I built it in mid-2007, and it launched with a yawn. A year later, it still had only 150 users, a dozen of those being my own friends.

But it's recently taken off like a rocket... and I'd like to report why.

1) following up on uninstalls
Whenever someone removes the app, facebook will ping a URL on my server. When my server gets that removal request, it deletes that user's profile and associated data, but... it also sends me an email message. In that message, I've coded in a link which composes a new message to the person that uninstalled the app. It reads (paraphrased): "you uninstalled [appname]. Can you tell me why? please let me know how I can improve it."

I get about 50% response from those requests for feedback, and the suggestions have been invaluable. I used all that early feedback to make simple - usually subtle - changes to the interface, to improve useability and enhance that first impression.

For example, one response was "I uninstalled because I couldn't figure out how to do X". Replying immediately, I not only explained to the user how to do X, but I moved the button that let you do X to a more prominent and obvious location, and changed the text caption to be more clear.

I repeated that process 100 times, and what a difference it makes - even I who built it, find it easier to use and navigate than I did in its first draft. Now my app also has a very low uninstall rate, indeed less than 5% - which is practically unheard of for F8 apps.

2) respond to bug reports immediately
early on, when I got a bug report, I'd reply in person (by a message, not through the wall or discussion board) to the reporter, and do my best to have it fixed within the hour. Some of those people who inadvertently stumbled upon an obscure bug went on to do some darn thorough QA and reported many more, just for my benefit. They went from sending a snarky "hey buddy, your crap is busted" to being an ally in serious distributed, free QA testing. Having gotten involved, I engaged them on a personal level, and they stayed involved. Some of those bug reporters were among the app's first "fans".

3) stuffed the name with a keyword
* real app name changed to respect the ToS. by the way, it's not a horoscope app either, I'm abstracting as an example
The name of the app used to be "iZodiac"*. But no one looking for horoscopes would actually search for "iZodiac"... One day I did a search for "horoscope" and it didn't even show up, even though "horoscope" is in the description, summary, etc. So I changed the name of the app to "iZodiac: a daily horoscope". Now when you search "horoscope", it's there near the top of the results! The effect was almost immediate - new installs and traffic jumped within the hour and stayed up. Shows that the search algo at facebook relies very heavily on the app name, and not on words in the description.

4) broadcast to feeds
I had originally created the app with "stories" being broadcast to the users feed. But when facebook switched to the "new" interface, the broadcasting functions stopped working so I disabled the stories and forgot to fix them later. Well I remembered a few weeks ago, did some research in the F8 wiki and got them working again. The effect was not immediate, but it did happen quickly - a day later new installs per day had doubled, and within 3 days the number quadrupled.

5) meatspace spamming
I almost forgot to mention this one.
I printed a bunch of business cards with the app name on it, a little logo and short description and the URL. I brought a bunch of those cards to pubcon - if you met me there I might have passed you one! Hence in early November a whole bunch of influential people, who all presumably have heard of the internet, many of whom also having a facebook account, were solicited to try the app. I believe getting some "influencers" to try the app may have had a Gladwellian effect - but it's hard to make direct correlations with traffic growth as with those factors previously mentioned.

* * *

these improvements were all pushed in late October and early November. I used to have email alerts whenever someone installed the app, but I had to turn it off because my Inbox was going "bing! bing! bing! bing! bing!" like a slot machine on ludes, which interrupted my work.

At the beginning of October, the app was getting only 2 or 3 new users per week. Now it's getting over 150 per day, and it's accelerating. If you look at the "insights" graph it looks consistently low for a year and a half, then suddenly bends skyward in November 2008. Much of that is seasonal, but I believe the tactics mentioned above played a big part.