I don't see the evidence here as being compelling at all.
Even if you could prove that Facebook is filling the gaps between new users and churned users by creating fake accounts, you also have to draw a line between those fake accounts and the company's bottom line.
Fake accounts don't buy products, so if advertisers are making decisions off of CAC, numbers potentially inflated by fake accounts (reach, clicks, engagement, etc.) are secondary.
Facebook's ad power comes from their egregious data collection and lack of privacy concerns. The more they know about you, the more relevant your ads will be and the more likely someone seeing your ad will be to buy. And if advertisers can draw a line that says "if I spend $50 on Facebook ads, I'll increase my bottom-line line $100", they'll spend money until it's no longer profitable to do so.
That said, it's not like all Facebook advertisers are acting rationally in that manner. I don't know what % of Facebook revenue to coming from unsophisticated advertisers, so it's possible that they could be making a lot of money from people focusing on fuzzy metrics like reach.
> The more they know about you, the more relevant your ads will be and the more likely someone seeing your ad will be to buy
How much ACTUAL evidence is there for this?
Take Amazon for instance... they know everything I have bought from them ever and employ some mega-smart people but the number of times I buy something and they then recommend either absolute shite or the same damn thing is not an insignificant number.
Another thing is if I am an advertiser and FB says "your ad was seen by 20 million people", how can I dispute this? It's a closed system.
I am sorry... but from my comfy armchair position I'm calling shenanigans on mass-data collection in advertising being able to move the profitability needle in a measurable way.
I think you are missing the point here and thinking too much about pay per impression. Professional ad buyers don't measure their ad effectiveness in "people reached". That is for chumps who want to be ripped off, and large companies that have too much to burn on their marketing departments.
The people who use ads effectively are buying pay per click and measuring the real conversions to measure effectiveness. Bots don't buy the products after they click through, but the real clicks often do. (For some categories you can get as high as 10% conversion rate to a real sale from real clicks that you buy because ad targeting on platforms like Facebook and Amazon is really, really good. If your ad is good and your product is good then the ad targeting will get your ad in front of the right people a lot of the time.)
If you are smart you don't pay for impressions, you pay for clicks and conversions, and you measure how many of those clicks convert into sales to make sure you are coming out net positive. Sure there are always going to be some fake clicks but there are still enough real clicks that convert to make it worth it. A bit of click fraud is just the cost of doing business.
Lets say I'm selling a product that has a 20% profit margin. I might be willing to pay up to 10% of my profit margin on ads if the increased number of real conversions that result will more than double my sales, allowing me to make more money overall than I would have otherwise.
There are definitely a lot of people out there who are foolishly wasting money on buying impressions, but there are also tons of people making ludicrous amounts of money from well targeted ads that convert into increased sales of their product
You are over-simplifying. Almost all of the metrics Facebook will feed you when you setup an account are on a view-through attribution basis. Yes, there will be some clicks, but if you attribute revenue from Facebook by clicks only, it will look terrible. This is an old problem for anyone that has spent time with display ad platforms. So, you install a facebook pixel on your checkout, and if the person making a purchase saw your ad on Facebook, Facebook's reporting console will take credit for that. Thats view-through attribution and it highly overstates the credit Facebook should receive for making the sale, but there's not much you can do to measure it otherwise. So what do you do?
The answer is, you setup a holdout test. Facebook has a black-box solution for helping with this, but it is a huge conflict of interest to let Facebook handle your test. The thing to do then is set up the test on your own platform. That is a massive undertaking, and any business that doesn't have a significantly sized engineering and analytic team doesn't stand a chance.
So yea, people look at things like reach, and view through conversions, and it would be very hard not too or they might not have the means to do something more sophisticated. Fake accounts undeniably puff of numbers for these people, and they should be taken to court for not giving a shit about that fact.
Facebook are 100% not going to lie to you through their experimentation platform. The whole point of it is to help people accurately measure the results of their ads.
In general, it will be extremely difficult to run accurate tests on FB (or indeed Google) ads without using some of their infrastructure, as these platforms can balance users appropriately in terms of likely response to your ads (for instance, ensuring that both ads are shown to people with approximately the same click probability).
Without this, you run the risk of making bad decisions because one ad got served to users who were far more likely to convert.
Of course you should definitely run your own tests, but I wouldn't completely discount the platform's tools as they do have advantages which are impossible for you to replicate (i.e. balancing users in terms of response rates in each condition).
Yea, but what are they using to analyze your results? You literally have no idea how robust (or not) their statistical methods are. It's been my experience that 3rd party solutions are notoriously shoddy in this department. They have no incentive to be robust.
AFAIK (it's been over a year), they use intent to treat approach, and only remove users from the experiment after they have already been matched to an ad. This allows them to ensure that the users are matched in terms of expected conversion rate, which allows for valid results.
I believe that there were rumours of using bayesian analysis in the future, but I believe it's a two-sided p-value based on the differences in your chosen outcome.
I worked in ads.
You individually is not what makes advertising useful or lucrative. Aggregate numbers do.
How many people see ads on tv, and buy the thing they saw right away? You guessed it, it is probably less than a percent.
There are so many impressions in the world, from billons of users. You cannot expect each of them to do meaningful conversions. On a super off the cuff calculation, assume each user see 100 ads a day, 2b users makes it 200b impressions. If each user generously converts once a day, that is 1%. If they convert every other day, it is 0.5% per impression.
I particularly left clicks out went straight to conversions. Usual flow goes like this: query, matched query, impression, click, conversion... The funnel is super large and drop offs are huge.
Are ad-hating HN users ever going to say that ads might work sometimes, instead of relying on random anecdotes and 0 data or facts from an extremely small and biased sub-population who is never the target of advertising campaigns anyway?
Or are you really arguing that >$500B businesses like Google and FB are complete fraud?
I hate ads, but I agree with you. I hate them because they work. Or at least, all this money wouldn't be flowing into the industry if it really wasn't working. Maybe FB is fraudulent due to the accusations in the article (inflating numbers to advertisers), I don't know; that's unrelated to the issue of giant, innovative tech companies basing their entire business models on advertising things to their non-paying users, and visitors, and also anyone else who ever visits pretty much any other website on the Internet.
Advertising isn't fraud. Neither is data collection. It's just unfortunate. I wouldn't even be surprised if many Google and Facebook employees agree it's unfortunate. They're probably thinking: what else exactly can we do to sustain all these non-paying users?
It's really not even the data collection that bothers me so much as the fact that it's being used for such a... dumb purpose, I guess. The best minds of our generation are dedicating their life's work and all of their brainpower to thinking about how to get people to see or click more advertisements. What a waste.
I didn't claim that ads don't work... far from it: I've seen how easily people are persuaded to part with money so I believe ads DO work!
What I have issues with is using masses of data to persuade someone to buy something having any more significant ability to make them part with their cash than doing it contextually.
I realise I said "measureable effect" and I kind of meant that in my original statement but I also mean measurable enough that my privacy is traded for an extra fraction of a percent.
By way of example, lets say a contextual banner ad at the top of an article for the new Hyundai i30 Fastback N (I just bought one of those :) with a discount code to buy one might persuade 10% of people to click and 10% of them to buy. (Illustrative figures, don't throw your back out over their accuracy)
If adtech companies hoover up data across the web about my buying habits, browsing habits, sexual habits etc. can they legitimately say they can make a measurably larger percentage of people buy that car (or handbag, shoes whatever)?
And if so, what kind of numbers? Is it 10.1%? 30%?
And is that trade-off worth me allowing them to do that?
The issue isn't making a larger percentage of the people buy, it is a larger percentage of the people advertised to buy.
Alcohol companies don't want to waste their advertising money on people who belong to a region that doesn't consume alcohol. Car companies don't want to advertise to the poor with bad credit - by contrast scammy used car dealers want to advertise to those with bad credit and not reach the rich.
Of course there is a real downside to this: you will miss someone who should be a target because they look like the type of person is not. The person who just left his restrictive religion. The poor person who is starting what will turn out to be a successful business. Thus big companies often to have a component of ad budget that reach everybody without concern for who buys - just to make sure they don't completely miss someone worth targeting.
It works (Has a positive RoI) up to a particular spend/impression, and doesn't work after a particular spend/impression.
The purpose of all the data collection/targeted advertising/remarketing, is to increase the efficiency of these ads, and to increase the threshold when it goes from 'works' to 'doesn't work'.
To put it another way - I can't think of many businesses where showing ads for 1 cent / 1000 impressions would not be worth it. I can't think of many businesses where showing ads for $1000 / 1000 impressions would be worth it. Depending on the business, there's some inflection point between those two numbers. More targeted advertising makes the implicit promise that the inflection point for <your particular business> is higher, then in an untargeted advertising platform.
What incentives are there to claim that their company can sell your company access to their secret sauce which greatly improves some metric which is notoriously hard to measure?
It's very easy to measure the success of online advertising campaigns, because you can directly measure how many people your ads are reaching and which of those people end up buying your stuff.
Even if it weren't, I still don't see the incentive. If you don't think that it matters what your secret sauce actually is, why use mass data collection rather than something more popular?
No it is not easy to measure. You can measure immediate success, but not delayed success. I just bought my first order for a product that I last saw advertised on facebook a couple years ago. At the time the product looked interesting but didn't fit my needs, but a couple years latter and I remembered them. Since I've moved and used firefox directly to access their website it will be very hard to put my order with the ad (they have probably closed the books on the original ad where I'm marked as a click but no buy)
I was all geared up to play devil's advocate here, but any response I could formulate seemed to distill down to data being extremely powerful, and so undermined the thesis.
Still though, if Facebook has solved the "50% of my ad spend works, I just don't know which 50%" issue for companies like Mercedes Benz and Coca Cola then that would be mighty impressive.
I'm working with comscore right now (an aggregator of stats for advertisers) and can confirm it's all a house of cards. there's no quality controls or oversight into what's reported.
In my companies and for my clients, I track up-tick in sales for each campaign on each platform. Sometimes the problem is conversion (that's on me/us), and sometimes bad targeting or whatever.
If FB isn't one of the lowest channels (platforms) for CAC that is greatly exceeded by LTV (or even <3 mo breakeven), then it isn't a channel you use for that product!
I don't think Amazon is a good comparison here. Yes, they have a lot of data on the things that you buy but at the core, Amazon is a logistics company. Advertising is not their profit center.
Facebook and Google are both advertising companies. That is how they make money so naturally that is also where a lot of their R&D spend goes.
Imagine you're Coca-Cola. You have an on-line ad budget for this quarter. Where should you spend it? Probably the places that have the most reach: Google and Facebook. Do you know how many cans of Coke or other drink brands you will sell based on these ads this quarter? No. Is there any way for you to figure it out? No. All you know are the numbers Facebook tells you. And those numbers look pretty good. This is how "sophisticated" ad buyers work.
It's actually far easier for a tiny business to measure the conversation rate between clicks and actual purchases. Big ad buyers throw stuff at the wall to see what sticks, and they rarely bother truly measuring because that's often impossible.
The entire ad industry is based on fuzzy metrics. Within that fuzz is an awful lot of room for fraud.
Is this how you think ads work? This is horribly naive. If an ad exec at a large corporation thinks this way they should be fired immediately.
To answer directly : of course you can measure how effective your FB ads are. Pick two markets with similar demographics and a timeframe. Run FB ads in one market as an experiment and traditional media ads in another market as a control. If sales increase in the experiment market vs the control that is strong evidence that FB ads are more effective than traditional marketing. For higher confidence repeat the test with multiple markets and multiple controls.
That experiment took me 30 seconds to think of. A sophisticated marketer can probably come up with one that gives even more solid evidence.
Yeah, I also worked in ads and your experiment for a brand of any reasonable size is effectively impossible. You will have a challenge:
- Finding a time where there isn't some global event (e.g. oscars, olympics, etc)
- Weather effects (it was hot/cold so they went out/didn't go out)
- Local effects (Warriors just won the championship)
- SUPER local effects (it was prom at this high school)
- Brand history (Coke does better in ATL than Pepsi)
- etc.
And then, even if you did, ACTUALLY being able to measure impact on anything but ultra-transactional things (e.g. buying a movie ticket) is super hard/expensive too. You're going to run a survey after each experiment? And THEN, even if you DID run a survey, maybe it worked for one brand and one brand only! Maybe it doesn't even cross over to other brands in a category, let alone other categories.
Look - scientifically speaking, sure, you should be able to do what you describe. Almost none of the time does that work in the real world.
Are you saying a good ad manager shouldn’t try to do measurements though? That seems like a cop out. Why would you be making the big bucks if you can’t control for all this noise?
Ad campaigns are super expensive. If I were an exec at a company managing a campaign I’d want to see solid metrics and impact on sales, otherwise what’s all this money for?
In other words, if you can’t tell the difference between spending the money and not, don’t spend it.
Yeah, it's very very difficult for brands which are bought offline.
That's why Google are working with mastercard to get hash credit card transactions.
Additionally, Amazon are probably in a really good place to do marketing for CPG brands, as people actually purchase those on Amazon and thus they can track conversions (whereas FB and Google aren't great for these kinds of advertisers).
>Pick two markets with similar demographics and a timeframe. Run FB ads in one market as an experiment and traditional media ads in another market as a control. If sales increase in the experiment market vs the control that is strong evidence that FB ads are more effective than traditional marketing.
This sounds like an incredibly naive and expensive experiment, filled to the brim with noise.
Their video platform was based on lies and junk metrics. People still spent millions and in many cases revamped their entire organizational structure around video only to see it fail.
Even if you could prove that Facebook is filling the gaps between new users and churned users by creating fake accounts, you also have to draw a line between those fake accounts and the company's bottom line.
Fake accounts don't buy products, so if advertisers are making decisions off of CAC, numbers potentially inflated by fake accounts (reach, clicks, engagement, etc.) are secondary.
Facebook's ad power comes from their egregious data collection and lack of privacy concerns. The more they know about you, the more relevant your ads will be and the more likely someone seeing your ad will be to buy. And if advertisers can draw a line that says "if I spend $50 on Facebook ads, I'll increase my bottom-line line $100", they'll spend money until it's no longer profitable to do so.
That said, it's not like all Facebook advertisers are acting rationally in that manner. I don't know what % of Facebook revenue to coming from unsophisticated advertisers, so it's possible that they could be making a lot of money from people focusing on fuzzy metrics like reach.