Before Google indexed the web, Larry Page and Sergey Brin tried to sell their PageRank algorithm to Excite in 1999 for $1 million, then dropped the price to $750,000 — Excite CEO George Bell turned them down because the search results were too good and users would leave the site too quickly

In 1999, two Stanford PhD students walked into the offices of Excite — then one of the largest portals on the internet — and offered to sell the technology that would later become Google. According to widely circulated accounts, the asking price was $1 million, later reduced to $750,000, though the exact figures have varied in different retellings. Excite CEO George Bell declined. The reason has become one of the most frequently cited misjudgments in business history: the search results were reportedly too good, and users would find what they wanted and leave the site too quickly.

The story has been recounted in various tech histories and interviews over the years. The exact dollar amounts have been reported differently across sources — sometimes $1.6 million, sometimes $750,000 — but the core narrative remains consistent. Excite passed because PageRank threatened the metric Excite was being valued on: time-on-site.

The metric that ate the company

Portals in 1999 lived and died by stickiness. Excite, Yahoo, Lycos, and AltaVista were all chasing the same thing — eyeballs glued to pages stuffed with weather widgets, horoscopes, stock tickers, and banner ads. A search engine that worked too well was, in that worldview, a leak in the bucket. Every second a user spent finding their answer somewhere else was a second not spent looking at an Excite ad.

Bell wasn’t being stupid. He was being rational inside a flawed frame. His board, his investors, and his ad sales team were all measuring the same thing. PageRank’s superpower — getting users off the page faster — read as a bug, not a feature.

This is what makes the story useful rather than just funny. It’s a textbook case of anchoring on the wrong KPI. The whole industry had agreed that engagement equaled value. Anything that contradicted the agreed metric got filtered out, no matter how obviously revolutionary.

Why smart people miss obvious things

Research on cognitive biases in entrepreneurial decision-making describes exactly this pattern. Status quo bias favors the familiar option. Anchoring locks judgment to an initial frame — in Bell’s case, the frame that portals win by holding attention. Confirmation bias then quietly screens out evidence that the frame is wrong.

Page and Brin walked in with proof that their algorithm returned dramatically better results than Excite’s own engine. Bell reportedly told them that if Excite’s search was that much worse than Google’s, he didn’t want it. The anchor held. The deal died.

A similar dynamic shows up in finance all the time. Analysis of critical thinking in financial decision-making points to the same trap: snap judgments built on the metric in front of you, instead of the metric that actually predicts the outcome. Excite was optimizing for the quarterly ad-impression number. Google was optimizing for whether the product worked.

The math Bell didn’t do

Excite was acquired by @Home Networks in 1999 for approximately $6.7 billion. By 2001, Excite@Home filed for bankruptcy. Its assets were sold for pennies. Google, meanwhile, went public in 2004 at $85 per share. Today, Alphabet stands among the most valuable companies ever assembled.

Run the counterfactual. For $750,000 — a relatively modest sum even by late-1990s standards — Excite could have owned PageRank outright. Forget the company valuation; the keyword auction model that Google later perfected as AdWords would, by itself, have rewritten Excite’s revenue model. The thing Bell was afraid of (users leaving faster) turned out to be the exact behavior that made paid search the most profitable advertising mechanism ever invented. Short queries, fast exits, high intent, high conversion.

Google didn’t lose money when users left quickly. Google made money because they left quickly — and came back frequently.

The pattern keeps repeating

The Excite story isn’t a one-off. Similar narratives have emerged around other major tech companies: Blockbuster’s reported rejection of Netflix, Yahoo’s alleged missed opportunities with both Google and Facebook, and Kodak’s invention of the digital camera in 1975 only to shelve it for fear of cannibalizing film sales.

Each refusal has the same shape. A dominant company evaluates a disruptive technology using the metric that made it dominant. The technology fails that metric — because it’s measuring the wrong thing — and gets rejected. The company that built the technology then goes on to redefine the industry around a new metric the incumbent never saw coming.

This is why the entire field of modern SEO exists. PageRank treated links as votes, and that single idea reorganized the entire web. Anyone running a site today is still working downstream of that 1999 decision — whether they’re building link building strategies or trying to understand how different SEO types affect ranking. The web Excite tried to protect no longer exists. The web Google built replaced it.

What Bell got wrong about engagement

Here’s the deeper miss. Bell thought engagement meant time-on-site. Google understood engagement meant trust. A user who finds what they need in three seconds and leaves doesn’t churn — they come back tomorrow. A user who’s forced to wade through irrelevant results for ten minutes eventually finds a competitor.

The same misreading shows up in workplaces today. A recent Forbes analysis of workplace engagement data from Kahoot! found that nearly half of leaders describe themselves as fully engaged while a third also report burnout — measuring presence instead of meaning. Excite measured presence too. It mistook captivity for loyalty.

Anyone building a digital product right now is making some version of this call. Do you optimize for the metric your board wants this quarter, or the behavior that actually keeps customers around for a decade? Companies running an SEO campaign face this exact tension — quick wins versus durable authority.

The bias under the bias

There’s a layer beneath the metric problem worth naming. Bell wasn’t just defending a KPI. He was defending the version of the internet he’d already built a career inside. The portal model had made him a CEO. PageRank, if it worked, made the portal model obsolete. Accepting Page and Brin’s pitch wasn’t just buying a technology — it was admitting his current product was bad.

That admission is the hardest move in business. It’s why incumbents almost never disrupt themselves. The same pattern shows up in how organizations handle AI bias today — the systems most in need of correction are run by the people whose authority depends on those systems being trusted.

Page and Brin, for their part, didn’t really want to sell. They wanted to finish their PhDs. The $1 million ask was almost a hedge — if someone said yes, fine, they’d go back to Stanford. Excite’s no forced them to actually build the company. In that sense, George Bell may be the most important accidental venture capitalist in tech history. By refusing to buy Google, he funded Google’s existence as an independent company.

The lesson that’s still expensive

The numbers from this story get repeated because they’re shocking — $750,000 for a company that would eventually reach a market capitalization in the trillions. But the shock obscures the practical takeaway. Every company is currently rejecting some version of Page and Brin’s pitch. Someone is walking into a conference room right now with a tool, a process, or an idea that threatens the incumbent metric. The default response is the Excite response: this works too well, it disrupts our funnel, it cannibalizes our model, pass.

The test isn’t whether the new thing fits the old scoreboard. The test is whether the old scoreboard is the one customers actually care about. Bell answered that question wrong in a single meeting in 1999. The cost compounds daily, twenty-seven years later, every time someone types a query into a box and gets the answer in 0.4 seconds.