Hah, HN gives me the same feeling pretty consistently. If/when I leave the game industry I'll be disappointed if I have less than $500K total comp as a result of years of reading HN comments on what decent engineers should get paid.
I love this comment. I've been running my own small game studio for nearly five years in San Francisco and this resonates deeply. After reading HN for a while I constantly question my life choices work-wise.
Well don't feel too bad, we also pay $2000-5000/month for 2 bedroom apartments.[1] California also has a tax rate at the same level as Canada. Being a digital nomad you can keep the tax rate and cost of living pretty low.
But typical bigco sr. eng total compensation is around $250k. On top of that JS engineers tend to get paid less than mobile or backend engineers, but business owners make the most out of all of them.
[1] Depending how long you want to commute to work and where work is.
Well it's not too late to go to a larger place :). I've found it isn't that different, especially when they are a few years before IPO. AirBNB has a really small engineering team for their size too and are very web heavy.
If I was going to do a small startup again, I would found a company. Going from 40 employees / series A -> $1 billon unicorn, I realized that I would of made more at a big co even if I was able to sell all of my vested shares.
3 jobs out of 3,000 in a category does not invalidate the results and is within acceptable error (and even if there was a super-mega-ultra outlier as a result, that's why the median is used instead of the average). Although I made my best effort, it's infeasible for anyone to get 100% perfect categorization, which is why I would not have made this visualization without having a sufficiently large sample size.
I agree that a few misclassifications is not a huge problem, but it suggests a bigger methodological flaw. Looking at your source code, it seems your definition of engineer is
This seems like a poor definition of engineer to use. It includes "VP Engineering" but excludes "CTO", it includes "Growth Engineers" (marketers) but excludes "Code Ninjas" (developers). Short of word2vec, I don't think you're going to be able to automatically classify people as engineers based on job title.
That heuristic is good for 99%+ of the data set (and probably higher accuracy than a word2vec approach anyways). That's more than acceptable for this analysis.
You don't even need word2vec good old cosine sim or an svm on the descriptions would classify theses. But if you're stuck with just the titles, just make some more patterns. There's not much you can do without building a simple dictionary.
Startup has opening -> Can't fill it with people in their network -> Can't fill it with people who come to them -> Resorts to Angel List posting -> Angel list posting stays online longer because they have trouble filling it -> high turnover means the job ad stays up forever or keeps getting reposted.
Great jobs tend to skew toward the left side of that funnel = sampling bias.
Pretty much. Another major assumption here is that Angel List postings are legitimate - even though some companies post "job openings" online to vacuum up resumes for a database.
Angel List's salaries are literally a joke. It's not hard to find something like "senior full stack developer; salary: $1,000 equity 0.0%" or "co-founder salary $0, equity 0.0 - 0.05%"
It's completely useless. I can't imagine who you'd actually attract with these ads, because you literally can't pay the rent with this money and the equity for senior positions is insulting.
And they're really itching to work for basically charity for someone who will greatly undervalue them. Sure. I saw some unicorns by the lake the other day too.
What I wanted to do was map offers to startups (so I could at the least tag smaller startups), but the AngelList API intentionaly omits the startup name/information from the Jobs endpoint. I'm annoyed.
Nice charts, this is really cool. One thing, though - it's weird to have "researchers" not compared against engineers, since at lots of places they're functionally just two different titles for the same job. Is there a reason for that choice?
Equity percentage graphs of a large number of companies, with no additional information, are completely useless information. I'd much rather have 0.01% of Google than 1% of whatever your startup is.
I'd be curious if the distribution of salary data from H1-B filings was more or less the same. Anecdotally for brand name companies, eg Netflix, the selection bias of Angel List data to the lower end seems to be confirmed. The database publicly searchable and I had seen a visualization once but can't find the link.
These are angelList companies so the safe assumption is there is significant less than 200 employees. How many 200 person startups are their relative to 1-10 person startups, 1 out of 100 or more. Salary is low too so safe assumption.
