Count what you produced, not the hours

Hours are the easiest thing to measure and the least useful

An hour of study can contain almost anything. It can be forty minutes of choosing a resource and twenty of skimming it. It can be a podcast running while you cook, which is worth something but not an hour of anything. It can be a grammar chapter you read attentively and could not use the next day. All of that lands in the log as the same number, and the number goes up, and the number rising feels like progress even on the days when nothing happened.

That's what makes hours a vanity metric. Not that they're false, but that they can't distinguish between the versions of the hour, so optimising for them quietly pushes you towards the comfortable version. Passive input takes no effort to sustain, so an hour-based target fills itself with input and leaves your production exactly where it was.

Count things that had to be made

A better unit is anything you produced that didn't exist before you started: sentences written, messages sent in the language, minutes actually spoken, a paragraph composed without looking anything up. These share a property hours don't have. You can't accumulate them by being in the room.

They're also honest in a way that stings productively. A week of listening to podcasts logs as seven sessions and zero produced sentences, and seeing that zero is more informative than any amount of felt progress. It's the same reason a bank balance beats a sense of how carefully you've been spending.

You don't need many. One sentence a day is a real target on this metric, and it's a target nobody can meet passively.

What output counting protects you from

The comfortable plateau, mostly. Input is pleasant, gets easier as you improve, and produces a convincing feeling of fluency because comprehension keeps growing. Production doesn't grow that way. It stays approximately where you left it, and it announces itself the first time you have to say something out loud to a person who is waiting.

Output counting also protects you from a subtler trap: study that's really preparation. Choosing an app, reorganising a deck, reading about learning methods, planning a schedule. All of it registers as time spent on the language, and none of it produces a sentence. A metric that only moves when you make something reveals the difference immediately.

How to keep it honest

Count units, not effort. "Wrote three sentences" is a count. "Worked hard on writing" is a feeling, and feelings drift towards whatever the day happened to contain.

Count corrected output separately, if you can. Output nobody looked at is still worth more than an hour of input, but output that came back marked is worth several times as much, because that's where mistakes get identified instead of practised. Ten unchecked sentences and ten corrected ones are not the same week.

And don't let the metric turn into a streak you defend by cheating. Padding, copying yesterday's structure, or writing something you already know how to write all keep the number moving without producing anything new. The unit has to require a choice you might get wrong. That's the entire point of counting production: it's the only kind of practice that can fail, which is why it's the only kind that reliably teaches you something.