The pie chart indicates that only 15% of companies use CSS? That's clearly the wrong way to display this data. Very few of the technologies covered are mutually exclusive.
1. In some cases the technology choices aren't mutually exclusive, as you point out. A better representation would be a bar chart that tops out at 100%
2. Not all technologies are equally detectable. The ones that are harder get penalized, and those that or impossible to detect are penalized 100%. In categories where this is endemic, the ones that are detectable appear to be dominant when that is hardly certain. Frameworks is probably one of the most egregiously misleading of the bunch.
I think if you combine Ruby on Rails and 'Ruby on Rails Token' it's larger than PHP. It's strange to me that they track this by their own detection algorithm instead of aggregating based on framework.
And if that's true, why aren't there more php articles and comments on HN? Are php developers that uninterested and apathetic about their own language?
The PHP community isn't terribly vocal. Probably because it's been around for ages and the dos and don't are very well understood by people who care to know.
I personally don't read many PHP related articles because, well, there isn't much new to learn. It's an old/stable language with well understood constructs.
Part of the problem is that the Rails community being as vocal as they are make it seem like everyone is using rails. In this case perception is not reality.
You do have a point. Pretty same is the case with perl I think. Libararies are already there for ages and matured.
I'm not into web development stuff, but I thought currently big projects prefer not to use PHP. People say PHP helps them to find cheap programmers, but I dont think good PHP guys can be hired at low rate anyway.
The list includes Dropbox, Scribd, WePay, OMGPOP, Loopt, and Weebly, but a lot of sites are missing out on potentially useful information--only 18% of 136 sites were detected as signed up with webmaster tools.
I used to sign up for webmaster tools but after a while I just didn't find the utility in it anymore. I'm able to control which domain Google sees as canonical (bare vs. www) without using webmaster tools. That was why I started using it in the first place. Though it's nice to see crawl errors or other potential issues.
From what I understand GA isn't showing search keywords now for users that are logged into their Google account (totally frustrating change as more of your users seem to be logged in now). Webmaster tools does kind of show this information but without any of the neat metrics that go along with a GA report.
I've also heard horror stories. There was an article (or possibly just a commenter) on here not too long ago that was describing how their position on the SERPs dropped dramatically just a week after signing up for webmaster tools. That kind of thing sticks with you, even if it's not true (and I do suspect there were other variables involved that he wasn't telling us).
So, it's not all about whether someone is savvy or not. ;)
Edit: Almost forgot! A killer feature of WT that I think no one should do without is that it shows you what pages on the web are linking to your site. I'd forgotten how awesome that is.
That information is not even close to being as complete as what they report to you in webmaster tools. There might be 2000 pages pointing back to your site and that query will return 50 (just an example, not hard numbers).
It's definitely not better keyword analysis, it's better then nothing keyword analysis. Not only are the aggregate numbers unhelpful, the "top keywords" are usually pretty far off from reality.
Maybe I'm missing something - I'm no Google Analytics expert. Here's my reasoning as to why I like Google Webmaster Tools better than Google Analytics ...
- Impressions
- Clicks
- CTR
- Avg. position (SERP)
To me, the Google Webmaster Tools information is more actionable. I have a number of different pages optimized or certain key phrases. I can look at keyword rankings and prioritize SEO. I can spot high impression keywords that I don't have a targeted page for, and create one. I can look at the CTR and see which page descriptions could be tweaked to get a higher CTR. The GA information is interesting, but ultimately not as useful for my needs.
The site won't show those that aren't doing non-meta ways so I suspect the actual registration would be higher than this. Although you can see the actual trending at http://trends.builtwith.com/Web-Master/Google-Webmaster
Seriously though: There is really too much noise in some of these diagrams to get much out of them. What surprised me most was nginx overtaking Apache in this sample group.
It's nice to see builtwith on the HN homepage, thanks for submitting this page Alex. I'm reading through everyone's feedback and will make changes accordingly, thanks everyone for commenting.
A few notes and things I will clean up -
* The pie charts I admit appear misleading thanks for affirming this, for widgets and technologies they show the leading tool but for document based technologies like CSS/JS they don't look right. 15% of sites aren't using CSS, > 85% are, and I suspect the rest are either dead sites or some other issue. I'll look at swapping to bar graphs which some of you have mentioned might be a good alternative
* PHP usage is based on server side implementation so a server might support PHP it doesn't necessarily mean the site is written in PHP
* The source of the list is yclist.com from about 3 months ago
I think the pie charts aren't just broken for document based technologies, they are wrong for anything where a user can have more than one of the tools. For example on our site we are using multiple types of advertising (including our own), multiple types of widgets, etc.
That being said, very cool that you were able to pull this together - the raw data here is very neat and it would be cool if we could even dive in more and end users could play with it in different charts, etc.
Put up a copy of the categories and the technologies in each please. I want to go through and use d3 to make it look prettier and convey more information.
I thought it was very telling that out of the 136 companies surveyed, only 2 had a payment system. Speaks volumes about the sort of companies YC is trying to incubate/build.
Only 2 were using a payment system that the reporting site tracks. That doesn't mean only two were accepting payments.
Notice that both that showed up were intrusive payment systems. What I mean by "intrusive" is that they take you away from the site you are shopping on to enter your payment and shipping information on the payment system site. My guess is that this is the only kind of system the reporting site detects, because you can find it by noting the links to the payment system site.
If a site were using a transparent system, where the payment and shipping information are posted back to their own server and they deal with it on their own backend, it probably would not show up.
For instance, if your cart checkout just posts back to your PHP or Ruby code, and that code uses Braintree's PHP or Ruby API to process the payment through Braintree, there is no way the reporting site is going to know that you are using Braintree. Same if you are using Merchant e-Solutions, PayPal's PayFlow Pro, or something similar.
"detected using at least one of the technologies we track in this group" - does this mean some technology usage could be higher but it's absent because they're not tracking it?
Frameworks - How the hell do they "track" some of these frameworks here because the produced html/css/js could look exactly the same coming from say flask on python and jersey/jax-rs on java?
- PHP might be the most practical choice for some of these companies. The founder or lead tech might be most proficient in PHP for real-world applications (even if he adores Python or Ruby).
- PHP might be the language that everyone on the team knows so they choose it to keep everyone equally involved in the production of the site.
- The people commenting on HN aren't necessarily people who start companies (truer now than it ever was).
Edit: Also, it might be the case that some of these servers purposely lie about what language is used on the backend to thwart script kiddies.
Yeah, they're splitting hairs on the Ruby count (not sure why). But I'm assuming that at least the number of PHP installs is somewhat accurate which is still a good number of the sites. Once you add the two Ruby counts together it does win out over PHP (which is more in line with what people would expect from SV types).
People that want to get stuff done don't waste their time hating on other people's tools.
Almost everyone on this site, including devoted PHP devs, recognizes the fact that PHP isn't exactly an elegant and well-designed language.
The pointless "hate" however only comes from a small group of people for reasons really only known to them. But if you look outside the world of programming languages to the kind of people that loudly proclaim to hate something, I think you'll get a pretty good idea of the kind of people we're talking about.
These are not the majority of users of HN. They just feel safe here, like hooligans feel safe in a stadium filled with supporters of the same team.
I would guess that a lot of Python sites are not getting counted for some reason, maybe because Python doesn't broadcast its presence the way PHP does. I'm pretty sure Dropbox and MixPanel at least make heavy use of Python, but neither is listed as a Python user.
It's not the most used language, actually, though I don't think it's your fault for misreading the chart. Ruby on Rails and "Ruby on Rails Token" were split into two different categories, based on how they figured out the site was using Rails. If you combine those two categories, ruby has higher usage than php.
Not to defend the php hate, of course. Just to point out that it's not actually the most popular language, despite the misleading graph.
First of all, as mentioned below if the server advertises PHP support, it's considered that the page is made with PHP (so a standard Apache install that comes with mod_php will make it think you're using PHP).
And second, I put my company's (admittedly not YC, but hey) homepage there just to try and since we're using a hosted Wordpress for the landing page so that business people can change the copy easier, it thought we're using PHP when in reality the app doesn't have a single line in it.
Yeah - I noticed the same thing with Python / Django being so poorly represented - seems like there's an disproportionate number of Python related posts and discussions on HN - but it could just be my internal bias as a Ruby guy :)
The pie charts seem like an odd way of counting some of these; for example, one site could use Facebook Like, Google +1 and a Tweet button all at once, but showing them as slices in a pie chart implies to me that the options are mutually exclusive.
The pie charts also don't match the percentages in the tables. Took me a second to notice, but the pie chart shows the relative percentage of usage by sites using a particular category, while the table shows % sties out of total # of sites.
Not just odd, but completely wrong. For data that doesn't add up to 100%, they should be using a bar chart or some other representation of data. Unfortunately, this is the sort of sloppiness that makes me wonder what other amateur mistakes might have been made in collating the data.
Obviously this analysis is meaningless if the company isn't building an internet-facing web application. And even if they are it isn't that interesting, as it doesn't capture anything about the back-end. Stick a nginx caching reverse proxy and change a few HTTP headers and you've got something that looks completely different as well.
- X-FRAME-OPTIONS: IMO, more people should know about and use this.
- Adroll: Would be much more effective if it didn't keep showing me ads for products that I already use. I wish it had an opt-out so I'd quit seeing the same ads over and over again everywhere I go online.
P.S. I know others balk at the pie charts (for good reason), but for me it's irrelevant. I go straight for the numbers and the charts add a nice visual break so that it doesn't look like a wall of text. So at least they have that going for them. :-)
Most advertisers that do display retarget attempt to build segments and suppress ads to buyers/logins from potential new customers. this can be difficult though with cookie deletion and people using multiple devices/browsers.
Interesting information that I'm genuinely interested in but the execution and delivery leave me scratching my head.
The operating system and web server graphs make no sense.
mod_ssl? varnish? Makes me question what was asked and who was answering.
I'd also contend pie charts aren't the best way to represent this information, especially in the categories where there are only a small number of responses.
Most likely all information was gathered automatically. According to their FAQ (on the right side of the page) they check the homepage and attempt to detect the presence of any of the listed technologies.
I agree pie charts are probably the worst way to visualise the statistics. For example, it gives no sense at all of the ubiquity of any technology.
Also the accuracy seems to be lacking. I was surprised by seeing their (automated) claim that only 85% of the sites use CSS. That seems strange since I can't actually remember having seen a site recently that doesn't use CSS on its homepage. Checking some of the unlisted sites confirmed my suspicion; they just failed to detect CSS on those pages.
Finally, these statistics only give insight into what is used on the home page of the given domain. The home page may just be part of a (separate) marketing website, so don't be fooled into thinking these statistics represent the products/services built by these companies.
Disregarding all the problems with these stats, it is indeed interesting to see a comprehensive list like this. I find their list of technology trends [1] even more interesting, though.
Great having statistics like this, but for most of the statistics the pie chart is rubbish.
It only makes sense in the case where the companies really only uses one of the technologies posible.
Bar graphs would have been better.
EDIT: Seems like I need to refresh the comments before i post my own, archangel_one beat me to it.
These are startups. IIS may have taken some of the market for more stable companies but it's not really practical for cash-strapped startups to use IIS. Also, college talent typically has greater affinity for Linux.
The way it's detecting Django (checking for the Django CSRF cookie) is false-negative prone. For example, our site (crocodoc.com) uses Django but is not registering for it; probably because we don't have any forms on our homepage so no CSRF cookie.
