Why Facebook is not overhearing you

I’ve seen numerous posts recently that Facebook listens to their conversations in order to target ads even better. At first, I’ve spent some time researching geofencing advertising technologies, spent some time in the machine learning space and I use Facebook ads as well.

The answer is very simple: Facebook is not overhearing you through your phone. Therefore, let’s try to break things down. No deep researches, only general information.


It is almost impossible to extract words and clear sentences from the phone when it is in your pants pocket and offline, of course. Those who believe that Facebook had created to such a “listener” clearly did not try to come to an agreement with Siri, Alexa or Alice. By the way, all those voice assistants only work online. If they “listen” to you in real-time and process your speech in the cloud, Facebook will go broke at least on traffic, I’m not even talking about all the hardware that would have been required, and don’t forget about the electricity. It would make more sense to mine cryptocurrency with all that hardware (even with the current market state).


Advertising technologies on low-frequency queries work only if they are extremely cheap. The advertiser will not pay more than the relative figure of 5–10 dollars per thousand ad impressions. Facebook, with all its efficiency, most likely keeps the cost of a display at a level no higher than two or two and a half cents, optimizing to the utmost, and selling this ad for pay-per-action.

“Listening” to your life will obviously cost more than the 10 cents a day that you can generate for Facebook as much as possible by flipping through the tape. And then, 10 cents, this is if you enter the audience of high-income US citizens.


Let’s visualize. The Facebook corporation, which employs more than 30,000 people (this is a completely clogged Olympic size stadium), using Apple phones (130,000 employees) and Google (85,000 employees), which 150 additional corporations generally are involved in the development process, listens secretly to the conversations of abstract Alice and Bob in the bar somewhere in London. Complementing the visualization, the action takes place in Silicon Valley, California, where every day any things that someone somewhere was trying to keep silent pop up. And no one, no one whispered it to the journalists who are awake at night, they are still looking to pinch Zuckerberg’s nose.


Your digital footprint and without “wiretapping” is phenomenal. You scroll through Facebook and Instagram tapes, like, share, send messages, stay on some posts for a longer time, click on notifications at once. You google the information, you read the news and write emails, you go to the sites, you make purchases online and offline (yes, Visa and MasterCard also sell your data from the bottom of your heart, along with the owners of the loyalty card programs).

Then a few clever data scientists sat down and made predictors that understand when someone should show what kind of advertising in order to maximize the chance of selling this product.

Here’s an example: Alice, a highly paid manager, bought a car three and a half years ago. Recently, Alice paid several bills at the service station but did not extend the insurance. Two friends who were with her in the same geolocation, at this moment googled “BMW X5” and “Audi Q7”. Alice comes home and sees an advertisement for the “Mercedes GLS”, about which she spoke with friends in a bar. Hold on a second, Facebook has been eavesdropping on her! Well, unfortunately, the answer is no. The add suggestion is simply based on a complex scoring model.