Thread by @eriktorenberg

Through Naval, I discovered the term "Curation Businesses" which describe things like Universities, Accelerators, and VC firms

These businesses (think top accelerators, universities, etc) attract the best people in the world to apply for them.

Their alumni then go onto be immensely successful, which improves the flywheel and further attracts the best people at an even bigger scale.

It’s unclear to what extent these institutions actually help the people become more successful beyond the credential that they bring.

Their brand serves as a credential for the people who join, which helps them convince other recruits/customers/investors to get involved.

Alumni are incentivized to promote the credential widely and speak highly of it, as to strengthen their own status accordingly.

Their near blind loyalty is rational: If they invested a lot to get that credential, and it becomes diluted, then they wasted their time and money.

Which is why once you get curation businesses off the ground, they are very difficult to disrupt, because everyone has incentive to keep the whole charade going.

Like money, a credentials is an illusion that everyone who has it or wants it is incentivized to keep perpetuating

To build curation businesses, you need some hack to get the best people to join. After you do, they’ll graduate and then be role models and the fly wheel will begin. Until then, It’s truly fake it til you make it

Hack = align w/ already incredible brands, or have a new approach,

If you’re building a curation business, the main (if not only) moat you have is brand.

Your biggest specific vulnerability is alumni NPS—

If it’s strong, the business is unbeatable. If it’s weak, the entire edifice is at risk.

There are a few ways to have high alumni NPS.

1/ Give someone their start or otherwise be an inflection point. Believe in them before others do and they'll be immensely grateful:

2/ Be legible (cadence, credibility, curriculum) so that people have similar experiences and can bond over them.

Legibility provides social context for other people across cohorts (even years apart) to build meaningful relationships.

3/ Be so curated that every person is legit or provide so much value that the best join

The tradeoff is "signal" vs "utility"

Signal itself has utility, but diminishing returns. You want to increase utility as much as possible without sacrificing signal.

The scalable curation businesses start with signal & then use signal to build utility

First YC cohort was super high signal (in retrospect at least!) & they used that to justify scaling 200x ("statistically, one of the companies can be the next Airbnb!")

One reason why this works is that startups is a power law game, and the most important thing is capturing the winner.

Also, given the best unicorn rate among VCs is 5%, it's unclear what signal means. Thus better to make more good bets than "sure" ones.

Why can’t 100K or 1M students a year attend Harvard?

It's partly true that credential quality is tied to its exclusivity or its size—

But it's also true that credential quality is tied to the magnitude of its biggest outcomes

Keeping same quality bar, Harvard should 10-100x.

Focus too much on utility (get too big), and you kill the signal.

Focus too much on signal itself without bolting on utility, then it becomes like Harvard — people know it has no purpose other than the signal and that itself weakens the signal.

It becomes a bubble.

Certain communities have similar dynamics as nightclubs: You get the VIPs in the door that'll bring everyone else, & everyone else brings the $ (or the data, or the hires) that make the community more valuable for the VIPs. If you do it right, you'll build strong network effects.

Curious to hear ideas on how to build curation businesses. We have very few disruptive new universities, accelerators, or Paypal Mafias, which means we understand very little about how to build curation businesses.

By this message I don't mean that they don't add value. It's just unclear whether signaling is 20% or 50% etc of the value. The ones that succeed will offer the most value separate from their signal.