Last Thursday was a red letter day for Skribit as we officially let the flood gates open — okay well they were always open.. we just started telling people about it. As our first real marketing move, TechCrunch got the exclusive of our launch complete with some promocodes, a system we initially built for our appearance at WordCamp Birmingham. Several people have been wondering (a la Skribit suggestions) how the launch went from a stats perspective and if we had to deal with any performance issues and so on.
Here goes.
Promocode Mixup
Right before the TechCrunch article went up at 9am PST, I had been searching for Skribit on Twitter as I usually do, and saw a tweet about us from Dr Peter Meyers (@dr_pete).

I sent him a direct message, offering him the promocode we created for the TechCrunch launch post, which was set to give a completely free Pro account (usually $24.95/year), and mentioning how we were set to launch a bit later that day.

I tried it myself and while the promocode registers on our end and shows a $0/year price, it’s when the transaction must be completed through Amazon Simple Pay that I noticed this issue:
Input parameter promotionAmount is too low (Minimum promotionAmount is USD 0.01)
Doh! We use Amazon Simple Pay in a subscription role so that’s why we don’t bypass Amazon completely when using a complete discount promocode. In short – when the promocode system was built we didn’t think we would use it to offer completely free Pro accounts.
At this point the TechCrunch post had just gone live and the best thing I could do was temporarily alter the promocode to show a $1 price (the lowest we can do otherwise, it doesn’t like cents) as well as manually upgrade users to the promised free Pro account.
Sorry International Users
Another issue we noticed during the launch was that non-U.S. residents could not purchase Pro accounts as Amazon asked them for a U.S. address. At the time I decided to manually upgrade the affected users to a free Pro account. Before the launch, we had paying Pro account customers but I guess none of them were outside of the United States!
So, did your server catch fire?
Prior to the launch, we made sure to employ some basic performance tweaks to Skribit (a Ruby on Rails application). First and foremost, that included simple page caching of the homepage and making it static. Initially we had signup via Twitter OAuth do a callback to the homepage, triggering a dynamic, server-side element on the page which page caching broke. We changed the Twitter callback to use the signup page instead and we were able to continue page caching the homepage, along with some static pages like the About and FAQ pages.
For the record, entire page caching in Rails is fairly easy. It’s the pages that need to be selectively dynamic based on the current logged-in user that are a bit trickier. We’re still working on that aspect with some fragment caching, and looking into Rails plugins like Cache Money, Cache-Fu or Interlock that can cache Rails finder lookups, serve as “intelligent fragment caches” and so on.
I moved all of our images to Amazon CloudFront, which we had already been using in some capacity for hosting the javascript used by our widgets, and most javascript used throughout Skribit has been hosted by Google for a while now.
Back to launch day: our servers did not explode or catch fire. We have a 2 VPS setup served up by Media Temple. Each VPS has a custom install of Ubuntu with Rails done by Apache and the lovely Phusion Passenger. However, mid-way through the launch we hit the daily limit on Google SMTP and Skribit was unable to send automated emails (new account, et cetera). That was taken care of by delayed_job and they all started getting sent the next day. We are currently looking into moving our email system onto SendGrid (a TechStars startup that just received $750k of funding).
The next day however I was woken up at 5:40am by a downtime notification text message from Are My Sites Up (I also use Pingdom religiously but do not pay for SMS notifications through it). I SSH’d into both of our servers to find that the database server was running 6.0+ load averages (100% load would be a 4 load average as it is a single processor, quad core machine), but had lots of memory to spare. Lots of poking around and I solved it by editing ~/etc/my.cnf and changing innodb_buffer_pool_size to give it more memory. We had left innodb_buffer_pool_size it at the default which only gave MySQL a few paltry megs!
After that the database server went to around a 0.1-.2 load average and the app server a bit higher.
Show me the stats
Even with that being said, our site survived the TechCrunch traffic without a hitch. The TechCrunch post created just shy of 20,000 pageviews from 3,500 unique visitors that day.

If I could have changed anything, it probably would have been better to launch on a Tuesday instead of a Thursday. While it is to be expected that traffic from such press will spike then fall back down to more or less normal levels, we could have received slightly higher residual traffic had the weekend not arrived two days after the launch. That traffic also turned into around 300 tweets mentioning Skribit (mostly bot/automated RTs of TechCrunch to be fair), 33 new bookmarks on Delicious.com (although skribit.com has a total of over 700) and various conversations on sites like Hacker News and Facebook.
Here’s what that turned into:
168 people redeeming our promocode through TechCrunch, 18 people redeeming a different promocode from my blog post and several others from Lance’s blog post. Then a separate 35 or so accounts I manually upgraded to a free account after responding to their tweets, emails and comments. More impressive is the total number of new accounts during our launch period alone: 938 new users. All starting from a TechCrunch post. I’m eager to see what numbers we can get from more substantial marketing efforts.

New Users and Blogs (timeframe: Aug 09 – Dec 09)
We make it easy for users to register for Skribit by using Twitter OAuth (in addition to OpenID). It turns out 34% of new users during our launch made use of that and signed up through Twitter:

I was planning on showing off some pretty screenshots from KISSmetrics, which we are beta users of, but unfortunately they have a large volume of data to process through so we do not yet have any up-to-date information to show off our launch day conversions and such. Below is our signup flow as captured by KISSmetrics with everything from the beginning of the month to a portion of launch day traffic, defined as people first hitting the homepage, making it to the plans page and then finally the signup form.
Data from beginning December 09 to part of launch day. KISSmetrics still processing..
What’s Next (aside from continuing to build out useful features)
TechCrunch was the first “getting the word out” initiative as they generally won’t cover you unless you give them the exclusive. The next steps involve doing more of the same with blogs in different niches and doing one of a few things: helping them start using Skribit on their own (there are quite a few large bloggers I believe would benefit tremendously from Skribit – the type that are usually inundated with tips/requests from their fans) or offering them a set of free Pro accounts to giveaway on their blog, and do interviews with bloggers blogs as a way to promote Skribit. Here are a few recent interviews on some blogs:
- Israel Startup Weekend: An Interview with Skribit Co-Founder – Paul Stamatiou
- “On the Couch”: Interview with Paul Stamatiou – Co-Founder, Skribit.com
- Your Hidden Potential: My 100th Interview: Paul Stamatiou, A Georgia Tech Graduate Talks about his Start-up Skribit- A Website that helps bloggers cure writer’s block
And all along the way we’ll keep making changes from user feedback. Most recently, a new user was wondering if Skribit was compatible with WordPress.com blogs. We had an FAQ page floating around but I went back to it, restyled it and am in the process of building it out and linking to it more prominently on the homepage and elsewhere.
I also have yet to work on a big email marketing campaign about our launch (thanks for the credits MailChimp!) but I’m not sure sending an email to thousands of people right before the holidays will lead to a good open rate.
Checkout my other blog posts about working for Skribit here.
Questions or just want to show that you enjoyed this post? Drop a comment below!























