Thread by @WolfieChristl

Yes, much of today's surveillance-based advertising is something between an empty promise and a giant fraud.

But the fact that we're all getting badly targeted ads does not mean that it doesn't work, because this is how it is intended to work, if it does.
ft.com/content/b013d9…

It's a misunderstanding that surveillance advertising aims to accurately target everyone. If at all, it aims to accurately target a low percentage of addressed users in order to make them act how it wants them to act.

Spamming everyone else is always part of the calculation.

Much of the data is flawed or fraudulent in today's broken web economy, and many metrics, clicks and even users are fake.

At the same time, attempts to optimize behavior in digital environments based on mass profiling and testing can be successful and lead to many side effects.

Of course, surveillance-based 'personalization' is not a remote control for people and it cannot change deep-rooted beliefs at the flick of a switch.

But it can influence how some people use digital environments, how they click, watch, subscribe, install, post, like, share, buy.

This is what marketing and advertising have always tried to do.

But I'd argue the closed loop between mass targeting and measuring how people act, combined with testing and control over digital environments and their interactive nature, has turned it into something different.

The results are often just annoying or make us think our phone is listening.

I'm afraid, surveillance marketing 'works' best to target vulnerable groups, people who easily fall for sensationalism or scams, who are angry, who are desperate because of finance/health/love/life etc.

There would be more to say. Different types of surveillance marketing and 'personalization'. Collateral damage at several levels for society and for how our economy evolves...

Anyway:

"It works / is great / is a problem" vs "it doesn't work / who cares" would be too simplified.

Also important to understand:

Marketers use discrete targeting categories less and less. If they do, they're meaningless proxies.

As Google, FB and others have now largely automated targeting, marketers mostly just let the platform 'learn' and then find them targets by itself.

This doesn't primarily rely on 'years' of platform data, but on constant access to data on real-time behaviors from outside the platforms, i.e. data on web/app activities, data uploaded by marketers etc, to close the loop between targeting, measuring behavior and optimization.

Agreed. While they train throw-away models for each campaign, operators probably also extract persistent data/knowledge.

In addition, automated targeting/optimization may be 'biased' towards the operator's business goals, intentionally and/or implicitly.

https://twitter.com/BuggeErik/status/1370701848957612032