Choosing one metric per habit

Goodhart's law, in the usual paraphrase, says that when a measure becomes a target, it stops being a good measure. Habit apps reach that failure in a week. You attach pages and minutes and a mood and a note to "reading," and then you either fake the stack or you skip the whole row because the form is a second job. One behaviour, one metric. If you need a second number, you have a second habit, or you have a project.

Harkin's 2016 meta-analysis of progress monitoring assumes there is a progress to monitor. Two progress bars on one cue create two feedback loops and a tie. Carver and Scheier's control theory needs a reference value. Pick one. Phillippa Lally's 2010 participants defined a single daily act in a stable context. They did not define a dashboard.

The metric should move when the behaviour moves

If the behaviour is "open the book after coffee," the first metric is yes/no. When that is boringly true, switch to pages or a timer, not both. Yes/no, quantity, timer, or a 1-to-5 scale is the menu. This post is the constraint: one item from the menu.

Wendy Wood's context-cued automaticity is specific. The association is between a setting and an action, not between a setting and a spreadsheet. "After coffee I read" can become automatic. "After coffee I read twenty pages in twenty minutes at a 4/5 focus with a gratitude line" will never become automatic, because it is a ceremony. Fogg's tiny-habits work is the same constraint from the other side: ability has to stay high. Extra fields lower ability.

Charles Duhigg popularised cue, routine, reward. Extra metrics usually sneak in on the reward side ("I want the chart to look serious"). That is vanity, which habit analytics already told you to ignore. The reward for a reading habit is that the book is open, plus a log that took one second.

Split the row instead of stuffing it

Walking can be steps (Health) or "ten minutes after lunch" (timer). Those are different projects. Apple Health and habits that log themselves can own the step floor. The cued walk can own the timer. Sleep duration can be a Health measurement. Wind-down can be a yes/no. Weight can be a weekly quantity. The daily eating behaviour is a different row, and how to log weight without spiraling is the caution.

Gollwitzer's implementation intention should name one then: "After I shut the laptop, I start the ten-minute timer." If the then contains "and," you are stacking. Stacking is allowed later, on purpose, as habit stacking that does not collapse. It is not allowed as a single metric with two heads.

A 1-to-5 next to a quantity is the usual trap for exercise: minutes plus "how hard was it." If intensity is the actual goal, log intensity and let duration be messy. If showing up is the goal, log yes/no or minutes and let intensity be messy. You cannot steer both with one glance. Athletes with a coach can. You, at 7 a.m., cannot.

Habit you meant	One metric	The extra metric to exile
Read after coffee	Pages, or a timer, not both	Mood, "deep work quality," word count
Walk	Minutes of the cued walk, or steps	Both, plus a heart-rate story
Meditate	Timer	A calmness scale every sit
Shutdown	Yes/no at the same clock	Productivity rating of the day
Water	Quantity	A hydration "score" the app invented

How to write a habit in one sentence will expose a stuffed metric: if you cannot say the habit without "and," the metric is already two. Rewrite until the sentence has one verb.

Locke and Latham's goal-setting work still applies to finite projects (finish the draft, run the race). Those goals can have several KPIs on a spreadsheet you open weekly. Daily habits cannot. Goals versus systems versus habits is the split. Do not import project management into a row you hope will become automatic.

If you already have a bloated habit, do not delete the history. Export, then create a new row with one field, keep the old row archived, and run thirty days. Comparing two metrics on the live row is how people never start the cleaner one.

A weekly review is allowed to look at a second number without putting it on the daily row. Steps can live in Health. The cued walk can live in the tracker. On Sunday you may glance at both. That glance is a system, not a habit. Goals versus systems versus habits is the sort. Daily automaticity cannot carry a committee of KPIs. Sunday can.

Habit AI is a useful example of the constraint in product form: a habit has a type (yes/no, quantity, timer, or scale) and a target, not a bundle of parallel scores. Pick the type that matches the sentence. The science is the constraint. The app is just a place to store it.

One metric, one month, then you are allowed to change the instrument if the behaviour outgrew it. You are not allowed to add a second instrument because a chart looked empty. Empty charts are how you notice that the cue failed. That is the point of monitoring, not a reason to decorate the row.

Read next

Yes/no, quantity, timer, or a 1-to-5 scale is the menu this rule picks from. How to write a habit in one sentence is how you know you only have one verb. For a tracker that stores one type per habit, see how Habit AI works.