Head of Design at Facebook: How Her Team Uses Data To Make Decisions

http://www.zurb.com/article/515/podcast-of-julie-zhuos-talk-on-how-facebo

by mikemcaveeney • 16 years ago

49 18 16 years ago

notmyname• 16 years ago

One thing that is implied is a strong understanding of who the users are and what they want from the product. I'm always encouraged to hear of large, influential companies talk about these sorts of things because I hope that these practices become more standard. Understanding users rather than guessing at what they want and measuring the effects of changes made seem to be obvious, but I'm always surprised at how often it isn't done.

away• 16 years ago

Perhaps the founder does not want to pander.

lpolovets• 16 years ago

FTA:

"Facebook had a really boring deactivation page when you close your account... [One of the designers created a page] which has all your friends waving good-bye to you as you deactivate. Give you that final tug of the heart before you leave. This reduced the deactivation rate by 7%."

While that's a nice reduction in deactivations, it also strikes me as somewhat manipulative. It feels like many of the "social" companies see users solely as metrics to optimize, not as human beings.

snprbob86• 16 years ago

You say "manipulative", I say "persuasive".

Is it manipulative to optimize the tag line on your landing page? You're trying to persuade a visitor to buy your product. Or are you trying to manipulate them to buy your product? The same principal applies to keeping customers.

I'd call this manipulative when it becomes disingenuous. For example, had they put actual speech bubbles above my friends' heads and led me to believe that they actually were asking me to stay. Until then, this is roughly akin to saying "Hey man, don't leave the party yet. We haven't had cake yet!" To which I'd reply "OK, well twist my arm why don't you" and then eat some cake.

exit2• 16 years ago

> You say "manipulative", I say "persuasive".

i say false dichotomy. you may have persuaded your friend to stay at the party. you didn't do her any favors if she has to get up early tomorrow. a user at that final screen has clearly made up their mind.

also, saying "* will miss you" is not too far from putting "i'll miss you" in a speech bubble above them. i wouldn't miss >%50 of the people on my fb list and it misrepresents me to say i would miss them.

kurtosis• 16 years ago

Well the deactivation page did automatically say that the users "friends" would "miss" them. Did they actually check with these users to see if they would want their image and identity to be used in this way? If not, then it seems pretty disingenuous to me. How do they know that these users will actually miss them? Why is facebook assuming that my friends would not respect my decision to leave facebook if that was the case..

dhs• 16 years ago

If I'm told by a machine that "X will miss you", I do consider that disingenuous. There's a programmer using a program, trying to get an emotional reaction from me by the entirely speculative claim that another human being - probably emotionally close to me, but more probably not emotionally close to them - will miss me, suggesting an emotional connection human->machine->human where none exists, trying to exploit my natural tendency to anthropomorphize. And that kinda rubs me the wrong way.

exit2• 16 years ago

farmville is basically one long manipulative tug

izaidi2• 16 years ago

It's interesting that the only time Facebook seems to want to offer you any kind of emotional experience is when you're trying to leave the site, and the experience is designed to be a negative one, rooted in guilt.

akd• 16 years ago

I find that checking Facebook is almost always an emotional experience. I see so many more new babies than I ever did before Facebook.

kgrin• 16 years ago

Interesting contrast between this and the stories about teams at Google gathering data on optimal line widths or hues of blue...

swah• 16 years ago

How is this generally done? Do they add JS to every UI element to log what the user does and send this info with the requests, or they can find that out only from the http server logs?

dnewms• 16 years ago

I wonder how many users visited that deactivation page just to see "x will miss you" once they heard about it, without any intent to deactivate.

sasmith• 16 years ago

Excellent point, but it shouldn't impact an A/B test. Specifically, the experimental and control users should be equally drawn to look at the deactivation page.

However, this could give a false impression of improvement if you just look at the data before and after the launch. It's a good reason to also look at the total number of deactivations / visits to the deactivation page.

mwexler• 16 years ago

It's not just enough to have or even "use" data. They carefully chose metrics that matter. In their business model, for example, they saw that photo sharing was a primary driver of longevity and of driving additional usage down the social graph.

Therefore, the metrics where carefully chosen to foster those behaviors. Having a goal and then choosing measure to support it is surprisingly rare, not just among designers, but across businesses.

And just picking things like "Volume sales... and Profit!" as your metrics are kind of specious. Yes, every business wants those, but that just delays the real question: what behaviors by your customers will make that happen for you? And once you think you have those, look for metrics that diagnose that those behaviors are happening, or can help you diagnose why they aren't.

We all have data. Picking useful metrics and linking them, well, that's insight.

Head of Design at Facebook: How Her Team Uses Data To Make Decisions

Related Stories