2022.
Uber
After former engineer Susan Fowler described harassment and retaliation, the law firm Covington & Burling investigated. Uber’s board unanimously approved its recommendations.
The recommendations pointed to gaps in complaint handling, HR records and tracking, and board oversight. They also called for reallocating some of the CEO’s responsibilities and using the existing COO search to bring in a leader to oversee day-to-day operations and help carry out the reforms.
Growth doesn’t build the basics for you. Someone has to own complaint handling, records and oversight.
Source: Uber board statement and Covington & Burling recommendations, June 2017.
Toyota
Toyota recalled millions of vehicles, and its president testified before Congress in February 2010. In March, Toyota announced that regional safety executives would take part in global recall and safety decisions.
Akio Toyoda testified that growth had outpaced the development of Toyota’s people and organization, and that recall decisions had lacked the customer’s perspective.
Growth that outruns your systems shows up later as a crisis.
Source: Akio Toyoda’s testimony to the House Oversight Committee, February 24, 2010; Toyota announcement, March 30, 2010.
Quibi
Quibi raised $1.75 billion, launched in April 2020 and announced on October 21, 2020 that it was shutting down.
The product was built on the premise that people would watch short shows on their phones while on the go. It launched during a pandemic, when people were home, and at launch it had no way to watch on a TV. Founders Jeffrey Katzenberg and Meg Whitman wrote that it failed either because the idea wasn’t strong enough or because of the timing.
Name the assumptions the whole plan depends on, test them before the big spend, and be ready to change course when the world changes.
Source: Quibi’s shutdown letter and Los Angeles Times coverage, October 2020.
Amazon Fire Phone
Amazon launched the Fire Phone in 2014 and disclosed an estimated $170 million in charges, mainly for inventory and supplier commitments. Fast Company later reported that its hardware group, Lab126, was restructured.
Based on interviews with dozens of current and former employees, Fast Company reported that Jeff Bezos drove the phone’s 3-D feature and that team members questioned its value to customers.
Give the product team a real way to challenge a major feature before the company commits to it.
Source: Amazon Q3 2014 filing; Fast Company, January 2015.
Space Shuttle Columbia
On February 1, 2003, Columbia broke apart during reentry and seven crew members were lost. Foam from the external tank had damaged the left wing during launch.
Foam shedding had happened on earlier flights and had come to be accepted. NASA missed chances to get images of the wing during the mission. The investigation board found that the management practices overseeing the Shuttle Program were as much a cause as the foam strike.
Treat a repeated “small” problem as a warning, not a known quantity, and have leaders go looking for bad news instead of waiting for it.
Source: Columbia Accident Investigation Board Report, Volume I, August 2003.
Space Shuttle Challenger
On January 28, 1986, Challenger broke apart 73 seconds after launch and seven crew members were lost. An O-ring seal in the aft field joint of the right solid rocket booster failed.
Morton Thiokol first recommended against launching below 53°F. Its managers reversed that recommendation while engineers kept objecting. The Rogers Commission found that the NASA officials who made the final call didn’t know about the first recommendation, the continuing objections or the joint’s recent history of problems.
Build a clear path for a technical “no” to reach the people making the decision, and expect the people closest to the hardware to push back.
Source: Report of the Presidential Commission on the Space Shuttle Challenger Accident, 1986.
Mars Climate Orbiter
After a nine-month trip, the spacecraft came in far too low at Mars and was lost on September 23, 1999.
Ground software supplied thruster data in pound-force seconds, while the interface called for newton-seconds. The error built up in the trajectory estimate for months. The investigation board also cited inconsistent communication and training and incomplete end-to-end checks of the navigation software.
Give every handoff between teams a named owner, and check that what one team sends matches what the other expects before anyone relies on it.
Source: Mars Climate Orbiter Mishap Investigation Board, Phase I Report, November 1999.
Boeing Starliner
On its December 2019 uncrewed test flight, a mission timer error kept Starliner from reaching its planned orbit, and it never docked with the space station. A separate software error affected service module disposal.
A joint NASA-Boeing review made 80 recommendations. The largest group, 35, covered process and operational improvements. A separate testing group added a full end-to-end test before every flight.
The problem usually isn’t one bad line of code. It’s the process that let it through.
Source: NASA flight summary; NASA review briefing coverage, July 2020.
Netflix
Netflix raised prices on its DVD plans and announced it would spin DVDs off into a separate service called Qwikster. Its U.S. subscriber count fell by about 800,000 in the third quarter, and it canceled Qwikster within weeks.
Netflix said the backlash came mainly from the DVD pricing changes and, to a lesser degree, from the Qwikster plan. Reed Hastings wrote, “I slid into arrogance based upon past success.” After a 2014 price increase, Netflix gave existing members two years at their old price.
Being right before doesn’t make you right now. Keep checking.
Source: Netflix shareholder letter, Q3 2011; Hastings letter, September 2011; Netflix annual filing.
SpaceX
Three Falcon 1 launches failed to reach orbit. The fourth reached orbit in September 2008. Musk later said he had virtually no money left and that a fourth failure would have been the end.
On Flight 3, leftover thrust from the first stage made it hit the second stage after separation. For Flight 4, SpaceX increased the delay between engine shutdown and stage separation. Musk has described the early failures as design errors.
Find the real cause quickly, fix it and verify the fix before the next attempt.
Source: SpaceX Flight 3 mission summary; Musk interviews.
Summaries are based on official reports, public statements and news coverage.
None of these teams lacked talent.
In most of these cases, the warning existed somewhere in the organization. It didn’t reach the right person, or it reached them and got explained away. That’s the part we look for.
Find the problems before your investors do.
Every engagement starts with a short conversation and a fixed quote.