What product leaders and founders learn from their biggest failures

Executive overview

Failure is inevitable in product work — but most professionals only share their wins. This compilation surfaces raw, specific stories of failure from senior product leaders across Airbnb, Google, Facebook, Intercom, Quibi, Duolingo, and Toast.

The recurring theme: failure teaches things success cannot. Trust must be earned before change can be driven. Product bets built on fear or magical thinking fail regardless of execution quality. And experiments only generate learning if designed to fail conclusively.

The measure of a failure is not the outcome — it's whether you learned from it and designed the conditions to learn.

Leading through trust

• Katie Dill (Head of Design, Stripe) joined Airbnb and came in "swinging" — implementing change immediately without earning the team's trust.
• One month in, five designers and an HR partner confronted her with a written list of grievances.
• The core feedback: she hadn't brought the team along; they had no sense of shared goals or that she cared about them.
• She shifted to listening first, understanding individual motivations, and moved in the same direction but with the team.
• A few months later, the design team had the highest engagement scores in the company.
• Lesson: you can inflict change on people, but sustainable change requires trust.

When stage fright meets catastrophe

• Paul Adams (CPO, Intercom) froze mid-talk at Cannes — the world's largest advertising festival — in front of thousands.
• He had over-rehearsed word-for-word for the first time, breaking his usual ad-lib style.
• He walked off stage, still mic'd, cursed audibly; the audience laughed.
• He turned around, walked back out, and finished the talk.
• Key insight: even the worst-case scenario is survivable — the trick is to adapt and keep moving.

Fear as a product strategy

• Adams worked on Google Buzz and Google Plus — both driven by competitive fear of Facebook, not genuine user need.
• His own research showed users needed better small-group communication tools; the product built was the wrong solution to the right problem.
• Google Plus was run with extreme secrecy and internal antagonism, which compounded the failure.
• Lesson: products built from fear rather than user insight rarely succeed, regardless of resources.

The arms race that kills

• Tom Conrad (former CPO, Quibi; engineering leader, Pets.com) describes how three overfunded pet e-commerce companies in 1999 each raised $50M+ and outspent each other into collapse.
• None could win an irrational advertising arms race; excess capital led to unwise decisions.
• Chewy later proved the business model was sound — the timing and execution were the problem, not the category.
• Pets.com went from nothing to public company to shutdown in 19 months; unusually, they returned remaining capital to investors rather than spending it all.

When the unit economics are broken

• Quibi raised $2B to build a bespoke mobile content library, betting that premium original content alone could drive subscription and retention.
• They produced 70 shows in 18 months — more than all major broadcast networks combined in a year.
• Conrad's insight: no amount of good execution can fix a broken foundational equation. The math required $6-10B, not $2B.
• Launched two weeks into COVID, preventing production of the live daily content that distinguished them from YouTube.
• Lesson: identify whether the company equation is fundamentally broken before optimising execution.

Failing conclusively

• Sri Bachu (former Head of Growth, Ramp) argues that failure is not failing to drive revenue — it's failing to learn.
• Growth experiments have a ~30% success rate; the majority of things tried don't work.
• The failure mode to avoid: inconclusive experiments that get repeated every time a new executive arrives with the same idea.
• To fail conclusively in B2B (where sample sizes are small): maximise the treatment effect — throw all plausible tactics at the hypothesis at once.
• If the strongest possible version of an experiment fails, you can confidently kill the idea.
• If it works, run a follow-up stripping out the least effective tactics to find what drives the result.

Jumping to solutions

• Jay Z (former Airbnb, Head of Product at Webflow) identifies solution attachment as the hardest habit to untrain in new PMs.
• Airbnb Plus launched to address a trust gap in the platform — but was built solution-first, driven partly by fear of managed marketplace competitors.
• The right problem: users needed confidence in what they were getting. The solution chosen: physical inspections.
• Inspections didn't match Airbnb's operational strengths and the unit economics never worked.
• Cheaper, more scalable alternatives existed: lockboxes, local cleaning partnerships, better use of existing guest reviews.
• Lesson: validate unit economics early; match solutions to the company's actual capabilities.

The A side and the B side

• Gina Gotthelf (early growth leader, Duolingo; COO, Latitude) coined the "A side / B side" framing: most professionals only share their highlights.
• Her own B side includes: dropping out of college due to depression, applying to 100 companies and hearing back from almost none, losing her work visa twice, being laid off multiple times, going six months without pay at Tumblr.
• She entered Duolingo as a consultant helping them grow in Brazil; they expanded her remit country by country until she was running global growth.
• Duolingo mistakes she was involved in: a failed social dueling feature, a schools platform that didn't gain traction, a China launch that hit 1M downloads then got government-blocked, an India launch that served the wrong base language.
• Dogfooding failure: the growth team delayed running a badges experiment for six months, then ran the most minimal possible version — a sign-up badge with no social visibility or collection mechanic. It failed. They moved on for eight months. When they eventually returned and dogfooded properly, they recognised immediately that the experiment had been badly designed, not that badges didn't work.
• Badges eventually had a significant positive impact across multiple metrics.
• Lesson: don't let MVP mean "so stripped-down it can't possibly test the real hypothesis."

Never do the rewrite

• Maggie Crowley (VP of Product, Toast) asks every PM interview candidate: "What's the worst product you ever shipped?" — because shipping something bad is a prerequisite for becoming good.
• Her own example: a codebase rewrite estimated at six months that took two and a half years and still wasn't complete.
• Root causes: arrogance, skipped discovery, no one-pager, insufficient technical and design scoping upfront.
• Sunk cost fallacy kept the project alive long past the point of rational continuation.
• Lesson: do not rewrite. If pressed to, scope it exhaustively first. Side-by-side rewrites almost never recover to neutral, let alone positive.