Hacker News Reader

Comment by NathanKP | Hacker News Reader

    Facebook's Ads Manager claims a potential reach of 41 million 18- to 24-year olds
    and 60 million 25- to 34-year olds in the United States, whereas U.S. census data
    shows that last year there were a total of 31 million people between the ages
    of 18 and 24, and 45 million in the 25-34 age group, the analyst said.

I doubt that Facebook is purposefully lying about their numbers, but the fact that they estimate their reach to be greater than the census results means there must be a lot of bot accounts on Facebook.

Assuming that Facebook isn't lying, and they actually see as many accounts as they claim to reach this data would suggest that at least 25% of the accounts on Facebook are alt accounts or bot accounts. And that is assuming that everyone in the target demographic who was in the census is on Facebook. Facebook must be greater than 25% bots.

View on HN

Lying aside there's still a point at which this is negligence. When your value prop is a set of numbers, and that number is obviously, impossibly incorrect, at what point do you have a duty to try to correct that number? At what point do your customers have a legal leg to stand on?

Not only that i had a Facebook campaign that consistently under matched Analytics reports. Either Google under reports Facebook clicks or FB overstates theirs.

Doesn't FB require you to be 13?

What are the chances that the self reported 18-24 group is full of younger teenagers and kids?

Just like the age checks on Youtube or Steam.

I'm 101 years old, per Valve's superior analytics :)

yeah, I just read the other day about a 3rd grade teacher in North Dakota or somewhere that has all her students posting on Twitter and Instagram and such... Pretty sure I was 8 in third grade. I was wondering about the 13y.o. requirement.

BOTS. I believe Facebook/Google/et al can and do detect bot traffic, don't believe me? Write a simple scraper to record google search results. They just don't use these techniques when applying it to their own revenue generating ads.

> I believe Facebook/Google/et al can and do detect bot traffic, don't believe me?

How? Clicking an ad would lead you to a CAPTCHA before going to the target site?

And a lot of the bots are probably primarily used for other services like Tinder which require Facebook logins.

Easy. If it is a bot - don't charge. Probably a few lines of code. if ( is_bot() } { charge = 0 }. Also, be transparent: you could even show a report of how many bots it removed. BUT this would cut their ad revenue to ~0.

> if ( is_bot() } { charge = 0 }

That's some funny pseudo code. The assumption being they can write an is_bot function that is reliable.

More like:

if (is_bot_probability >= 90%) { charge = 0 }

A lot of people in this age bracket have 2 active accounts for work and personal.

Yeah, I doubt this is malicious.

Lots of grade school aged kids have an alternate Facebook - one for parents/grandparents/relatives, one more locked down for their friends.

Folks who have to be on FB for work often make a second, work-only account. I still see plenty of small businesses running a "personal" profile instead of a page.

This would suggest that an enormous amount of facebook accounts were alt/bot accounts. I simply cannot think of a way that Facebook would be completely oblivious to this, at best this is severe negligence from them, and at worst outright fraud.

I'm sure they're aware, but can you propose a way to prevent this without massive false-positive rates? There's not going to be a very easy way to determine if "Jane Lastname" and "Jane Middlename" are the same person or just roommates who happen to share a first name.

It really is for the advertisers to determine if a "like" from "Jane Lastname" AND "Jane Middlename" is worth anything to them. Orrr, are they throwing money into a blackhole.

So do adults that lead a less-than-traditional lifestyle or otherwise would like to separate work or family from their larger social life.

I've mostly seen this with occultists and LBGT community, though I have seen a few artists do the same just to keep fans and intimate folks separate. Understandable, since simply saying one has had a bad day at work can be taken as a personal insult by some (I've seen it happen).

I'm European, and I have a bullshit US address registered on my Facebook, just to trick the bots etc.

Also possible that FB and Instagram might have some double-counting of personal accounts occurring as well.

More importantly, I'm curious whether there are any ad managers who actually weight those estimates heavily for more than directional guidance as to whether they have enough reach to be worth testing and getting actual benchmarks. If so, I'd love for them to chime in with how they use that data.

I don't rely on those numbers for anything critical as getting benchmarks from a live test is really the best way to go about it.

I run mostly direct response campaigns, so conversion focused. Take it for what it's worth.

I would never trust the numbers reported in any ad interface to accurately reflect reality. But, why should I care if it reflects objective reality? I'm only trying to get X conversions for $Y.

For me, building the "fraud" or poor accounting into my economic model is how you play the game. I don't care if FB sends me 500 visits at $1 each, with 20% bots. I just care that I spent $500, for $conversion_result.

People who do branding campaigns may feel differently because they're reporting on how many humans saw their message. I don't think that's a big deal, but I also think branding campaigns are a waste of money anyhow.

Fully agree with your view on the DR side of things. There are obviously questions as to the incremental lift of a given view-through conversion (especially for campaigns where the vast majority of conversions or VTs).

However branding campaigns definitely can work, they are just hard as hell to measure somewhat accurately without LOTS of budget, fancy software, focus groups and statistical modeling. You need a large amount of data to notice the needle moving, and that just isn't always feasible.

Reply on news.ycombinator.com