6 ms·

Hmmm a few years ago a friend who worked there told me they experimented with having two tabs: one with their machine learning algorithm, and one that was like

Hmmm a few years ago a friend who worked there told me they experimented with having two tabs: one with their machine learning algorithm, and one that was like Twitter and just had 'everything, by everyone, in chronological order.' Do you know what happened? <i> Duration per user on the site went down </i> by a statistically significant margin. One of the things they do well is micro-optimize many different features and use machine learning & data analytics to provide the "best" features, whatever that may be. What they DO suffer from sometimes is lack of prioritization of the correct features (like messaging, removal of certain features, pages products, etc.)

wheaties 13y ago
The problem with time on site is that it can also mean "time spent looking for what I want to see but can't 'cause I can't find it." For a commerce site or a saas site this is terrible user experience. For FB, it's terrible user experience too but a great metric to show advertisements. What a bad incentive they have.
pcl 13y ago
Duration on the site might not be the right thing to measure. I mean, I guess that probably turns into more dollars for them from ad impressions. But not could be that time on site was shorter with the other model available was because their users got what they wanted faster. Intuitively, this doesn't seem like it would be the case for the change described, but I hope that they are measuring some approximation of user engagement levels in addition to just time on site.
DanBC 13y ago
I spend several minutes everytime I use Facebook wondering how the hell I can achieve something, or where on the page the button is hidden, or what this inscrutable icon will do. "Time on the site" is similar to the supermarket trick of moving items around to make customers look for things, thus exposing them to more 'buying opportunities' of stuff they wouldn't normally look at. It's probably great for them, but it's not so fun for the user / customer.
- e12e 13y ago
  It's funny. when I started using fb I just kept an eye on the email notifications: a permanent stream of all activity (including posts later deleted by the poster). It worked perfectly for me; it was before the worst of the farmville-style app-spam -- but I'm sure killing that was just a (bayesian or not) spam-filter away if I'd felt the need. With what fb has morphed into, I generally assume the only way to share stuff with friends (in a way that they'll actually have the opportunity to see it or choose to ignore it) is to tag them or share directly on their wall. In other words; I might as well send an email with a couple of cc's tacked on. I don't really mind that they're sabotaging their own product; I much prefer a decentralized, self-hosted web. It does mean I'll see less from many friends though - people that (rightfully) don't bother setting up a home page.
jellicle 13y ago
> Duration per user on the site went down Well, yeah. Because people were finding what they wanted in a shorter time frame. Which grocery store do you want to shop in, the one where everything is hidden and there's only one cashier so you have to spend a long time there, or the one where you can find everything quickly and there's 20 cashiers open? The marketers say, "store number one". I say number two is better.

wheaties 13y ago

The problem with time on site is that it can also mean "time spent looking for what I want to see but can't 'cause I can't find it." For a commerce site or a saas site this is terrible user experience. For FB, it's terrible user experience too but a great metric to show advertisements. What a bad incentive they have.

pcl 13y ago

Duration on the site might not be the right thing to measure. I mean, I guess that probably turns into more dollars for them from ad impressions. But not could be that time on site was shorter with the other model available was because their users got what they wanted faster. Intuitively, this doesn't seem like it would be the case for the change described, but I hope that they are measuring some approximation of user engagement levels in addition to just time on site.

DanBC 13y ago

I spend several minutes everytime I use Facebook wondering how the hell I can achieve something, or where on the page the button is hidden, or what this inscrutable icon will do. "Time on the site" is similar to the supermarket trick of moving items around to make customers look for things, thus exposing them to more 'buying opportunities' of stuff they wouldn't normally look at. It's probably great for them, but it's not so fun for the user / customer.

e12e 13y ago
It's funny. when I started using fb I just kept an eye on the email notifications: a permanent stream of all activity (including posts later deleted by the poster). It worked perfectly for me; it was before the worst of the farmville-style app-spam -- but I'm sure killing that was just a (bayesian or not) spam-filter away if I'd felt the need. With what fb has morphed into, I generally assume the only way to share stuff with friends (in a way that they'll actually have the opportunity to see it or choose to ignore it) is to tag them or share directly on their wall. In other words; I might as well send an email with a couple of cc's tacked on. I don't really mind that they're sabotaging their own product; I much prefer a decentralized, self-hosted web. It does mean I'll see less from many friends though - people that (rightfully) don't bother setting up a home page.

jellicle 13y ago

> Duration per user on the site went down Well, yeah. Because people were finding what they wanted in a shorter time frame. Which grocery store do you want to shop in, the one where everything is hidden and there's only one cashier so you have to spend a long time there, or the one where you can find everything quickly and there's 20 cashiers open? The marketers say, "store number one". I say number two is better.