"We only hire from top schools" is lazy practice that results in less-than-top teams. This thread is a rough attempt to explain why 🧵👇
Quickly about me: for undergrad I went to NAU, the smallest of three state universities in Az. I did my Masters and PhD at Stanford.
At Stanford I TA'd, I co-taught courses, I sat on the masters admission committee in CS as a consulting faculty, and I've been involved in hiring hundreds of folks as a founder, tech exec, and board member.
So I've had broad exposure to admissions, curriculum, students, and workplace performance and my experience is that while top schools produce great students, focusing just on them is a good strategy for building not-truly-exceptional teams ...
Why is that?
The core problem is we build recruiting policies around the mean student. Yet hiring is about the tail.
For example schools often present comparative metrics such as GPA or admission test scores as 25-75% ranges, thus minimizing perceived overlap.
However, populations are distributions. And hiring should be focused on the tail (e.g. upper 10%). Take the following two normal curves with a mean difference of 50%.
While purely illustrative, even when the mean student at a top school is far better than the mean student at a lower tier, the top students in the lower tier are likely better than the mean student at a top tier university.
Further, my experience is that admissions at top schools are far less forgiving of "life getting in the way" or flubbed standardized test scores, so I suspect the distributions actually look more like this:
Again, this is just illustrative, but an important point. Top schools are often similarly blinkered in admissions as companies that only hire from "top schools". So the distributions are likely far narrower. And a lot of great talent ends up at other schools.
It's also worth noting larger schools (e.g. TAMU as ~70k students) have many more students in the tail than a smaller schools (Princeton has less than 10k students).
By the numbers there is likely a very strong pool of talent in lower-tiered schools, particularly when plucking from the top. And importantly, there are a lot more of them (both schools and students)!
Of course you have to be a good recruiter to find them. But that's the job!
Grade inflation at top schools further obscures the top quartile in top schools. Getting an A+ at Stanford was hard, an A was easier to get at Stanford than at NAU (so many more given out), and getting a B at Stanford was routine.
As a hiring manager, if you get a student with a B or B+ average from a top school with grade inflation, you really don't know much about them. If you get a student with a high A or A+ average from a lower tier school without it, they are quite likely to be good.
Grade inflation plus massive oversubscription of talent from similarly lazy companies results in an increased chance you're actually hiring closer to the mean than the tail from the top schools.
My anecdotal work experience fits this. While we had many talented employees from top schools (of all degree types). A lot of our most impactful employees weren't. And some didn't have degrees at all.
So here is where I net out on all this -- Build your hiring policy around the top students of good schools than the "good students" at top schools.
If not, you're doing it at the detriment of your company, and your team. And you're a worse recruiter for it.
Final point. I don't believe GPA correlates strongly to work performance as this thread might suggest. I don't believe a university degree does either. I'm just using it as an exemplary metric for the discussion. But that's a separate topic for a separate thread /fin
• • •
Missing some Tweet in this thread? You can try to
force a refresh
0/ VCs benefiting from offering bad advice to founders is such a tired canard. But it’s repeated enough that it’s worth addressing in some detail (a thread)
1/ VCs avoid investing in companies that don’t have potential for venture economics. We vet heavily for this. Founders' intentions is a huge part of the calculus. If there is early misalignment in vision and expectation, something went very wrong during the investment process.
2/ Different investors have vastly different risk tolerances and return expectations. There are many funds that invest in smaller TAMs or focus on efficiency over growth. It is incumbent on both sides prior to investment to ensure alignment.
1/ Advice on finding “the right idea” is often a variant of the tired painkillers vs. vitamins cliche -- focus on a problem you’ve experienced or others are experiencing yada yada
Yet, many impactful ideas were more akin to aesthetics, how founders felt the world should be.
2/ Take compute virtualization. For decades, VMs were an elegant oddity. The VMware team viewed them as the right way to abstract compute but weren't quite sure the killer use case(s). And yet that led to multiple deca-billion dollar markets.
3/ Many platforms and OSes have followed similar paths. They didn’t address a particular burning need now, but if adopted they would change how we did work and thought about work.
I’ll have a go - raw numbers are correlated with turnout, not relative popularity - 94% of democrats voted for Biden and there are 5% more democrats than republicans nationally - the only demographic that voted majority trump is white males (and they only make up 30% of the pop)
- exit polls contradict a number of these claims including LatinX voters outside of Florida and oil-dependent Tx (both that went Trump) - absolute gains for blacks are correlated with record turn out, exits polls show overwhelming backing of Biden
- much of the anomalistic skewing happened during mail ins in an election where the incumbent told his constituency not to trust ballots over mail
1/ Unlike traditional software, margins, scalability, and defensibility for AI companies are usually a function of the problem -- not the technology. And often the problems are ugly.
2/ Many AI problems show long-tailed distributions of input data (meaning most inputs happen infrequently), and supervised learning is not well equipped to handle it.
0/ A very interesting trend to watch is the disappearance of pre-sales, and what it means to GTM tooling. (a thread)
1/ Pre-sales has traditionally been the function that brings a new customer to technical and financial close. Generally it involves an account rep (AE) and a sales engineer (SE) to educate a customer on the product, do technical demos, PoCs, integration etc.
2/ However, with the shift to bottom-up this role is changing. And in some cases going away entirely. This is happening so fast that a surprisingly high portion of new founders don’t even recognize the “pre-sales” / “post-sales” lingo.