What to track when learning a language: 7 metrics that matter

Most language tracking advice tells you to track everything, which in practice means you track nothing for more than three weeks. The useful question is narrower: which numbers, when they change, should change what you do tomorrow?

These seven pass that test. The rest are trivia.

Input hours

The foundation. Time spent listening to or reading your target language at a level you mostly understand.

Track this because it is the metric with the strongest research backing as a predictor of eventual comprehension, and because it is entirely within your control. It is also the only number on this list that maps cleanly to published milestones — the FSI's 600 to 2,200 hour range depending on language difficulty.

What to do when it changes: if weekly hours drop, everything else is noise until you fix that.

Consistency

Days active per week, not total hours. Two learners doing four hours a week are not equivalent if one does it in four sessions and the other in one Sunday marathon.

Spaced exposure beats massed exposure for retention — this is one of the oldest and most replicated findings in learning research, and it applies to language input as much as to flashcards. Four forty-minute sessions across a week produce more durable gains than one long block.

What to do when it changes: a falling day count with steady hours means you are binge-compensating. Lower your daily target so the ordinary days survive.

The active versus passive split

Passive input — a podcast while you cook, a show on in the background — is not worthless. But it is not equivalent to focused attention, and if you log both as the same thing your total will drift steadily away from your actual ability.

Record them separately, even roughly. A month that reads "22 hours active, 30 hours passive" tells you something useful. A month that reads "52 hours" tells you almost nothing.

What to do when it changes: if passive is more than about two-thirds of your total and comprehension has stalled, you have found your reason.

Comprehension percentage

Every few weeks, watch ten minutes of the same show and estimate what proportion you followed. Not word by word — the gist of what is happening and why.

The same show is the important part. Comprehension varies enormously with genre, speaker, accent, and topic, so comparing a nature documentary in March against a courtroom drama in June measures the content, not you. A fixed reference point removes that variable.

What to do when it changes: rising numbers mean your content is now too easy. Move up before the input stops teaching you anything.

Content difficulty

Roughly, how hard is what you are watching relative to where you are? Children's programming, slow scripted drama, fast comedy, unscripted conversation, news — these form a rough ladder.

This one matters because hours only count when the input is comprehensible. Time spent in front of content far above your level is closer to noise exposure than to practice, and the tracked hours will look identical to productive ones.

What to do when it changes: if you have drifted upward and comprehension dropped below roughly half, step back down. Frustration is not a sign of progress.

Speaking minutes

If output is a goal, track it separately and honestly — actual minutes spent producing the language, whether in conversation, shadowing, or talking to yourself.

Most learners dramatically overestimate this. A weekly hour-long lesson in which you speak perhaps twelve minutes is twelve minutes, not sixty. Writing the real number down is uncomfortable and useful in equal measure.

What to do when it changes: if input hours climb for months while speaking minutes sit near zero, you are building comprehension that has nowhere to go. There is a fuller treatment in how to track conversation fluency over time.

Vocabulary encountered

Not vocabulary memorised — encountered. If you use a subtitle tool or reader that logs unknown words, the rate at which you meet new ones is a decent proxy for whether your content is at the right level.

A falling rate of unknown words in the same kind of material is one of the earliest measurable signs of progress, often visible weeks before you consciously feel any different.

What to do when it changes: a rate near zero means it is time for harder input.

What not to bother tracking

Streak length past about thirty days stops carrying information and starts carrying anxiety — at that point it mostly measures how afraid you are of breaking it, which is a poor reason to watch something you do not want to watch.

Words "known" from an app is another. It measures recognition in a highly cued context, which correlates weakly with understanding the same word spoken quickly in a sentence.

And self-rated fluency on a ten-point scale moves with your mood more than your ability. Rate the input, not the feeling.

Keeping it sustainable

Seven metrics sounds like a lot, and it would be if you recorded all of them daily. You should not. Hours and consistency are the two worth capturing every day, and those are exactly the two that a tool can capture without you doing anything. The rest are periodic check-ins — comprehension and difficulty every few weeks, speaking minutes weekly, vocabulary whenever your reader surfaces it.

That split is the whole trick. Automate the daily numbers so they survive your bad weeks, and reserve your actual attention for the periodic ones that need judgement.

Tracking Languages handles the daily half — it detects the audio language on Netflix, YouTube, and Prime Video and counts every minute automatically, with per-language totals, daily goals, and CSV export if you want to analyse the rest yourself. For the wider picture on what to track when learning a language, start with the complete guide to tracking language learning.