Hacker News Reader

Comment by tyingq | Hacker News Reader

Facebook has been caught dead to rights a few times, like this one:

"According to a recently published report, Facebook says they reach 1.5 million Swedes between the ages of 15 and 24. The problem here is that Sweden only has 1.2 million of ’em"

"Facebook’s Ads Manager says that the website is capable of reaching 41 million Americans between the ages of 18 and 24. The problem is there are only 31 million Americans of that age." [1]

Nothing ever happens, other than maybe some grumbling from paying advertisers.

[1]https://mumbrella.com.au/will-facebook-ever-stop-bullshittin...

View on HN

>Why should a person's mere nonexistence prevent you from serving an ad to her? You have built a scalable technology platform; it hardly seems fair that your growth should be limited by the human species's comparative lack of scalability.

(Via Matt Levine, tongue firmly in cheek)

http://www.thevab.com/wp-content/uploads/2017/09/Facebooks-R...

I didn't read your source. But I'm sure that for those age gaps the number of accounts could be explained by underage kids lying about their age to be able to create an account.

a lot of people create alternate personas for varying reasons. There are a sizable cohort of people that like to create accounts for online fantasy. also, there are many game accounts (some that even produce money)

trying to police them off would be a mistake by facebook.

>a lot of people create alternate personas for varying reasons.

I'm sure they do, but I doubt 30% of them do. That's what you'd need to balloon from 30 million to 40 million. The 30% also assumes that EVERY 18-24 year old American has a Facebook account in the first place.

Another way to slice those data - 33% don't have accounts, 33% (9.9 million) have just one account, and 33% have the remaining 30.1 million accounts spread between them - on average, 3.04 accounts per person in that group, but one or two people/groups may have thousands of accounts.

Edits: Correcting math. Thanks, kbenson.

Well, to correct it entirely, you should probably use thirds instead of 30%, since that was either an allowance for simplicity or an error originally. I believe the correct answer is just under 3 accounts per person in the group with multiple accounts.

30% of the populace has 3+ accounts? That doesn't make sense - it doesn't follow power-law distribution.

Most like the .01% have hundreds to thousands of accounts. What does that imply?

If somebody has hundreds or thousands of accounts, it implies they are running bots ... which I guess brings the whole argument back to the point that active humans with multiple personas are not responsible for inflation of reach stats.

You're implying that identities created using persona management software [1] are bots? I guess you could though they'd likely pass a turing test.

[1] https://www.dailykos.com/stories/2011/2/16/945768/-

I'm implying that if you have hundreds of accounts, you must be using software to manage it and aren't actively engaged with all of them yourself. Therefore, they are no better than bot accounts.

You'd be surprised. A lot of people are technically inept.

I've seen people create new accounts every time they lose their password. I also had a friend who'd create a new account every time she had a personality change of some sort (new hair? new profile!). Another friend owns a couple businesses and has accounts for each one (Pages are completely lost on him...). My aunt has entered her birthday incorrectly so she's supposedly 19 years old right now. A neighbor has a "family account", which is really just the wife posting pictures of kids.

Don't take "technically inept" to mean "my abilities, just a bit lesser". Take it to mean "no ability to reason about software applications whatsoever".

If they aren't logging into those accounts any more, then Facebook's advertising isn't "reaching" them. That explanation doesn't paint Facebook in a good light either, because presumably they are measuring how many impressions they serve and to who, right?

I agree, it's not good on Facebook to advertise inactive accounts (maybe they don't?), but it's not really in their immediate interest to do so.

facebook makes around 80% of its revenue from mobile

Start by assuming virtually every one below 21 that has an FB account has at least a couple of accounts - One with their public persona and one their parents won't be allowed to see. Add in accounts pet owners have for their pets etc. Then add in the bot farms. You'd get there pretty fast.

Alternative accounts are much more common on sites like Twitter and Instagram. I don't think it happens as frequently on Facebook.

That would be because Facebook actively makes it difficult for you to have multiple accounts.

It's not amusing to me that fake numbers can be used to justify a higher valuation, especially when those fake accounts are used to scam or mislead the "real" ones.

At some point should there be regulation on what terms like MAU means?

I don't think MAU numbers are being used to justify valuations. They may inform analysts' understanding of growth trends, in general. But, most analysts are looking at revenue, revenue/user, and margins.

This is the classic foible of calling for the mere existence of some regulation and assuming that is the same thing as the world really behaving according to some good rule.

Right now, investors researching Facebook stock have to be aware that published metrics include botspam. Suppose an authority imposed a regulation to force Facebook to publish "bot free" metrics.

Then what will happen? Legal will impose a demand on engineers, and the engineers will scratch their heads and do what it takes to satisfy legal, and legal will scratch their heads at the result and write up whatever report it takes to satisfy the regulator.

In the end, the relationship of the published number to reality will have changed in some difficult to predict, probably undocumented, way. Will investors then really have an easier time doing their research?

They haven't really been caught on anything. These aren't ad measurement results saying they did reach more people than exist, this is a reach estimation service. It's a technically complex problem that needs to take an arbitrary targeting spec with boat loads of dimensionality and return a reach estimate in 100s of milliseconds. It's based on a sampling method, of course. The reach estimation tool is a planning tool for approximations and nothing more. That's why advertisers aren't making a big fuss.

It's not that challenging to present real stats on active accounts.

OTOH, it's pretty tempting to use numbers already pre padded for you.

Reply on news.ycombinator.com