A Team, Not a Family

In 2009, Netflix published a 125-slide deck on how it ran its culture. Sheryl Sandberg called it the most important document ever to come out of Silicon Valley. It defined a performance-filter culture in unusually blunt terms — and defined “performance” with no metric at all.

What traveled, and what stayed home

I went looking for that language on other companies’ careers pages — embedding each Netflix idea and its paraphrases, then flagging any company sentence that came close. Most of the deck never left the building. But a few ideas did, and the lineage is legible: Coinbase lifted the severance line almost word-for-word; Meta turned the “no rules” ethos into “we have values, not rules.”

The first view sorts the deck’s ideas by how closely anyone else’s language matches. Each concept leads with Netflix’s original line; where a company borrowed it above the validated threshold, that lift sits beside it, bold. Underneath, in lighter type, is the adjacent language — the softer echoes that gesture at the same idea without quoting it. The height of that stack is the volume of the ambient version; the faintness is how loosely it matches.

The pattern is stark. The harsh mechanics — the keeper test, “a team, not a family,” context-not-control — stayed Netflix’s own, with almost nothing echoing them. What diffused widely was the softer material: “we don’t have rules” and “dream team” language turns up across eight or nine companies, but mostly in the faint, generic register — “an amazing team,” “warm inclusive people.” “Talent density” itself never traveled verbatim; Stripe’s “talent radar” and Shopify being “intentional about talent” are as close as anyone comes. The sharp idea decays into ambient HR-boilerplate. (Phrases like “raise the bar” and “only the best” show up everywhere too, but those are industry convergence — Amazon’s, not Netflix’s — so I leave them out.)

The second view asks a different question: when did each borrowing and echo first surface? Plotting them by year against a continuous snapshot record, the chronology guards against the obvious overclaim. Some adjacent language predates the 2009 deck entirely — Amazon was already writing about deciding “based on our core values” in 2007 — so this is convergence, not a clean chain of copying. The one unmistakable lift, Coinbase’s severance line, lands late, in 2024.

And there’s a bend in the timing worth naming. Reading the corpus as a whole — the full performance-language axis across all fourteen companies, not just these ten concepts — this language softened in 2023 and returned to its early-decade highs in 2024, holding there through 2025. It’s a modest swing, not a surge: 2024 recovers the ground of the 2021 peak rather than breaking new. But the shape is suggestive, because of when it lands — just after the 2022–23 tech layoffs and the end of cheap money. Copy follows the market rather than leading it, and blunt performance-filter language becomes sayable again once the people it describes have less room to object.

The scoreboard that isn’t there

All of it borrows the language of objective performance, and none of it defines how that performance is measured. Across roughly 2,600 culture chunks in this corpus, the copy invokes merit again and again — and not once names a metric.

The deck’s most-quoted line makes the borrowed authority explicit: “We’re a team, not a family. We’re like a pro sports team, not a kid’s recreational team.” The coach’s job, at every level, is to hire, develop, and cut smartly so there’s a star in every position. It’s a vivid frame — and it smuggles in an assumption that doesn’t survive contact with the work.

A pro basketball player has a regulation goal: get the same ball into the same hoop, on the same court, as many times as possible in 48 minutes. Points scored, points prevented — the scoreboard is right there. But how does a company decide it has the “best” talent? How do you measure a software engineer on an infrastructure team against one in customer support, or one in user experience? Different goals, different surfaces, no shared unit.

By Netflix’s own admission, it largely doesn’t. The system rests on the keeper test — would a manager fight to keep this person if they tried to leave? That is a manager’s gut with a name, exactly the room for subjectivity the sports metaphor was supposed to remove, and the discretion that fills the gap is where bias enters. Netflix’s deck even refuses to keep score — “we have no bell curves or rankings or quotas such as ‘cut the bottom 10% every year’” — while still demanding you be a “high performer.”

None of this means performance isn’t real, or can never be measured. It means the companies borrowing this language while claiming to be perfectly objective about it are, mostly, flattering themselves. And it shows up in who actually gets cut: for many tech workers, a performance conversation is more likely to follow a budget cut than a missed deadline.