Founders have been buzzing all week about the hoopla created by President Obama’s visit to Reddit. The event not only created maelstrom of traffic, but spurred a lot of discussion on scale strategies for handling such traffic. Its hard not to get sucked into the conversation but keep it academic because startups in general spend way too much time thinking about and planning for that moment when traffic explodes but 9/10 times that isn’t likely to happen.
For those of you that aren’t familiar with Reddit, the 134th most popular website on the Internet (according to Alexa), it is a community of users who submit stories, links, ideas and then generate discussions around them. Stories are either “up” or “down” voted depending on whether you like them or not. The good stories percolate to the top and even to the front page where they will be seen by 20 – 30% of Reddit’s traffic. Hitting the front page is like being on the cover of the New York Times, everyone will see it.
President Obama participated in what is called an AMA or ask me anything. The format informs Reddit readers that a person will be answering questions and allows them to submit those questions in advance. Then the AMA featured guest goes through them and selectively answers them.
CNET covered the event and reported the following stats which are pretty insane:
Within the first 5 minutes, Redditors had posted 37 comments. That number grew exponentially over the next hour. After 10 minutes the number was 278; 30 minutes in, the number jumped to 5,266. By the end of the first hour there were more than 10,000 comments posted in the thread.
The President’s AMA generated more than 5 million page views as of today, Sept. 1st. At one point, the AMA page was generating over 100,000 page views per minute requiring that Reddit operations add an additional 30 servers to handle the load.
Those are staggering numbers to most people, but not to those who run Reddit. The site generates more than 270 million page views a month. Steve Huffman, co-founder of Reddit recently presented lessons he learned while building and growing his site. As I read through the article I was reminded of the many discussions I’ve had with startup founders who worry about scaling right out of the gate and devote too many resources to it.
Its hard not to get caught up in building scale into your 1.0 product especially after reading Steve’s article and fantasizing about all of those customers who are going to love your product, but you should consider this warning. Building scale too early will delay your ability to reach the market which could have a greater and more devastating impact to your business than not being able to scale.
My advice doesn’t preclude preparing for scale or in fact utilizing strategies to deal with it, just don’t devote too many resources to it. Here are some things I do recommend doing. You should instrument your platform early on so that you can detect areas of poor performance. Knowing where your weaknesses are is an important first step in fixing issues down the road.
You should also deploy out to customers in a staged fashion. Don’t just let anyone signup and use your product out of the gate, instead roll it out in waves. This way if you reach maximum capacity, you can close the flood gates and repair or scale out to alleviate the problem.
This advice isn’t necessarily for everyone, some startups by nature need scale out of the gate. For most startups however, this advice will likely be useful. President Obama will not be using your product any time soon and its even more unlikely that your product will generate hundreds of millions of visitors each month. Worry about the real problems in the near term and devote your energies toward creating a product with value and if you succeed, then worry about how you will scale it.