Jeff Bezos and the Regret Minimisation Framework Regret Minimisation

The walk in Central Park

In the spring of 1994 Jeff Bezos was a thirty-year-old senior vice president at D.E. Shaw & Co., a New York hedge fund that treated the stock market as a computer science problem. He was well paid, well regarded, and comfortable. He was also quietly obsessed with a single data point. Internet usage, he had read, was growing at 2,300 percent a year.

He took the idea to his boss, David Shaw. The two of them walked in Central Park for roughly two hours. Shaw, by Bezos's own account, was thoughtful rather than dismissive. He agreed the online commerce opportunity was real. He also pointed out, reasonably, that it would be a better idea for someone who did not already have a good job. He asked Bezos to think it over for a couple of days.

Bezos went home and ran a different calculation.

The framework

He imagined himself at eighty years old, sitting in a chair, looking back at his life. From that vantage point, two choices sat in front of him. He could stay at D.E. Shaw, miss the bonus for the year he quit, and never find out what the internet would become. Or he could try to build an online store, probably fail, and come home.

At eighty, the failed startup barely registered. Not trying registered a lot.

He later called this the regret minimisation framework. The point was not to minimise risk. The point was to minimise the kind of regret that compounds for decades. Most short-term career decisions look terrifying at thirty and vanish entirely at eighty. The ones that look safe often turn out to be the ones you quietly wish you had made differently.

He resigned.

The drive west

Bezos and his then-wife MacKenzie packed a car and drove toward Seattle. He had chosen the Pacific Northwest for a mix of logistical reasons: proximity to a large book wholesaler in Oregon, a smaller state population to limit sales tax exposure, and a deep local pool of software engineers. During the drive he worked on the business plan in the passenger seat while MacKenzie drove.

The first Amazon.com office was the converted garage of a rented house in Bellevue, Washington. Desks were made from doors. The company sold books first because the global catalogue was already digitised, no single incumbent dominated, and the median order was small enough to make the fulfilment economics tractable. The site went live in July 1995.

It grew faster than anyone, including Bezos, had modelled.

Amazon.toast

Amazon went public in 1997 at eighteen dollars a share. By 1999 it was a dot-com totem. The stock peaked above a hundred dollars. Bezos appeared on the cover of Time as Person of the Year.

Then the market remembered how to count.

Between 1999 and 2001 Amazon's share price collapsed from roughly 106 dollars to about 6, a drawdown near 94 percent. Bezos's own paper net worth fell from more than 10 billion dollars to under 2 billion. Analysts wrote the obituary. A famous Barron's cover called it Amazon.bomb. Internally, employees started calling the company Amazon.toast.

Most of the public conversation assumed Amazon would not survive the bust. Many of its contemporaries did not. Pets.com, Webvan and eToys all folded in roughly the same window. Amazon cut about 15 percent of its staff. It raised convertible notes in Europe just before the capital markets window slammed shut, a timing call that in retrospect probably saved the company. It killed unprofitable product lines and pushed each remaining unit toward real unit economics.

And then it kept spending.

The long game

The real battle during the crash was not operational. It was philosophical. Wall Street wanted a quarterly profit. Bezos wanted to keep funding fulfilment centres, software, third-party marketplace infrastructure and, before long, servers nobody outside Amazon was asking for.

He sided with the long game and kept telling shareholders exactly what he was doing. Every annual letter from 1997 onward appended the original 1997 letter in full. Its central claim, repeated for a quarter of a century, was that Amazon would be measured over the long term, that it would keep making bold investment decisions where the probability of success was reasonable, and that it would occasionally be wrong in loud, public ways.

Amazon turned its first profit in the fourth quarter of 2001. The number was small, roughly five million dollars, and it was not the point. The point was the signal: the model worked, the long game was real, and the company would survive to compound.

From there the aperture widened on a schedule that in hindsight looks almost suspiciously confident. Marketplace in 2000. Prime in 2005. Amazon Web Services in 2006. Kindle in 2007.

The accidental cloud

AWS is worth pausing on because it was close to an accident.

In the early 2000s, Amazon's own internal engineering teams were slowed down by their own infrastructure. Every team that wanted to build something had to re-invent storage, compute and messaging. Bezos and a small group of senior engineers pushed the company toward building those pieces once, cleanly, with documented APIs, so that any team could plug in.

Somewhere in that work the obvious second question got asked. If this infrastructure was useful to Amazon's internal teams, would it be useful to external developers too?

AWS launched publicly in 2006 with S3 and EC2. For several years it looked, from the outside, like a quirky side bet by a retailer. Inside, it was quietly becoming the most important thing the company had ever built. By the late 2010s AWS was generating the majority of Amazon's operating profit and funding retail expansion, Prime Video, hardware and, indirectly, Blue Origin.

Blue Origin and the second clock

Bezos had founded Blue Origin, his private space company, in 2000, during Amazon's worst year on record. For most of the next two decades he funded it by selling roughly a billion dollars of Amazon stock a year. The company developed slowly, almost deliberately so. Its tortoise logo and motto, gradatim ferociter, step by step ferociously, were not an ironic gag.

In July 2021 he flew to the edge of space on a New Shepard suborbital flight. The same month he stepped down as Amazon CEO and handed the role to Andy Jassy, the executive who had built AWS. He kept the executive chairman seat, and the time.

Day 1

If regret minimisation is Bezos's framework for life decisions, Day 1 is his framework for running a company. For decades the main Amazon headquarters building was literally called Day 1, and he kept a sign on his door with the phrase. In his 2016 shareholder letter he spelled out what Day 2 meant. Day 2, he wrote, is stasis, followed by irrelevance, followed by excruciating painful decline, followed by death. That is why it is always Day 1.

The practical content of Day 1 is more mundane than the slogan. Obsess over customers rather than competitors, because competitors do not pay the bills. Resist proxies: do not let the process substitute for the outcome it was meant to produce. Embrace external trends quickly, because the world will not wait for your roadmap. Make high-velocity decisions, especially the reversible ones, because speed compounds.

And disagree and commit. When a decision owner picks a direction you disagreed with, say so for the record, then commit fully. The alternative, Bezos has argued, is the slow politics that quietly kill most large companies.

What travels

Most founders will not build a trillion-dollar retailer or a private space company. The parts of Bezos's playbook that travel are smaller and more usable.

The first is the eighty-year-old test. Before any decision that feels existential, ask which option future-you will regret more. Almost every founder in the world has made the wrong version of this call once, and known it immediately.

The second is time horizon arbitrage. Most of the market is optimising for this quarter. If you can genuinely optimise for the next five to seven years, you compete against almost nobody worth the name. The hard part is not seeing further. The hard part is withstanding the quarters in between.

The third is the appendix habit. Bezos reprinting his 1997 letter every year was partly literary and mostly operational. It committed him, in public, to a thesis he could not quietly walk away from when a given quarter got ugly.

In 1994, on a two-hour walk in Central Park, David Shaw told Jeff Bezos that building an online bookstore would be a better idea for someone who did not already have a good job. Bezos agreed. Then he went home and became that person.