Last week has given our whole team some grey hair and a lot of big smiles. After giving a nice interview to ArcticStartup blog-magazine on Monday, we were doing our usual product development things. User interviews, a blog post about font differences, Android support and so on.
The show begins aka HackerNews Effect
On Thursday morning, I got a pat-on-the-back-mail from an old schoolmate about the post on ArcticStartup. After reading it, I opened up Google Analytics and saw a nice flow of sign-ups and new tests, but nothing extraordinary.
But then things started to happen. Someone posted the article on HackerNews, later on Reddit. The article moved to the first page of HackerNews and even though we were in the bottom third (20+) of the news list, our website was swamped with a constant flow of visitors. This was great news! Luckily, our landing page was hosted on GoDaddy and ready to take on visitors.
What was not so great was that the load of users that signed up for the service and requested their webpage to be tested was too much for our trusty web app. Response times lagged (nearly everything on the front-end kept on working albeit slow). We were not ready for popularity yet. To be honest we were not yet looking for that in such large quantities.
In 4 hours we found out about 4-5 bottlenecks in the system that we managed to clear by adding additional resources. While the team was focused on keeping the system going, I hopped in to inform the incoming flock of the delays that awaited them. After dropping to the second page of HackerNews the load receded to acceptable levels and our app managed to churn through the lined up test requests. We stayed up late (surviving on pizza) and made sure that the app would make it through the night.
It ain’t over until the fat lady sings
Friday morning greeted us with a nice “tail” of visitors. Some bugs in the back-end, but nothing too serious. Then it started again – the Spanish website genbeta.com wrote a review about us that resulted in the next “wave”, which turned out to be even bigger than the “HackerNews effect”. The app was performing a little better, but the hardware was simply not ready to take the punishment.
The Spaniards liked the article so much that they kept on tweeting about it. It spread like forest fire and this wave did not slow down nearly as fast as the day before. The signup count passed Thursday’s and kept on growing. One nice surprise was the crazy-good conversion rate (visitors who created an account) that was well over 50%.
As Friday passed, we hoped the weekend would give us some room to breathe. We turned our focus on getting the stability of the application to a level that would not require manual maintenance over the weekend. We identified that we needed to patch the iPad shot worker to handle hundreds of oncoming tests and decided to take it off from public use until the fix is validated.
The weekend showed receding volumes of traffic and we guessed that by Monday it would be “over”. The following Monday and Tuesday greeted us with significantly lower spikes, but a steady “base” flow of new users that is not decreasing. Wednesday was “flat” but on a new level. We were getting sign ups at a pace that we would have been singing about a week earlier.
During the whole course of the adventure, we received quite a bit of feedback both through portals and e-mails. It was pointing that we were moving in the right direction. For instance:
- A few people ranted publicly that they could not pay to get ahead of the line (a good alternative view for our freemium model!)
- A bunch of “pat on the back” e-mails
- Feature requests that gave a good input to our development backlog
- The HackerNews effect is not as big as some people think, especially if you’re not number one on the page