Why Great Product Managers Design for Failure

When Alex Cowan started his first tech company at 19, he didn’t have a grand plan. “I was like a vacuum cleaner salesman,” he says. “I would go door to door and just be like, ‘Hey, does anybody need their computer fixed?’” That entrepreneurial spirit — to see a problem, try something, learn from the result, and try again — now defines his teaching at the University of Virginia Darden School of Business.

Cowan teaches four courses in the Data Analytics and Decision Sciences Area: Digital Product Management, Digital Product Design, Coding with GPT, and the Digital Capstone. His mission is to help future business leaders learn to think experimentally — and interrogate every idea.

“We shouldn't take anything for granted that it's a good idea,” he tells Darden’s Brett Twitty in the latest episode of Office Hours, a faculty spotlight series. “Every idea is bad until it's proven to be good, rather than good until it's proven to be bad.”

Learning to Manage by Experiment

Cowan’s concept of “hypothesis-driven development” underpins much of his teaching and writing. “Everything’s an experiment,” he says.

“The best bet for most people going into a tech business, whether that’s as an entrepreneur, or as a consultant, or building software inside a company, is to treat everything as an experiment and front-load the difficult questions about ‘How do we know if this is working or not?’ And, ‘How do we oblige ourselves to minimize waste?’”

From Fixing Computers to Teaching Management

Cowan came to Darden as a practitioner. “I sold my last company the day I started at Darden,” he says. “That was in 2015.” Before that, he had taught workshops in his native Silicon Valley on product design and management in support of his work on hypothesis-driven development.

And even though he had managed companies, it wasn’t until he joined Darden that he fully understood the concept of general management.

“I thought it meant, ‘I don't like the rental car that I got, I want to speak to the general manager,’” he says. “What it means is that you're training somebody to be a fiduciary of earnings, which sounds boring, but that's basically what I was always doing [as a serial entrepreneur].”

In other words, general managers strike the right balance between revenue and costs to drive earnings. They have what Cowan calls "a coherent economic definition of success" rather than just focusing on one function such as sales or engineering.

His courses, both at Darden and online through Coursera, prepare students for a business world transformed by AI and automation.

“Instagram was 11 people when it sold for a billion dollars,” he says. “Being a good general manager isn’t so much now about managing a giant army of people. It’s about getting to a really good, pertinent idea, testing it, and executing really well with an interdisciplinary team, and amplifying them and making them effective.”

For Cowan, general management means understanding every piece of the puzzle — not being the expert in all of them, but being able to “substantially understand and thoughtfully engage” across various disciplines with your collaborators.

He identifies two traits that set great general managers apart from the rest: Empathy and expertise.

When it comes to digital innovation, Cowan says he tries to foster two skills in his students. “First, is an experimentation mentality, and second is the creative confidence to, where applicable, get into the ‘technical’ details and participate in the process.”

And there’s good news: Cowan sees product management as “the everything job” of the future — a form of general management that can't be automated away. He believes the most valuable professionals will be creatively confident general managers who can take responsibility for economic outcomes across design, coding and analytics, rather than specialists focused on narrow technical tasks.

Designing for Failure

In addition to teaching, Cowan serves as an investor-advisor to startups. He meets with these companies weekly — a middle ground between the “too slow” quarterly board meetings and the intensity of running a company full-time.

One company he advises, Jedburgh Technology, builds language-learning software for the Department of Defense. When the team debated adding a leaderboard to track performance and encourage competition among users, Cowan pushed them to treat it as a testable hypothesis. “You have to say, okay, why are we doing this? If it’s to increase engagement, how do we measure that?”

That creative confidence, he believes, is what distinguishes great product teams. “Good design is testable,” Cowan explains. “That’s the difference between design and art. Design has some kind of commercial objective.”

He adds, “You really have to carry your passion about the user and the success of the product into every little thing that you’re doing,”

What’s more, in his classroom, failure isn’t a bug — it’s the feature.

“In product design, we design for failure,” he says.

That mindset, he emphasizes, isn’t just for startups. Even the world’s largest tech companies live and die by their ability to test and kill ideas quickly.

“Nineteen out of 20 experiments at Google fail — they’re not better than the prevailing alternative,” he says. “They introduce a change, find that it doesn’t help with whatever they were hoping it would help with, and they kill it and try the next thing.”

The implication ripples throughout everything these companies do. "That's a very different thing than making tennis balls, or commercial real estate, or opening up restaurant franchises," Cowan notes.

Coding with AI — and Still Thinking Like a Human

Cowan’s latest course, Coding with GPT, embraces AI as a teaching partner rather than a replacement for human insight. “Every student has the greatest TA on Earth looking over their shoulder, available to answer any question, any time of day or night” he says.

But while the coding part is fast and relatively easy now with the assistance of LLMs, the hard part is still figuring out what to build, why, and how to debug the code, he adds.

Students in the course don’t just write code — they analyze, test and iterate.

The course focuses on four fundamentals: expressing design intent clearly, unpacking work into codable steps, choosing between technical alternatives with testable criteria, and analytical debugging. "That's a great way for a general-purpose coder to figure out, 'Am I good now? Do I know how to code?'" Cowan says of the last step.

In the final class, student teams must pair with someone from outside their team to make a change to their code and push it live — a test of whether they've built something amenable to the kind of change that’s inevitable for any digital product or feature.

“That is a major bellwether that engineering managers will use to see how healthy an individual team is,” says Cowan. For example, when a company hires a new developer, how long does it take that person to not only make their first changes to the code but also become effective? 

“A good product can become a bad product if it becomes encumbered by lots of features that aren’t very relevant,” Cowan says. “So we spend a lot of time making sure the code is intuitive and easy to work with and amenable to change, because you have to expect the unexpected with code.”

The “Cotton Gin Moment” of White-Collar Work

With all the worries about AI taking away white-collar jobs, is there still a future for product managers?

Good news.

“The general job of product management, whatever the title is,  is actually a major growth job because it’s effectively general management, which you can’t automate away,” says Cowan.

“You can’t go to ChatGPT and say, ‘Hey, make me some money,’ and have it deposit money into your bank account. It just doesn't. You still have to ask it the right things, connect the dots      and execute.”

In contrast, he warns, the workers who are at risk of losing their jobs to AI are those who say, ‘’I don't want to have to be responsible for the overall outcome, I just want to do my thing. I just want to be in charge of accounting,’ or ‘I'm just a mid-level developer, I make code, that's my job, my job is code.’ Those are the jobs that are at risk because the AIs are getting better and better at just going and executing those things.”

He calls it “a cotton gin moment for white-collar work.” The automation of repetitive digital tasks, he predicts, will leave behind a premium on creative, interdisciplinary problem-solvers — “creatively confident general managers” who can connect technology, design and economics.

Alex Cowan is the author of the book Hypothesis Driven Development available on Amazon.

To listen to the full conversation with Brett Twitty, visit “Office Hours” with Professor Alex Cowan, Presented by Darden Ideas to Action.