Justin Smith, InsideFacebook (moderator), Blake Commagere Mogads / Zombies, Jason Beckerman TeachThePeople.com / Lotto, Jia Shen RockYou, Tim O’Shaughnessy Hungry Machine
a two minute history
since may 25 366M aps in the first 20 weeks
14M unique ap users in august
Invitations: originally no rules on invitations
no volume limited
starting to target
no built in stats
now need social incentives for invites
News and minifeeds introduced sept 2006
broadcasts your activity
worries about privacy
feeds: you can optimize
but selection algorythm is not published, depends on individuals, no built in stats
>only friends with the ap see your feed items
friends of logged in user or anyone with the ap
rules have not been well articulated, some people are abusing and getting blocked as too spamming
can get shut down.
Blake: facebook is getting better about letting you know what changes are coming. my focus was optimizing invites, and I’ve been using the standard invite interface provided. people have tried different stuff, but instead I’ve focused on how would this new invite control work vs. the old one. it’s worth doing A/B testing.
For notifications, as a mechanism for viral spread, I didn’t really use it, and i tried once and i went and played frisbee and my ap was blocked. my users were too eager to bite people.
Tim: good or bad, we lost the massive growth provided by invite process. it’s not that invites are not important, but if you look at what we’ve called up there is a decision point and they can choose skip. Notifcations, it can show up without being marked as spam.
Jia: form looking at all the different channels, invites, minifeeds, minifeeds is the only way to grow it outside fo the users. form an invite perspective we’ve spent time to make sure the selection process is fairly easy. Most people call it an invite process, but it’s really a way to spread the application. if it’s an event it’s invite, but if its zombies, its a bite, or a gift a gift…
minifeeds, we’ve tried changing the graphics that accompany and has a big corollary on how often people click through. tuning the images will improve your throughput by far.
People who have 3-20 wallposts are more likely to accept invites, people with real relatoinships accept, just data to support the theory everyone has put in your head.
Jason: bonus functionality works, when you invite more people you get tickets for the daily jackpot. they designed for daily engagement, the jackpot goes up every day, it’s good facebook measures engagement for them.
they’d like rollup messages, rarer than having every single activity in the feed.
_insert lunch drowsies. notes getting thinner…._
_again I’m blown away by how these folks study and tweak. they put major corporations to shame. a/b testing, user research and more._
Jason: if you could message your users, that’d help, even if it was only one or two times a week.
Justin: do you know what the future of analytics is on facebooks? what are you doing?
Blake: I’m grateful everytime they add anything, such as recently on pageviews. Some of these issues are a bit opaque because you are going through facebook to the user. I need metrics where I can measure activity so I can learn what features will engage. I don’t do as much a/b testing as I should. I know you should do it like on invites where it’s the biggest bang for the buck.
Jason: we just built out stats, because we want to understand where our invites are going, is it core users who really want bonus tickets, ro invites that never got used. it’s been really cool to have that data.
Jia: dont’ go and overdesign a anayltics sytem. we still do real time mysql system, eventually we’ll do somethign more but raw numbers speak for themselves. make sure you collect that information properly.
Tim: there are raw numbers we dont’ want to relay on facebook for, but then there are things facebook will build and we dont’ want to waste money building it.
Blake: dont’ spend a ton of time making something beautiful and complex since they (facebook) know what our pain is and will get it to us, what if you spent a month on analytics and you didn’t need to.
Jason: we focus on the data facebook will never be able to tell us.
Blake: We’d all like to know what the algorithm is for whats spam, but I understand they don’t want us toeing the line between spamminess. I odn’t think that algorithm will ever be shared, but we all have insight into ti, # of installed users or engaged, then number of notificaitons, then how many get marked as spam. you shoudl think carefully about notifications and think about if yoru toeing the line. I limit it even if the uses are crazy active, thinking I know that would annoy me…
Jia: we’ve gone through a lot of tuning and focused on only notifying when there is something useful, and blake and play frisbee together and we’ve gotten blocked and you have ot sit on your computer to see if your ap has gotten blocked, you don’t want to sit on your computer and watch it.
Blake: sometimes you are sending out only a few notifications and you see your spamometer going up, a few users can really shift the tide, facebook users have a low tolerance
Jia: when we launch a new ap we don’t use notifications.
Justin: what if you could show to non-installed users in the newsfeed
Blake: don’t underestimate the power of the newsfeed. if you had a clear argument for the ap on teh newsfeed it’s change things
Jason: we wouldn’t have to do invites, if people saw their friends winning money
Jia: I really couldn’t see us change our call to action in the minifeeds, I don’t think it’d change our strategy.
Tim: I think we’d see it as another kind fo invitation, we wouldn’t’ change strategy much.
Q: how far can these go with non-viral applicaitons? werewolves are naturally viral, BUT…
blake: aps that are not inherently viral can’t be made viral by optimizing the heck out of it. is there a reason for someone to want their friends to use it? is it so cool you get street cred for finding it? sharing photos, sharing music, because uses have a direct benefit fom it. You can’t make an indea viral, but it can make the difference between seeing a a good idea flounder.
Jason: I don’t know if it has to start as being viral if there is a value proposition, it can become viral.
Jia: that addresses the question of when will aps become utilities? they just won’t grow as fast, but they could still growth. opening up the minifeed could help utilities.
Tim: you can’t dress up a pig. but at least you can fail fast and cheaply. you can try the methods, but if the idea isn’t solid, no amount fo virality will help you.
Q; do you know what the drop off is with inviting ten or installing aps
Jason: we require it so we dont’ have a good number on that. but the growth is showing that people are using it.
jia: unfortuantely in the previous world, people just clicked hte next button. but people are getting more savvy. limiting to ten has been a good ting, because people are less pissed off and ignore everything.
Tim: look at the growth rates after the limit, it dropped dramatically.
Jia: if you have other incentives, then it’s good to have the invite at the top. but for things like events, where people want to use it over time, it just annoys people.
q: how many gets uninstall?
Blake: vampires gets the highest uninstall of all my aps, and its 13%, which is a good number.
jia: exposing the install rate is interesting to brand advertisers, but better would be how long to users keep the ap. it’ll end up being like total uniques, etc. those are the stats that really matter.