Good Companies Are Built Through Iteration: PayPal Changed Direction Four Times in Ten Months

Deep thoughts on AI and aspirations —— ByteDance Deep Thinking Circle

Today’s PayPal processes nearly a trillion dollars in payments annually and has a market cap of over two hundred billion dollars. Rewind to 1999, and it was called FieldLink, working on encryption technology for mobile devices. Over the following ten months, the company changed direction four times: mobile cash, PalmPilot payments, email payments, and focusing on eBay sellers. The fifth pivot came about six months later, changing the business model. Without these five changes, nothing that followed would have happened.

The people involved have told this story many times, and most people’s reaction remains the same: that’s PayPal, we’re just an ordinary company. The subtext is that pivoting is a remedial move for failures, while successful companies stick to one great idea no matter what. The reality is exactly the opposite: cases of thinking up an idea, charging ahead full speed, and achieving sustained success are virtually impossible to find in the real startup environment. Pivoting is a standard move for early-stage companies—it’s the norm.

First, Let’s Lay Out the Five Pivots

Looking closely at this table, there’s a counterintuitive finding: of the five pivots, only leaving the encryption platform was truly due to “product failure.” The subsequent ones weren’t remedial—they were three different types of corrections. The second was a speed correction; mobile cash wasn’t necessarily the wrong direction, but it required three years to deliver, and the company didn’t have three years. The third was a scenario correction. The fourth was a structural correction. These two are worth expanding on.

The Best Pivots Happen Before the Data Comes In

The Palm episode best illustrates this point. At the time, the email payment idea was already on the table, but the team’s attachment was to Palm: the engineering technology was proprietary, the Series A pitch was “this technology is hard to replicate,” and Palm itself promised to help with marketing. Everyone vaguely felt email payments might be more useful, yet thought it was too simple, anyone could build it, and it couldn’t become something big.

How was this resolved? Through a thought exercise. Someone proposed a typical use case: friends at a restaurant splitting the bill, transferring money to each other using PDAs. Following the scenario through revealed the flaw—even in Silicon Valley, at any random restaurant in Palo Alto, you couldn’t find a single Palm user at a table. This use case died on paper, never even making it to product launch for real data.

What I’ve observed is that most teams get this sequence backwards. Everyone treats “let the data speak” as a decision-making virtue, and as a result, a direction that could be ruled out through reasoning alone ends up taking three months to develop, launch, and obtain dismal retention curves before anyone will admit it. Data is lagging, reasoning is leading. For early-stage companies, the most expensive resource is time. Spending validation costs on a hypothesis that can be disproven by thinking is the most common waste.

The signal for the fourth pivot was even more subtle. PayPal was using an aggressive customer acquisition method: refer a new user, both parties get $10. Everyone said this was insane, but the math was clear: at the time, acquiring a customer through portal site ads cost over $40, while PayPal’s actual acquisition cost was only $12. The real surprise came in user composition—the influx was mainly eBay sellers, who even used referral bonuses as promotional tools, letting buyers register and take goods for free. The board’s initial reaction was confusion: who are these people? The crucial follow-up reaction was: wait, this is the only place we’re growing.

Planned growth only validates what you already believe; unexpected growth tells you what the market actually wants. The former is confirmation, the latter is information. Unfortunately, most people’s first reaction to unexpected growth is to explain it, rationalize it, and stuff it back into the original story, rather than admitting the original story was wrong.

The Hardest Thing to Give Up Is What Hasn’t Failed Yet

The closing move of the fourth pivot was the most ruthless cut in the entire story: completely shutting down the Palm product.

Note that Palm wasn’t a failing business. It had elegant technology, was hard to replicate, had exclusive partnerships—it was the company’s differentiating asset. Even Reid Hoffman disagreed with shutting it down outright at the time, arguing they should sell the technology for cash. Peter Thiel’s response was straightforward: you’re betting on where the growth market is, we can’t raise enough money to make this technology more impactful, we must shut it down.

This left a frequently quoted judgment: effort doesn’t equal value. Value is what others are willing to pay for. No matter how unique or clever something is, if no one will pay more than a dollar for it, it has no value. How much emotion and work hours you’ve invested is a separate ledger from what an asset is worth.

I’d add another layer. What’s truly hard to give up is never a failing business—everyone can see the death date of failing businesses. The hard part is that “not yet failed, but you know where the ceiling is” business: it’s still producing, still carrying sunk costs, still part of the funding story. This kind of business is the most comfortable and most dangerous, because every day it provides new reasons to “wait and see.”

Bill Gates did a more extreme version in the nineties. Microsoft was working on an online service codenamed Blackbird to compete with AOL. After seeing that the internet was the future, he directly had engineers’ hard drives wiped. The logic was simple: he knew engineers couldn’t resist finishing interesting problems at hand, and rather than relying on self-discipline, he physically cut off the escape route.

For early-stage companies, this is almost an iron law: if you don’t kill the old business, the transition won’t be clean. Leave one escape route, and the team’s attention will always have an escape route to retreat to. Half-hearted pivots are most likely to fail because they try to do both and do neither thoroughly.

Don’t Copy Survivor Stories Into Methodologies

A bucket of cold water is needed here.

Each of these five pivots could have killed the company. Surviving all five required both execution and luck in equal measure. The sample size is one. Copying PayPal’s roadmap as your own methodology is no different from buying lottery tickets with someone else’s winning numbers.

There’s also a boundary worth drawing. In PayPal’s five pivots, the underlying capabilities never changed: from encryption technology to payment transfers, what the team was good at and the company’s direction were continuous. If an adjustment requires changing even “what we’re good at,” that’s not a pivot, that’s a restart, and restart accounting follows a different algorithm. To judge what kind of adjustment it is, just ask: are we changing “who to sell to and in what scenario,” or changing the underlying capabilities as well? The former can be fast and frequent; the latter, you don’t get many chances in a lifetime.

Operationally, early teams should schedule course corrections rather than waiting for crises to force changes. Many of PayPal’s adjustments happened at board meetings, discovering the need for adjustment every meeting or two. The mechanism isn’t complex: first write down what you believe is the winning formula in a few sentences, then periodically answer only one question: has confidence in this formula increased or decreased? If it’s decreased, adjust—don’t wait until the data looks bad or the bank account is empty before acting.

As for mid-stage companies that are stuck, what often needs examining isn’t “should we pivot,” but “dare we shut down the old completely?” Wrong directions can be changed. The reason for not being able to change usually lies in your own hands: a business that hasn’t failed yet, sunk costs you can’t let go of, a team that still wants to wait and see. The truly valuable part of PayPal’s story is here—what it demonstrates has never been about how many times to pivot, but that each time they could pivot and pivot cleanly.