[ballenf]

How Facebook determines ad placement in news feeds:

> Millions of auctions take place every minute as users across Facebook load their feeds. ... The algorithm is constantly learning, using past results to inform how it weighs bids in the next auction. The intent, Facebook says, is to maximize value for everybody: to pair the advertiser with its likeliest customers, and to show ads that users want to see. And, of course, to make money for Facebook. [emphasis mine]

So FB wants:

1. advertisers to be successful so that they come back and attract others;

2. users to remain sticky and not be turned off or too distracted by an ad; and

3. sell each ad space at the highest price, so long as 1 and 2 remain true.

Interesting business they're in and an incredibly challenging (and fun/rewarding) algorithm to code.

[amelius]

The problem is that it leaks information. If I place an ad for paperclips targeted at cat-lovers, then if you click on my ad and buy paperclips, I know that you are a cat-lover.

[d4l3k]

They already allow you to target groups like that. The algorithm doesn't leak anything more.

[amelius]

I'm not sure what you mean. It leaks the targeted group, that was the point.

[mbesto]

I'm not sure I understand your point. It only leaks a subset of data (cat lovers who have an intent to buy paperclips) and is hardly valuable at scale to anyone for anything other than prospecting new customers.

[sdenton4]

I dunno, identifying groups like "Jew-Haters" seems useful beyond "prospecting new customers." For example, if the US government cared about tracking hate groups (as opposed to hiring them into the administration) they could use it to identify people to track.

[dingo_bat]

I wonder where we'd be today if all this effort was focused on something like automatic cars. Instead of spam.

[CalRobert]

Good question. I graduated college and desperately wanted to work on renewable energy research. After scads of rejections I quit ignoring the expressions of interest in my programming abilities.

I still wish I were doing something that actually bettered the world, but I have bills to pay.

Then again, maybe ensuring people graduate in debt is a great way to ensure we as a society do what folks with cash want us to instead of what we feel we ought to be doing.

[sebleon]

"The best minds of my generation are thinking about how to make people click ads" - original FB data scientist

[klokoman]

Because it is one of the few sectors where you can be a little creative and not obey the rules set by corrupted bureaucrats decades ago

[chii]

And instead follow rules set by investors or corporate vested interests.

[ISL]

Spam, spam, spam, spam, cars, spam, and spam.

There's not too much spam in it.

[nihonde]

And yet, advertisers still have no meaningful ROI, because it's never in the platform's interest to overachieve on this count.

And, users are still very much annoyed by distracting, irrelevant, or overly-targeted ads.

[maxpupmax]

Why do you say that there's no meaningful ROI? I know a lot of marketers and advertisers who have built great businesses off of Facebook Ads.

Agree that we need to improve the quality of ads across the board.

[nihonde]

I will answer your question with another: what metric are you using to measure ROI?

Edit: When you say you know lots of "marketers and advertisers" that are making money from ads, are you referring to resellers and providers of ad inventory and/or analytics, or are you referring to people who actually sell goods and services other than marketing and advertising inventory?

[akanet]

Yeah, this is a very strange statement. Many businesses are printing money advertising through Facebook.

[nihonde]

Then you should be able to answer my question: what is the metric for ROI?

[csallen]

Metric for ROI is profit from a customer vs cost to acquire that customer via Facebook ads.

[jaggederest]

Return on investment for advertising is very simple:

net present average customer lifetime value acquired per unit of advertising cost

If it's over unity, you're making money with advertising, if it's not, you're wasting money on advertising or need to increase your CLTV or decrease your discount rate.

[matsur]

People entering and converting through a purchase funnel?

[nihonde]

OK, so how is the success of that funnel measured vis-a-vis other promotional channels? And are you able to demonstrate definitively that a visitor from FB concluded a sale? If so, how?

[traek]

> OK, so how is the success of that funnel measured vis-a-vis other promotional channels?

What difference does it make? If it’s profitable, it’s profitable.

> And are you able to demonstrate definitively that a visitor from FB concluded a sale? If so, how?

Through the variety of existing measurement solutions, or a custom solution.

Twitter: https://business.twitter.com/en/help/campaign-measurement-an...

Facebook: https://developers.facebook.com/docs/facebook-pixel/pixel-wi...

Google: https://support.google.com/adwords/answer/1722054?hl=en

I don’t understand why you seem to think that companies are throwing tens of billions of dollars a year at Facebook (and Google, and Twitter) with not only no returns but also no ability to measure that lack of returns. It’s just such a ridiculous viewpoint to hold.

[nl]

OK, so how is the success of that funnel measured vis-a-vis other promotional channels?

It's trivial easy to add a tracking id to this (in-fact I think it is on by default)

And are you able to demonstrate definitively that a visitor from FB concluded a sale? If so, how?

I think people are a little confused about if you are trolling or you genuinely don't know.

Anyway, https://developers.facebook.com/docs/facebook-pixel/pixel-wi... (and plenty of other methods)

It's really, really easy. And that is why Facebook and Google are printing money - vendors know it works and can measure exactly how well, so they throw money at them.

[jaggederest]

Yes. Facebook gives you a unique visitor id when they see the ad that persists and is included in the sale form. It's completely linear, and is tracked from beginning to end. Every click is tracked, and users are tracked even if they leave the website and return later.

There's hard, direct evidence involved in every step of the way.

[leggomylibro]

Maybe users just don't want to see many of these ads, full stop.

So they're really pairing the ads with people who are least likely to complain. That kind of shifts how I view the incentives.

[golergka]

> And, users are still very much annoyed by distracting, irrelevant, or overly-targeted ads.

How would you know, if only users who are actually annoyed are vocal about this?

I, for once, mostly see ads from the companies which I am already a happy and enthusiast customer of, for new versions of products that I enjoy. I even follow the industry news to know about new versions of these products and their reviews - and when I can't buy them, I genuinly enjoy watching youtube reviews about these products.

Do you honestly think that a person like me would be annoyed by these advertisements?

Don't you think that a _lot_ of people can say the same about their own product category that they're enthusiastic about? In my case, it's music software and hardware, but it could be cars, power tools, clothing or some other hobby that people enjoy spending their money on.

Sorry for the rant - I'm just tired of seeing the notion about how bad and annoying the ads are, when I see zero confirmation of this in my reality.

[nihonde]

I get it. I like ads in Japan because they're entertaining. You're the first person I've ever hear of who enjoys being a target of online ads, so I'm still convinced that you are very, very unique.

[adambyrtek]

Work on real-time ad auctions is definitely technically challenging, but I'm not convinced about the fun part, and I definitely wouldn't describe that as rewarding (unless you mean that in a purely financial sense).

[ballenf]

I was thinking in terms of having the results be immediately measured and quantified. It's kind of a luxury to have such a quick feedback loop. That's all I meant by rewarding, wasn't thinking in financial terms even though that now seems like a more obvious implication.

[brianpgordon]

Given the number of levers they have available to adjust in terms of tuning factors to maximize engagement, and given the changes constantly going out to their platform, and given the changing landscape of the ad market itself over time, I doubt that it's exactly clear to a developer what they're doing right and wrong. Facebook probably has a full-time staff of data scientists to extract conclusions out of that pile of analytics data.

[wsinks]

Sounds like Asimov's 3 laws.