Somewhere inside every large platform there is a dashboard, and on that dashboard are a few numbers that a lot of people are paid to move upward. Those numbers are not "how happy are our users" or "how well informed is our audience." They are things like watch time and daily active users. And because the whole company organizes itself around lifting those numbers, they end up quietly deciding what lands in your feed. If you want to understand why you see what you see, start with what is being measured.

The metrics that run the show

A handful of engagement metrics do most of the work. Each is a way of turning attention, the thing the business actually sells, into a number a team can track and grow.

Watch time, or time spent. How long people stay watching or scrolling. This is the headline metric for video-heavy platforms because it maps almost directly onto attention, and attention is the inventory the ad business rents out. More watch time means more ads shown, so it is the number many systems are ultimately tuned toward.

Daily active users, or DAU. How many distinct people open the app on a given day, often reported alongside monthly active users. This measures reach and habit. A platform wants not just long sessions but people returning every single day, because a daily habit is far more valuable than an occasional visit.

Sessions per user. How many separate times a person opens the app each day. This captures the reflex, the automatic reach for the phone in every idle moment. Ten short visits can signal a deeper hook than one long one, because it means the app has become a default response to boredom.

Retention. Of the people who joined or visited, how many keep coming back over days and weeks. Retention is the long game. Growth is fragile if new users drift away, so keeping people is often prized above attracting them.

None of these is secretly sinister. They are ordinary business metrics. The catch is what they stand in for and what happens when you push hard on them.

Proxies quietly become the goal

Here is the subtle mechanism that shapes your feed. Every one of these metrics is a proxy. Watch time is not the same as "our users found this valuable"; it is a rough stand-in for it. DAU is not "people are glad they came"; it is a stand-in for that too. Companies use proxies because the real goals are hard to measure and the proxies are easy.

The trouble is that once a proxy becomes the target, effort flows toward the proxy, not the goal it was meant to represent. This is a well-known trap. Optimize hard enough for watch time and the system will learn to surface whatever holds attention longest, whether or not it is good for anyone. The measure and the mission drift apart.

When a measure becomes a target, it stops being a good measure. Push relentlessly on watch time and you get whatever maximizes watch time, which is not the same as whatever was worth watching.

This is not a claim that anyone decided to make the feed worse. It is that a system tuned to a proxy will find the content that moves the proxy, and that content is often the outrageous, the anxiety-provoking, or the endlessly snackable, because those reliably hold attention. The good and the compelling are not the same thing, and a metric cannot tell them apart.

How the number becomes your feed

The link between a dashboard and your screen is the recommendation system. In simple terms, it predicts, for each candidate item, how likely you are to do the things that move the target metric, keep watching, come back tomorrow, open the app again later, and then it ranks your feed to maximize those predictions.

So if watch time is the objective, the system is quietly answering one question millions of times a second: what should I show this person to keep them watching longest? Your feed is the visible answer. It is not a window on the world or a considered editorial choice. It is the output of an optimizer aimed at a number. The specific case of how watch time trains a recommendation engine is worth reading on its own: how watch time trains the algorithm, as is the broader machinery of a personalized feed: how the For You page works.

This is also why two people open the same app and see two different worlds. The optimizer has learned different things about what holds each of you, so it serves each of you a different feed, all in service of the same underlying numbers. The design that removes your natural stopping points, covered in why feeds are designed to never end, and the metrics that reward holding your attention are two halves of the same machine.

What this means for you

The useful takeaway is not outrage, it is orientation. Your feed is not neutral and it is not personal in the caring sense. It is tuned to move a number, and that number is a proxy for your attention, which is what the platform sells. Understanding that lets you read your own feed with clear eyes: when something grips you hard, it is fair to ask whether it is genuinely good or simply optimized to hold you.

It also points at a gentle response. Your attention is finite and valuable, the very thing this whole market is built to capture, so it is worth deciding on purpose where it goes rather than letting an optimizer decide for you. When you reclaim some of it from a feed tuned to a proxy, the honest next question is where to put it instead, and pointing that reclaimed attention at something you actually chose, like a short lesson on a topic you have been curious about with a tool such as NerdSip, spends it on your terms rather than the dashboard's. For the market this all sits inside, the pillar is here: what the attention economy actually is.