Presentation: Scaling for Your First 100k Users
Speaker: Matt Mullenweg
Event: Webvisions 2006
Date: Friday, July 21, 2006

I was impressed and intrigued by the best practices Matt Mullenweg presented despite the hardware-heavy titled presentation, “Scaling for Your First 100k Users.” As the founding developer of blog software WordPress, the Jazz musician turned blogging software developer has garnered world-wide recognition and a following of blogging elite. The Q&A after the session was equally as interesting and tech-filled with lots of questions from the audience about infrastructure and hardware. If you would like to start a blog, you should sign up at WordPress.com.

“The first 100k are going to be your most passionate users, your most unique users, and they’re also going to be the hardest for you to get to. The first 100k are like rolling the rock up hill, and after that you’ve got enough momentum that you’ll be fine.” -Matt Mullenweg

Mullenweg’s 12 Rules:

1. You have to be your most passionate user.
2. The act of writing things on paper frees the creative juices.
3. Obsess about the details. Pay attention to every single part of the interaction. See what the experience is like for the people using it.
4. Do your own support. Feel the pain of your users. Track the number of support requests, time spent, how much time you’re spending on it, document as you go along. Don’t bother making a lot of documentation before people are asking questions because it’s not going to match their mental model. Make it as easy as possible for people to contact you in every single way.
5. Blog every step of the way. People will root for you. People will help you. It’s a good way to find people to work on what you’re also working on. People will cut you a lot more slack.
6. Have a great tag line. If you can’t describe what you do in 5 words or less, simplify what you’re doing. Create contextual tag lines: Gutenberg for the web for librarians, the democratization of media for politicians.
7. Frame everything you’re talking about in a context for your users. People want to know what you’re going to do for them. “We make it easy for you to do what you want to do.”
8. Get out version 1.0 as fast as humanly possible. User feedback is completely impossible to duplicate. The people who give you a hard time will push the limits of what you’re doing. Be able to listen to the silent majority.
9. Measure your success. Identify your core metric and watch it religiously.
10. Know what to do if you are successful. Does your business model scale with the number of users?
11. Start strong and end strong. First impression is very important. Lead with the most compelling point for why people should use your product or service.
12. Be a pain killer, not a vitamin. Everyone has things in their life they care about more than you. Address a real need that someone’s having. People really love Aksimet because it’s a pain point.

The Q&A following Matt’s WV06 presentation was equally interesting. Here are some snippets:

• WordPress.com is stateless, it’s all php talking to the database. It has no load balancing it’s all round robyn DNS.
• A fantastic book to check out for scaling is Building Scalable Web Applications by Cal Henderson of flickr.
• Models that Matt looked to: flickr and LiveJournal.
• Matt learned that eventually you are going to have to partition the user data. It’s more and more painful the longer you wait, so if you do it from the beginning, you’ll be set.
• Uses four replicating boxes instead of one box.
• Recommended myISAM instead of innodb because of file system and memory issues. Matt said he would highly recommend using innodb for running a single blog, the file system is fat but fast.
• Get out of that one box mentality from the very beginning. Start with two boxes. Instead of buying on $150 box, start with two $75 boxes. Go with the cheapest boxes you can possibly use. It’s always to have more. Everything that will possibly fail on your system will, so just plan for it.
• A mistake Matt said he made in the beginning is that they spent a lot of money on their first servers–they were big expensive boxes that weren’t well optimized for what they were doing.
• A little background: Matt started out as a musician and then claims to have gotten side-tracked into “all this awful web stuff.” He used to trade web design services for music lessons.
• Start with fewer features that you think you’re going to need because you can always add features–people can’t stand removing features. It’s better to start with less than you think you need.
• Matt says, “pretty much every web host worth it’s salt has a WordPress installer.”
• You have a very real responsibility to make sure that data is never ever ever lost, and if you have any problems you will loose that trust.
• If it wasn’t hard, then it wouldn’t be worth doing.