How I Can Has Cheezburger became a web empire

Then we could talk about how we make systems to help the editorial team sort through the 16,000 content submissions we get every day, or how we have a public API so people can submit funny content from their applications.

The "wordpress blog with cats on it" is a very shallow view of the company. There's a lot more going on than that.

1) cheezburger.com, the one that our users use to make content via our tools (Caption tools, graph builder, etc.). It's also a niche social network, where people curate their favorites, message with friends, etc.

2) The internal site that our moderators use to sift through submissions.

Nothing shows up on any of our voting pages without having been screened by a moderator. Nothing shows up on any of our sites' front pages without having been voted on by the community. A lot of what makes things work is the mix of editorial guidance and community feedback. It's not a pure democracy, but it's not a pure dictatorship, either. It turns out that a lot of deliberate thought and hard work goes into running a funny web site. I'm constantly amazed at how our editorial team is perpetually refining their process and sharing knowledge.

The software does things like automatically filter out duplicates and allow moderators to filter by items that have been added as a favorite by members of our community, etc. We also try to automate as many mind-numbing brute force things as possible, to let our humans do what only humans can do (answer "is this funny?") so the submission engine does things like automatically resizes and watermarks images. I personally spend about as much time working on software for around 16 moderators than I do working on software for around 16 million end users. Most of the moderators are in the office with me, and I really love having the low communication overhead of just being able to walk over there and ask "what exactly do you want, again?"

Almost all of our front-end sites are running a single custom Wordpress theme (PHP) hosted by Wordpress VIP (even with the recent downtimes, nobody is better at hosting Wordpress than Wordpress/Automattic). We send content there through the WP XML-RPC API. We use some extra metadata to identify posts as "voting" or "front page", and our theme is able to filter by that. We've also written some custom wordpress code to do things like display the "you may also like", integrate with your cheezburger.com account to add favorites, etc.

Both the content backend and the wordpress frontend are generalized enough so that we can usually get a new site from concept to launch in around a week. Some sites (senorgif.com, for example) require backend changes (support for Animated GIFs).

From a process side, we try to be cutting edge without falling into trendy bullshit buzzwords. We use a variation of Kanban, where we visualize and limit work-in-progress to keep things flowing through our system quickly in a way that allows us to respond quickly to new ideas or emergent problems. Our team is distributed, we use Campfire for multiple rolling discussions. Tokbox for video chat, Skype for screen sharing, and Kiln for distributed source control and code reviews.

With a few exceptions (only one guy really knows Flash) every one on the team is a multi-technology generalist, working on the full stack, top-to-bottom, front-to-back. We generally work in thin vertical stripes rather than thick horizontal layers. As a result, I'm able to have an idea for a feature on Monday morning and have it coded, reviewed, tested, and deployed by Tuesday morning.

We have multiple layers of test automation and monitoring, which allow us to turn around those small releases very quickly with very little manual human testing. We have a goal to do IMVU-style continuous deployment, where every submit to our central code repository would, if all tests pass, automatically trigger a zero-downtime production code update, but we're not quite there yet.

Ben's remarks on that were a reaction to a bunch of candidates for entry level position complaining loudly about entry level pay. He noticed that being hung up on pay was inversely proportional to perceived awesomeness (although some of this may be sour grapes and/or confirmation bias). It turned into a bit of a fiasco and I'm sure he regrets saying anything about it.

The truth is, people at Cheezburger are paid at competitive market rates. The entry level positions are entry level. Positions for experienced professionals with specialized technical skills and relevant education are not. I have no complaints about what I earn. Benefits could stand to be a little better, but that's always tough for small companies.

I'm fully aware that I'm working to make Ben (and his investors) rich, I went to business school so I know how these things work. I'm a consenting adult. I get paid for doing work that I enjoy and hanging out with people whom I like. I'm totally OK with this arrangement.

When I tell people where I work, I get two distinct reactions. "I can has what?" or "OMG! That's so cool! What do you do there!" I think that what we're doing is interesting from both a business and a technical standpoint and I thought I would offer to share with the HN audience.

FYI for anyone else who is curious like I am about what those 53 sites might be: http://cheezburger.com/sites

I think it starts with the assumption that people enjoy sites that make them laugh. We all need a break from our hectic days for a few smiles. That's why sites like The Onion, FML, Stuff White People Like, etc have thrived. They have hilarious content that's updated often and draws people back again & again.

Advertisers love being any place there are masses of people - particularly a place that can matches their target demographics and fits their brand.

Right there, you could have a fair business. Funny content + advertisers.

But to make this business scalable, you need a lot of content. Writing content is difficult and expensive. You would have to hire a lot of comedy writers (a la College Humor) to constantly pump out great stuff. It's not impossible, obviously, but for every College Humor there are dozens of failed sites who just weren't funny enough.

There's a better way though. If you can encourage your audience to generate the content, you'll have a cost-effective (read: low-margin) way of creating great content, plus you'll get the audience involved and make them feel like part of the community. It's very satisfying to see your cat with your funny caption on icanhascheezburger.com. That's the kind of thing that can make you a long-term fan and repeat contributor.

To take this one step further, have the audience rate which pieces of user-generated content is good enough to be featured on the site. This is similar to the Threadless model. You don't need to actively pick the best pieces of content, you just let your audience do it for you. And your audience is probably the best judge of what they think is funny or not. It's crowd-sourcing at its best.

Now you have: (user generated content + user ratings = funny content) + advertisers

What's especially insightful is that Cheezburger Network took this another step further. They built a platform that allowed them to replicate this formula across various "humor niches," if you will. As MartinCron said, it's more than just Internet memes, it's any crowd-source-able humor.

The platform itself is what I alluded to in my previous comment. It allows them to cost-effectively launch new sites after they've analyzed new potential humor niches. They don't even have to be comedic geniuses (though with a failure rate of 20%, perhaps they are), they just have to launch lots of sites, then kill the ones that fail. This Darwinian approach will leave them with only high-quality content.

MartinCron didn't mention this in his description of the technology, but I imagine they also need an advertising system of some sort. Advertisers on icanhascheezburger.com may not want to advertise on ihasahotdog.com, for instance. Their ad system would have to be able to manage their ad inventory.

So here's what we have so far: (user generated content + user ratings = funny content) + (advertisers + managed ads)

And this is being replicated over & over again by their platform. I focus a lot on their platform because I believe that is what makes their business scalable. You could try building individual instances of such sites over & over again yourself, but it would take quite a bit of time. Automation is key to a scalable business.

On top of all this, it sounds like they're now adding a basic social network to further increase a user's involvement on the site. That will surely encourage current power users to become even more loyal. And the ability to share content on social media services will aid them with marketing reach. All smart moves towards scaling their business even further.

I hope that answers your question, medianama. A little long-winded perhaps, but I think I've covered most of the major points, as I see it. I'm sure others can elaborate even further.

If I ever find that guy, he's in for a world of pain.

Getting rich selling plastic sheets to indian farmers does not bother most hackers because it's completely outside of their domain. Building a website or iFart app is not.