β
Are our phones really listening to us, or have algorithms become so good that they can predict what we are likely to want? Here is what actually happens behind personalized advertising.
Most of us have experienced something like this.
You think about buying new shoes, taking a trip, changing your car or purchasing a new television. You do not remember searching for it, yet a short time later an advertisement for exactly that type of product appears on Instagram, Facebook, YouTube or another website.
The immediate reaction is often:
“There is no way this is a coincidence. My phone must be listening to me.”
The real explanation, however, is usually more interesting than secret microphone surveillance.
You think about buying new shoes, taking a trip, changing your car or purchasing a new television. You do not remember searching for it, yet a short time later an advertisement for exactly that type of product appears on Instagram, Facebook, YouTube or another website.
The immediate reaction is often:
“There is no way this is a coincidence. My phone must be listening to me.”
The real explanation, however, is usually more interesting than secret microphone surveillance.
The internet knows more about our behaviour than we realize
Every time we use the internet we generate numerous digital signals.
The websites we visit, products we view, videos we watch, searches we make, advertisements we click and applications we use can all help advertising systems estimate our interests.
The platform does not necessarily need a piece of information saying that a particular user wants a specific product.
A pattern of behaviour similar to that of other users preparing to make such a purchase may already be enough.
The websites we visit, products we view, videos we watch, searches we make, advertisements we click and applications we use can all help advertising systems estimate our interests.
The platform does not necessarily need a piece of information saying that a particular user wants a specific product.
A pattern of behaviour similar to that of other users preparing to make such a purchase may already be enough.
An algorithm does not need to read your mind
Modern advertising platforms process enormous amounts of data and look for patterns.
People preparing to buy a car, for example, may begin visiting automotive websites, watching reviews, checking insurance or researching financing weeks before they explicitly search for a particular vehicle.
An algorithm may recognize these signals before the user even consciously decides to buy.
This is one reason why advertising can sometimes feel almost predictive.
People preparing to buy a car, for example, may begin visiting automotive websites, watching reviews, checking insurance or researching financing weeks before they explicitly search for a particular vehicle.
An algorithm may recognize these signals before the user even consciously decides to buy.
This is one reason why advertising can sometimes feel almost predictive.
Retargeting is one of the main reasons
If you visit an online store and look at a product without purchasing it, the website may record that your device showed interest in the item.
An advertising platform can later display that product or a similar offer on another website or social network.
This is known as retargeting or remarketing.
It explains why a product you viewed once can appear repeatedly for several days afterwards.
An advertising platform can later display that product or a similar offer on another website or social network.
This is known as retargeting or remarketing.
It explains why a product you viewed once can appear repeatedly for several days afterwards.
What if I never searched for the product?
Advertising systems do not rely only on direct searches.
They may use broader signals such as interests, approximate location, content consumption, previous behaviour, device information and statistical similarities with other audiences.
If users with behaviour similar to yours frequently become interested in a certain product, the system may predict that the same product could also be relevant to you.
The platform does not need to know what you were thinking. It only needs to become sufficiently good at predicting what you are likely to want next.
They may use broader signals such as interests, approximate location, content consumption, previous behaviour, device information and statistical similarities with other audiences.
If users with behaviour similar to yours frequently become interested in a certain product, the system may predict that the same product could also be relevant to you.
The platform does not need to know what you were thinking. It only needs to become sufficiently good at predicting what you are likely to want next.
People around us can influence the ads we see
People in the same household may share an internet connection, visit similar locations and display related patterns of behaviour.
If someone close to you is actively researching a certain product, similar advertising may sometimes appear elsewhere within the same digital environment.
This can strengthen the impression that a private conversation must have been heard by a device.
If someone close to you is actively researching a certain product, similar advertising may sometimes appear elsewhere within the same digital environment.
This can strengthen the impression that a private conversation must have been heard by a device.
Are phones listening to conversations for advertising?
Phones and applications can access microphones when users grant permission, and microphones are necessary for calls, voice messages, video applications and voice assistants.
However, advertising platforms generally do not need continuous access to private conversations in order to create highly personalized advertising.
The amount of behavioural information already available through digital activity can often produce remarkably accurate predictions without listening to conversations.
However, advertising platforms generally do not need continuous access to private conversations in order to create highly personalized advertising.
The amount of behavioural information already available through digital activity can often produce remarkably accurate predictions without listening to conversations.
Why do these coincidences feel so powerful?
There is also a psychological component.
We see many advertisements every day that have nothing to do with us and usually forget them immediately.
When one advertisement happens to match something we recently discussed or thought about, however, we notice and remember it.
That single match can feel far more significant than hundreds of irrelevant ads we previously ignored.
We see many advertisements every day that have nothing to do with us and usually forget them immediately.
When one advertisement happens to match something we recently discussed or thought about, however, we notice and remember it.
That single match can feel far more significant than hundreds of irrelevant ads we previously ignored.
How much can digital advertising actually know?
Modern advertising increasingly focuses on showing an advertisement to people who are statistically more likely to be interested rather than displaying the same message to everyone.
This involves analytics systems, audience segmentation, remarketing, interest models and automated ad-selection algorithms.
For businesses, this can reduce wasted advertising budget.
For users, it can produce more relevant advertising, while also raising important questions about privacy and how much information is collected about online behaviour.
This involves analytics systems, audience segmentation, remarketing, interest models and automated ad-selection algorithms.
For businesses, this can reduce wasted advertising budget.
For users, it can produce more relevant advertising, while also raising important questions about privacy and how much information is collected about online behaviour.
How can users reduce tracking?
- review application permissions;
- disable microphone access for applications that do not require it;
- review privacy settings on major online platforms;
- limit advertising personalization where possible;
- clear cookies and site data you no longer need;
- periodically review which applications and services can access your information.
The internet does not read minds – but prediction is becoming very good
Perhaps that is the most interesting part of the story.
Technology does not need to know exactly what we are thinking in order to occasionally create the impression that it does.
It can analyze many small behavioural signals and compare them with patterns observed across enormous numbers of users.
The result can be an advertisement appearing at exactly the right moment.
What feels like mind reading is usually a combination of data, statistics, behavioural analysis and increasingly sophisticated prediction algorithms.
Technology does not need to know exactly what we are thinking in order to occasionally create the impression that it does.
It can analyze many small behavioural signals and compare them with patterns observed across enormous numbers of users.
The result can be an advertisement appearing at exactly the right moment.
What feels like mind reading is usually a combination of data, statistics, behavioural analysis and increasingly sophisticated prediction algorithms.