When you’re an engineer, you’re constantly being told to do things that are clearly not good for the user. If you’re building any kind of app or platform that makes its money from advertising, you are trying to maximize “time spent”—how long a user spends with your product. Time spent is considered a proxy for value delivered to the user, because if your product wasn’t useful, the user wouldn’t be using it.

Here’s how it typically works. An order comes down from on high: the board says to increase revenue. What's the best way the management team knows to increase revenue? To increase time spent. So they issue the order, which gets percolated down the tree, and now everyone is working on increasing time spent. This means making the product more addictive, more absorbing, more obtrusive. And it works: the user starts spending more time with the product.

But every worker knows this is bad. Every engineer and designer knows this is awful. They’re not happy making these features. But they can’t argue with the data. The engineer and the designer who care about the user don’t want to put these features out in the world. But the data says those features are increasing time spent—which means they’re good. Because more time spent means selling more advertising, which means making more money.