Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Facebook knows when you'll break up

cnn.com

Post reply on HN
Print view View original

11–19 of 19 posts

Re: Facebook knows when you'll break up

#11

Post by BrandonM »

While it's possible that more breakups happen on Monday, I postulate that Monday is actually just the day that people are most likely to report a weekend breakup. People who break up aren't rushing to Facebook right away. A breakup can be a hugely emotional event. If it happens on Friday, Saturday, or Sunday morning, it is likely to be followed by spending a lot of time with family, very close friends, or generally being in denial. When Monday comes around, it's back to the daily grind, it's time to face reality, and telling everyone you know is one way to come to grips with the situation.

I would also argue that the pre-Christmas spike likely isn't because people are cheap. I think the main factor is that most people spend the holidays with family. Two possibilities come to mind: 1) one or both of the people in the relationship decide that the relationship is not worthwhile enough to take each other home to Mom and Dad, or 2) meeting the family over Thanksgiving leads to a breakup (it was awkward, the family didn't like the S.O., the S.O. didn't like the family, etc.).

All-in-all, I agree with the sentiment of several comments here that between the poorly-ripped image and the unjustified conclusions they jump to, the article is mostly crap.

Re: Facebook knows when you'll break up

#12

Post by iamwil »

It would be interesting to be able to segment the graph by how long the relationship that was just broken up--mostly as a proxy for how long someone takes to recover from having broken up. Thus, you can find the best times of the year to be looking for singles.

If everyone took the same amount of time to get over the last relationship (which is not true), then we can just start looking for singles just a couple months after the spikes.

Re: Facebook knows when you'll break up

#13

Post by achompas »

tyrmored wrote: post #9

That's a pretty curious and probably flawed metric to choose. Surely it would be better to find the actual "is no longer in a relationship" updates? Sample size sucks too. Of only 10,000 status updates, how many would actually include those two phrases? I call bullshit.

I've been on Twitter for about 20 months and tweeted 1000 times.

Using that ratio (50 updates/month), we can estimate that their 10,000 status updates cover 16 people over a year.

My first instinct was "why not?" but thinking about this more, that looks kind of thin. Of course, I'm assuming a lot of things away...

Re: Facebook knows when you'll break up

#14

Post by ewjordan »

tyrmored wrote: post #9

That's a pretty curious and probably flawed metric to choose. Surely it would be better to find the actual "is no longer in a relationship" updates? Sample size sucks too. Of only 10,000 status updates, how many would actually include those two phrases? I call bullshit.

Is it possible they really meant "over 10,000 status updates" that include one of the phrases in question? Broken down yearly, 10,000 samples is an average of somewhere around 25 samples a day, and I'd imagine that on average far less than 1/25 of Facebook status updates are about breakups, so there would be only a handful of days with any hits, let alone enough to pick out a "Monday effect" or anything like that.

If it's 10,000 that match the query, then that's a more reasonable amount of data. Of course, the interpretation is still pretty shallow, and mere regex-matching for those phrases could be getting a lot of other crap, but at least it's a (just barely, given that they want to break it down day-by-day) decent sample size.

Re: Facebook knows when you'll break up

#15

Post by chronomex »

vaporstun wrote: post #4

CNN's crappy rip of the image cuts off everything past October losing some of the most important data (the lead up to the holidays and Christmas Day). They also cut off the months at the bottom so there's no way to accurately tell what is happening in what month. Here is their source site which has a better version: http://www.geekosystem.com/facebook-breakup-graph/

The original source (well, not quite, but I can't find a link to the actual source) has an even better copy: http://mathiasmikkelsen.com/2010/10/amazing-facts-about-face...

Re: Facebook knows when you'll break up

#16

Post by chronomex »

achompas wrote: post #13
tyrmored wrote: post #9

That's a pretty curious and probably flawed metric to choose. Surely it would be better to find the actual "is no longer in a relationship" updates? Sample size sucks too. Of only 10,000 status updates, how many would actually include those two phrases? I call bullshit.

I've been on Twitter for about 20 months and tweeted 1000 times. Using that ratio (50 updates/month), we can estimate that their 10,000 status updates cover 16 people over a year. My first instinct was "why not?" but thinking about this more, that looks kind of thin. Of course, I'm assuming a lot of things away...

It sounds to me like the 10,000 "status updates" include only status updates containing the keywords, probably from the Facebook search function.

Re: Facebook knows when you'll break up

#17

Post by electromagnetic »

michaelchisari wrote: post #5

All this shows is when people are statistically most likely to break up. This cannot be used to predict a specific situation (ie, "when you'll break up"), only to generalize about broad data.

Ah but it does tell you when to statistically 'work harder' on your relationship. IE Feb/March and Nov/December.

Although this can be compounded in a different way. The first break-up mountain appears after Christmas and spikes after valentines, translation: he buys crappy Christmas presents and he forgets valentines!

The second mountain is in Nov/Dec, translation: I spend all my money on her, she hasn't bought dinner in three months and now I have to get her a Christmas present too?!

Interestingly Aug/Sep/Oct were the months I noticed this year that I spent a vast majority of my time in doors. I'd gotten over the novelty of hot weather (I work outside) and had been cowering inside, which meant lots of movie theaters and dinners with the wife.

Re: Facebook knows when you'll break up

#18

Post by brianpan »

BrandonM wrote: post #11

While it's possible that more breakups happen on Monday, I postulate that Monday is actually just the day that people are most likely to report a weekend breakup. People who break up aren't rushing to Facebook right away. A breakup can be a hugely emotional event. If it happens on Friday, Saturday, or Sunday morning, it is likely to be followed by spending a lot of time with family, very close friends, or generally b…

I don't think that's the conclusion anyone is drawing. Most other articles use the phrase "announced on Monday". The journalist in the TED talk scraped status updates for "break up" or "broken up". This isn't supposed to be taken as hard evidence of anything.

Re: Facebook knows when you'll break up

#19

Post by BrandonM »

brianpan wrote: post #18
BrandonM wrote: post #11

While it's possible that more breakups happen on Monday, I postulate that Monday is actually just the day that people are most likely to report a weekend breakup. People who break up aren't rushing to Facebook right away. A breakup can be a hugely emotional event. If it happens on Friday, Saturday, or Sunday morning, it is likely to be followed by spending a lot of time with family, very close friends, or generally b…

I don't think that's the conclusion anyone is drawing. Most other articles use the phrase "announced on Monday". The journalist in the TED talk scraped status updates for "break up" or "broken up". This isn't supposed to be taken as hard evidence of anything.

I don't think that's the conclusion anyone is drawing.

From the article: Mondays, as if they weren't bad enough, are the most likely day to break up.

I was referring specifically to the shoddy article and its crap conclusions, not to the original data. The original data was interesting and potentially insightful.

Post reply on HN