Some of these graphs display several non-exclusive (sometimes totally independent) things in a pie chart.
For example, the second one shows "on-site", "full-time", "remote", "intern", and "visa". Remote/on-site seems like a mutually exclusive pair, but full-time/part-time seems logically and practically independent from that (and "part-time" isn't shown in the pie chart); "visa" probably logically implies on-site; "intern" likely implies on-site in practice; and both "visa" and "intern" probably imply full-time in practice, but not logically; and I think "visa" and "intern" are probably logically and practically independent of each other.
And many jobs list multiple languages, multiple frameworks, and even multiple of the things in the "databases" category.
I suspect each datum is best considered by itself, as "N% of postings mentioned X". Could be displayed as a bunch of bar graphs or something.
Currently the charts only show counts that have 15 or greater results which is why part-time isn't showing. The table below the charts allow you to filter/search by part-time though. I will add a notation to the page that indicates the 15 count min.
I can group some of the different types together but after digging through the raw data I saw that there are a ton of different edge cases. I didn't feel comfortable implying anything because of this.
I'd be really interested in seeing a breakdown of locations, if possible. Naturally, most of these positions would be in California (from what I saw yesterday skimming through the thread), I was surprised to see there were fewer than 10 positions in either Denver or Boulder yesterday.
My thought exactly. And it's even more specific than California - I've been pretty disappointed with the number of positions posted for LA, and haven't noticed an upward trend. I would be very curious to see how % and # of positions in SF vs. other cities has changed over time.
I will try and add this to the next revision. Location would be interesting to see. I will need to pull in a reference table of cities/states in order to properly match locations to the text in the job posts.
It gets many times a framework called "flex", but it's misscategorized heavily. "flexible hours", "flexible working hours" and "we are pretty flexible" are just the first 3 examples I opened.
Thanks for sharing this, cool to see others analyzing this data. I've been compiling monthly trends for a few years[1] of languages, frameworks and databases too. I see you define the terms to search for as well. I need to do a better job of grouping, but here's the raw list I use[2]. What was the most interesting thing you've learned from the data so far?
Very cool! I really like your "Rankings and movers" section and your compare feature is really nice. I was little surprised to see Python at the top and that a lot of companies are open to remote.
The most important raw stat is how many comments are there by month, over time? That will give some insight into how the tech ecosystem in general is doing.
I graphed this by month from April 2013 to April 2015 -- if you were to add the #s to today, you'd see a downturn in early 2016, but back up to April 2015 levels by now.
I got excited seeing Rust listed as a language so many times. But almost all the results linked didn't have Rust. Just the word "trust" (which appears a lot...).
It would be really interesting to see this for the "Freelancer? Seeking Freelancers?" posts. From what I've seen, people seeking work outnumber those seeking freelancers by something like 10 to 1. As a result it would take some aggregation to get an idea of the buy side of the market.
I wonder if you could boost the solidity of the language data by tracing backwards from framework. For instance, I see one posting in your data where the framework is known (Ember) but the language is not. Of course we know Ember is a JS framework, so you could answer both questions.
A big mapping of Python => (Django, Flask, etc.) Ruby => (Rails, Sinatra, etc.) ...
I think employers have gotten better with adding this information in some of the more recent months. Some of the older results lack a lot of data so it's not truly an accurate comparison.
As an example: May 2016 had only 74 results that were not matched up to a job type (full-time, remote, etc.) compared to Jan 2015 which had over 300.
Another factor is growing awareness of the Who's Hiring thread.
Anecdote:
My company isn't hiring any more or less people than they used to be, but we never considered the Hacker News Who's Hiring thread as a recruiting venue until I pitched it to upper management rather recently. So that's at least one example of new jobs appearing here that aren't necessarily new jobs in the economy, just new to Hacker News.
I'd be willing to bet other employers who began posting job ads here rather recently have similar stories. I doubt all the growth is strictly related to the tech hiring economy growing faster. Much more likely a combination of solid macroeconomic growth with Hacker News' (and similar venues') own growth combined with growing acceptance of advertising jobs this way by employers and employees.
Several of our applicants can be traced back to the Who's Hiring threads and I'm pretty sure at least one of our new hires in the last few months came from someone who found us via one of those threads. Not 100% sure; I'd have to ask the person on question if my assumption is correct.
But we only just started doing this a few months ago. Plus our hiring needs are pretty specific and aren't likely to align well with the majority of HN job seekers for a wide variety of reasons. As such we don't expect a flood of applicants from this venue. We don't need a huge amount of people though, so we're willing to be patient with it, keep posting our ad, and wait for the right handful of people to come along.
At the moment our highest ROI recruiting efforts are old-fashioned college recruiting events. We do a few different kinds at a handful of universities near us and found it to be a deep well of talented people eager to break into the field and we have a pretty solid training program to take entry level people and turn them into solidly productive engineers rather rapidly.
I really wish the "Who is Hiring" threads had a more unified formatting, to make it easier to search (and collect data) for certain things, especially location.
It's all loosely-typed interpreted languages for the web. You can see the bias in the existence of the "Frameworks" column.
This is San Francisco it seems. They aren't into making real things like spacecraft, factory robots, aircraft, engine controllers, CNC equipment, and disk drives. Real things tend to require factories, and that doesn't fit the usual get-rich-quick start-up mentality.
Heck, they aren't even into making any non-web things. They don't seem to do phone apps, firmware (BIOS, UEFI, etc.), 0-day exploits, regular old desktop applications, ASIC/FPGA/PCB dev tools, compilers, trading software, medical billing software, tax software, and timecard software.
From my perspective, there are no jobs in San Francisco. I couldn't stand doing web stuff. I don't see how people can tolerate it.
Having skimmed the jobs thread a few times I would say that's an accurate thing. Feels like 99% of the posts there are for website devs that know javascript.
I'm not sure if that's because 99% of software jobs involve website dev with javascript, or because the majority of the people that visit hackernews are. Hrmm?
For example, the second one shows "on-site", "full-time", "remote", "intern", and "visa". Remote/on-site seems like a mutually exclusive pair, but full-time/part-time seems logically and practically independent from that (and "part-time" isn't shown in the pie chart); "visa" probably logically implies on-site; "intern" likely implies on-site in practice; and both "visa" and "intern" probably imply full-time in practice, but not logically; and I think "visa" and "intern" are probably logically and practically independent of each other.
And many jobs list multiple languages, multiple frameworks, and even multiple of the things in the "databases" category.
I suspect each datum is best considered by itself, as "N% of postings mentioned X". Could be displayed as a bunch of bar graphs or something.