Happytweets Retrospective

During the Christmas holiday season, Portland found itself blanketed in one of the biggest snowstorms in the last century. Since the city doesn't generally anticipate that kind of accumulation, and that Portlanders generally don't fare well when faced with driving in such conditions, I had quite a bit of time indoors parked in front of my PC. More time than usual, anyway.

I'm not sure how I came about the idea that eventually turned into HappyTweets, but I had spent my first weekend of the snowstorm researching and applying a methodology that would eventually become the algorithm for the site. I can't say much just yet about the inner-workings of the site, as I've brought the concept to my employer and it has some potential. I anticipate to be sharing more on this in the coming months, but one small step at a time.

Word of Mouth, FTW

Happytweets took off in a way that far exceeded my expectations, especially for a quick, fun project I threw together in under ten hours. It truly speaks to the viral potential of Twitter and how all anyone really needs is an original idea, and a clever marketing angle. I don't consider myself to be all that clever, probably just lucky in this case, but I have definetely walked away with lessons learned as well as immediate action items to improve my algorithm.

The site has lost it's peak appeal, as I expected it would quickly, though it lasted much longer than I expected. It received a considerable amount of traffic in it's first month at about 708 visits per day. Traffic started slow, but really took off after I added a "Retweet This" feature, which made retweeting a user's score possible with just two clicks, and didn't require them to authenticate with Twitter. I released this shortly following the TwitterRank authentication scare, so I didn't want to send the wrong message.

Postmortem

My biggest takeaway is that I learned more about the Twitter API. Because I was relying on the account API (different from search), I was restricted to just 70 requests an hour. There were several times that the traffic load caused me to violate the request limit, which caused a negative user experience for some Tweeps. Some of them let me hear it, too.

Another is that I relied heavily on Twitter Search to get a sense of how the public was reacting to the tool. As to be expected, there were some skeptics, but in general, the reaction was pretty positive. And for the record, no, the algorithm is not based purely on emoticons. :)

The last big takeaway was that the agorithm did not perform well for users that have not posted very many tweets. I also wanted to create a graph that showed how your happiness changed over time, but found that there were not very many returning visitors, so this wouldn't be all that valuable.

Finally, I owe a huge shout-out to @semaphoria and @divamatrix for helping spread the word. The two of you truly reinforce the argument that Twitter influence can have a snowball effect.