How Much Are Google and Facebook Overstating the Effectiveness of Their AI and Ads?

Executive Summary

• Facebook and Google’s business model is based upon both ads and exaggerating to advertisers about its data’s reach and effectiveness.
• We cover how Facebook and Google get away with it.

Introduction

In the article Facebook’s History of Lying to Advertizers and Ad Buyers About Ad Reach, we cover Facebook’s long-term practice of providing false statistics to advertisers and ad buyers. In each case, when cause providing false information, Facebook blames an error or a bug. In this article, we will cover some actual ads that follow us around on many of the websites we visit and provide examples that question the effectiveness of these ads and the supposed “advanced AI” behind the scenes that supposedly target web visitors so effectively.

Our References for This Article

If you want to see our references for this article and other related Brightwork articles, see this link.

The Claims About Targeted Ads

The first claim is..

Google on Tuesday announced several new machine learning-powered ad offerings to help brands run more effective campaigns. The tools are designed to enhance ad campaigns across the tech giant’s Search, YouTube, and Google Maps properties, and will likely help it top the $21.4 billion in ad revenue it generated in Q1 2018.

Although the field of AI tech is still somewhat nascent, more marketers are beginning to realize how the technology can help with completing tasks more quickly, and this could contribute to uptake in Google’s new machine learning ad tools. – Business Insider

What is the problem with just accepting this claim?

Well, first, Google is targeting this message at both investors and advertisers. Google wants investors to push up its stock price. Second, Google wants advertisers to think that the ads they buy will be even more effective than ever.

This quote is from Business Insider, which, like most business and IT media, does not bother to evaluate companies’ claims and instead serves to repeat them. Business Insider also probably counts Google as an advertiser, and you don’t fact check advertisers, or you soon lose those advertisers.

We can see the claims regarding ad targeting by Big Tech in the following quote.

The search firm has officially launched Auto Ads, a system that uses machine learning to not only determine the types of ads you see, but how they’re placed. The AI technology will decide how many ads are appropriate for a page and where to put them. Advertisers have to give up control, but Google has bet that they won’t mind the results. A long beta test saw publishers rake in an average of 10 percent more revenue.

There is a concern that trusting AI could create problems. Beta testers complained about ads crowing their pages, and there’s always the concern that it’ll serve fake ads or other dodgy promotions. – Engadget

First, yes. Websites are crowded with ads, and it is degrading the user experience.

But second, all of this advanced AI is only providing an extra 10% of revenue? But I thought that AI was so powerful. Where are the breakthrough forecast accuracy gains? In many cases, if I were to find a way to provide only a 10% increase in forecast accuracy for a client using some different technique, the client would in many not go forward with the improvement, stating that the cost of implementing the change outweighed the benefit of the improved forecast accuracy.

I could include more quotes from other sources, but every one of them I found was basically the same — which is some person who does not know anything about AI, repeating a claim made by Google or Facebook. For years now, AI has been a magic bullet that you use to make promises that can’t be verified.

The Effectiveness of Online Advertisements

I do not have a Facebook account and do not click on Facebook Like buttons. However, I am an avid user of Google services. Google has a similar business model to Facebook in that they surveil their users and sell this information to advertisers.

Approach #1: Facebook

Facebook lures in users with their social media website, and through users clicking Like buttons on media. These actions allow users (even non-Facebook users in the media Like button) to be tracked across websites.

Approach #2: Google

Google’s approach is to either serve ads based upon Google searches — so within the search results or to share information about the user to websites where ads are placed.

Both companies drive revenue the same way, but each obtains its data somewhat differently.

The Presentation of Facebook and Google’s Ad Forecasting Prowess Versus Observing the Ads

When reading about how much money both of these companies make from ads, one could easily fall into the assumption that both companies are experts at serving ads to Internet users. But something happened when I was reading the accounts of Facebook and Google’s ad prowess. It appeared much less impressive when the recipient of these ads, versus reading about their supposed ad prowess.

Example #1: Google’s “Targeted Ads” Website Searches

The first modality we will observe is how Google got its start in serving advertisements, the good old Google search. 

As Google has become more and more anti-user, it now serves more and more ads. Notice that the first page is now all ads. 

It’s tough to see how any of these results require advanced AI forecasting that Google is supposedly masters at doing. I told Google what I was looking for, and Google is serving the highest paying advertisers. Naturally, as Salesforce is the largest CRM vendor, they can afford to pay the most, and they are in the top results.

The second page is supposed to be where the ads stop, but the most prominent result is also an ad. This is called a “quick snippet” by Google and is defined as follows.

How Google’s featured snippets work
For some searches, Google provides a quick answer or summary with a content snippet from a relevant website. These featured snippets are most likely to show up when your search is in the form of a question.

It is designed to look educational, but it is published by Salesforce, which means its entire purpose is to drive people to purchase Salesforce CRM.

This is a problem because now we have a financially conflicted entity that actually has an inferior quality CRM system, which is listed as the “horse’s mouth” on CRM systems. How about my research that indicates that CRM systems most likely have a negative ROI, as I cover in the article The Problem with CRM Systems that Salesforce Does Not Want You to Know? 

Trust that you won’t be reading that article on the Salesforce website.

And apparently, Google sees no problem with this. A company whose primary motivation to write about CRM is to sell CRM, and Google sees this as very normal and good. I am aware that Salesforce may have obtained this position through crafting the article to authentically be in this position, but Google should have pushed this result down through intervention, as it is not an unbiased source.

After this, we get into the authentic results. However, I skipped all of the ads, and always skip all of the ads as I am not interested in what people or companies have to say that have paid Google for their position.

But there is no magic to these ads. The person tells Google what they want to look at, and Google tries to divert them from looking at the authentic results as much as possible to view their ads, which is how Google makes its money.

Example #2: Google’s Targeted Ads From Cross Website Visits

The second modality of how Google serves ads is through cross-website visits. This is where the user is tracked across the Internet for all of their visits. As you will see, I am constantly served the same ads throughout my website visits.

The following are ads I was served from Google while visiting various sites. I had noticed these types of ads for some time and considered them a nuisance. And it was not until I began researching the topic of Facebook and Google ads and their revenue models that I began thinking about the ads I have been served.

Upon reflecting on my own experience with Google ads, I found it incongruous with these experiences that these ads were considered effective by advertisers and by those that wrote articles about online advertising.

As with many people, I have been keeping up with the coronavirus statistics using the worldmeter.com website. Ads that display on this website are a form of cross-site tracking. This is not really the Worldmeter website, but instead can be thought of as a Google carve out of the site. Google controls this area of the website, what is shown, how quickly a new ad is shown. Google then pays this website to show this ad. When Google shows ads following the first modality, it does not need to pay anyone, as in that case you are already on the Google website. 

Worldometer also places a cookie in my browser, which lies in improving my “experience.” I recently deleted all of the cookies in my browser, which numbered around 2500. Within just two days of using the computer, I was up to 282 cookies. A week later I was up to 1132 cookies.

All of the Surprising Things in My Brower’s Cookie Bin

The Safari browser is honest about what cookies are. As it says..

“They can be used to track your browsing.”

Chrome, made by Google, does not have this disclaimer when you go to clear the cookies in Chrome.

I thought I would provide that the Brightwork Research & Analysis website uses no cookies, proving it is unnecessary to have a cookie for a good website experience. However, when I cleared all the Safari cookies and loaded the Brightwork website, I found some interesting things.

First, the Brightwork website does install a cookie on my browser. This cookie is not for users, but to keep authorized users to stay logged in. 

Secondly, the cache files that you see are not cookies. The cache speeds up the site when visited again. But it means there probably was a cookie from that entity before I cleared the cookies.

Third, I found I had two Facebook cookies, even though I never go to the Facebook website and don’t have a Facebook account. This means some other website loaded these Facebook cookies on behalf of Facebook into my browser. It also means that Facebook tracks my Internet activity, and I was unaware of this.

Fourth, Microsoft’s Bing.com was also tracking me with a cookie. Again, I don’t use Bing. In addition to all of the other websites tracking me, I am being tracked by Facebook, Google, and Microsoft, while of these companies, I was thinking Google was only tracking me.

Time to Change the Name of the Privacy Tab?

Something amusing is that the “cookie bin” is called “Privacy.” As the tech giants are surveilling me and apparently everyone else and their uncle, shouldn’t that tab be called “undesired surveillance?” I think we are a bit past using the term “privacy” for what is happening to my browser. And if you try to surf the Internet with an ad-blocker, first, Facebook has technology that allows them to bypass ad blockers, and second, many sites begin screaming bloody murder and generate several pop up messages telling you to turn off ad blocking once you don’t accept their constant under the radar surveillance.

A major lie told by not only Google and Facebook, but uncountable websites that tracking you is necessary for the experience. The experience you are experiencing right now is provided sans cookie.

Notice when I try to turn off the ad, I am told Google serves this ad. Google is paying the site to run this ad for the advertiser called the K21 Academy. If I click the ad, then K21 Academy is charged for the click. 

Google does not give me the option of not seeing ads, only that I don’t want to see this particular ad. This is only allowing me to give more information about myself for future ads. 

These ads are based upon a forecast of my previous search history. Facebook and Google present this to advertisers as “targeted ads.”

These ads are inaccurate.

I am not interested in Oracle Integration Cloud Services training (although I did look up this topic in a search) and am not interested in any of the batteries displayed. If anything, these ads appear brainless — they merely represent previous searches that I have made in Google. In many cases, I will be shown ads for products for weeks that I have already decided I was not going to buy, even though I was researching them at the time.

Therefore, with all of the Big Data used by Google, they still only have a weak ability to forecast what I am interested in at the time. I also can’t ever recall clicking an ad. Maybe it happened at some point in my life, but it is not something I generally do. I like to search for my interests actively and don’t respond to websites putting things in front of me to click.

Like other web users, I am repeatedly shown these ads, and as I never click them, they would appear to have zero value to advertisers. Other people may be more susceptible to clicking ads. However, it is also known that a user has to be shown many ads to result in a click. For any website to make money from Google ads, it needs to have a very large volume. Therefore, many websites are significantly degrading the user experience, as well as the credibility of their website, for a small payout.

However, look at what is said about this targeted advertising capability.

Google created the first insanely lucrative markets to trade in human futures, what we now know as online targeted advertising, based on their predictions of which ads users would click. Between 2000, when the new economic logic was just emerging, and 2004, when the company went public, revenues increased by 3,590 percent. This startling number represents the “surveillance dividend. It quickly reset the bar for investors, eventually driving start-ups, apps developers and established companies to shift their business models toward surveillance capitalism.

Even Ford, the birthplace of the 20th-century mass production economy, is on the trail of the surveillance dividend, proposing to meet the challenge of slumping car sales by reimagining Ford vehicles as a “transportation operating system.” As one analyst put it, Ford “could make a fortune monetizing data. They won’t need engineers, factories or dealers to do it. It’s almost pure profit.” – New York Times

A big part of what Google and Facebook do is make advertisers think that their ads are far more effective than they really are.

This is another entirely unrelated site from Worldometer. Instead, it is a research site. However, it shows many of the same ads as on the Worldometer site. (I had also recently been looking at tents).

This means that Google is also controlling the ads on this website. I am being shown the same Google ads on multiple websites. However, how many times will I be shown the same small number of ads? As with the batteries, I have already moved on from the idea of buying a tent a month ago, but the same ads persist. And..

.they follow me on many sites that I visit. It seems I am continually shown the same ads. My previous tent searches on Google drive the Patagonia ad. The Priceline ad may have been driven by my Google searches for travel destinations. However, I am not a budget traveler and so not a good candidate for becoming a customer of Priceline. 

Over a month after searching for tents in Google, and long after my interest in buying a tent has passed, I am continually showed images of tents, to the point where I am sick of seeing the same ads for the same tents.

This is the supposed fantastic “targeting” offered by Google’s AI.

This ad to the right was served to me on the Engadget website while researching this article. It’s for Verizon. Is this ad-based upon any AI? It’s hard to say. However, it’s an ad for Verizon. I don’t see how I am a particularly good potential customer for Verizon as I use Google Fi and have for many years. Maybe it is targeted, or maybe it is just random.

I have also received ads for things like Huggies diapers. I am 51 years old and have never had a child. I am probably not a very good candidate statistically to start now. And if I were about to start a family, I think Google would know about it from my search history. So it is also difficult to see how some intelligence determines I am a potential customer for Huggies.

Let’s Visit a Website That is All About Serving Ads

I knew from experience that one of the best places to see extremely irritating and distracting ads was the Forbes website. Forbes also has its ads run by Google. However, if Google knows me so well, why was I shown an ad for ServiceNow? Am I really a potential customer for service desk software purchased by large companies? 

Forbes has a program called “Brand Voice,” where companies like AWS above publish advertisements designed to look like articles. And then, within these article/advertisements, Google serves ads. (Therefore, the entire page is ads). However, did it take advanced Google AI to serve me an ad for AWS, as I selected an article/advertisement page on AWS? 

Where is this advanced AI that is targeting me with things I want to see?

I am being served ads for things I searched for over a month ago, ads for things I would never be interested in like Huggies, generic ads for Verizon, and ads for the same thing that the article or search result is about. No AI is necessary for any of these results.

Considering the History of Forecast Accuracy Exaggeration

If we take Google and Facebook out of the discussion for a moment, forecasters have a long history of exaggerating their forecast accuracy. When a forecaster has poor accuracy, one of the first things they do is try to hide this by changing the forecast error measurement. My experience consulting in companies for several decades is that most do not know their forecast error and have a poor understanding of forecasting error measurement. Many groups eliminate measurement once the measurement undermines their claims that they are effective at forecasting. Forecast accuracy has a very strong hierarchical element to it. That is people high in the hierarchy will hold those lower in the hierarchy accountable for forecast accuracy, but the higher your status in a company, the less your forecast accuracy matters.

This applies to outside of companies as well. Firms like Gartner, Forrester, Deloitte, and McKinsey specialize in making promotional, financially biased, but ultimately inaccurate forecasts. However, they are not held accountable for forecast inaccuracy, because once you have a high enough status, your forecast accuracy is immaterial to the continuation of your status. Humans have a general pattern of preferring to not do their own thinking, but to “save energy” by simply agreeing with whatever considered the status quo and this is whatever the powerful and prestigious entities say is true, not what is in fact true.

Google and Facebook are now two of the most prestigious companies globally, and they, too at a point where their forecasts and what they say about their internal capabilities are rarely questioned.

Selling Exaggerated Claims of Advertising Accuracy and Effectiveness

In this case, we have two giant companies that sell a product: the ability to target, forecast, and what ads will be effective to show to which users. We also have advertisers and ad buyers, which, while they can measure clicks generated from both Google and Facebook ads, don’t have a very good way of tracking these ads’ revenues. And advertisers are not really paying for clicks. They are paying for revenues.

This video explains that Facebook ads are becoming better at creating “awareness” while Google is better at creating sales. One of the issues with Facebook ads is that the Facebook website has a contradictory purpose of use versus its purpose of the provider Facebook. Facebook users go to Facebook to keep up with friends and family, while Facebook’s only reason for maintaining the Facebook website is to serve Facebook users’ ads. This is one reason why a website like Amazon has such a high revenue per user. The Amazon website has no contradictory purpose. People go to Amazon, not for social interaction, but to buy things, and Amazon maintains its website to sell things to people.

Analysis of the ads shown in this article calls into question this claim that Google has some advanced AI targeting users. And the ads for Facebook are generally known as significantly less effective at generating revenues than those from Google.

Marketing Scam Companies

Something I realized by working with software vendors, is that there are plenty of scammy marketing companies that make a lot of promises about their ability to drive leads to your door. The joke is, you go to their office and everyone is wearing a turtleneck and the offices are very chic. These companies use a lot of marketing terminology and get the prospects thinking that the company must know what it is talking about. However, mostly what they deliver is leads that sales can’t convert. Naturally, these are the types of companies that are going to talk up Facebook and Google ads as it makes them seem cutting edge.

This is a quote that I have made anonymous because it’s better if the provider not be known to their employer.

Facebook charges 96c US for every click on an advert. For our company, requires on average of between 500 clicks to obtain a sale. So the actual cost to us around ~$450 per sale.

Call centres in India are only charging ~$200 per sale.

However, remember the assumption. Facebook’s AI is so sophisticated and its surveillance-based user data predictive.

Therefore why can’t it beat a call center that does not have these advantages and is calling from purchased lists?

Conclusion

Both Google and Facebook are engaging in extreme forms of surveillance. They are perpetually caught lying about what they are doing with user data and how many other sources they accumulate. Both Google and Facebook have created enormous revenue streams from advertising. However, what is not clear is that any great intelligence is being used to leverage this massive quantity of data or that their advertisements are anywhere near as effective as they claim. For the longest time I have been told that many other firms are lying about their AI capabilities, but with Google and Facebook..

“It’s legit,”

..and that only companies with such enormous quantities of data and financial resources can..

“Really do AI right.”

Do the examples I have shown in this article look like AI done right?

Billions from Glorified Bus Advertisements

What is also somewhat amusing is that with all the hubbub about Google and Facebook’s great companies, they make their money almost entirely by doing the same thing that billboard companies do, or ads on busses do.

Actually, it’s worse, as no ad on a bus has ever degraded my experience riding a bus. No ad on a bus stop bench ever violated my privacy or surveilled me or got in the way of me sitting on the bench. Yet Facebook claims its entire objective is to “Make the World More Connected.”

That is a lot of sexy talk, for such a banal final outcome.