Habit analytics: trends that are not vanity

A chart can tell the truth or it can sell you a feeling. Vanity analytics in a habit app look like a perfect week, a longest streak, a heat map that is all fire, and a ranking that says you are crushing it while the actual behaviour is a two-minute stub you no longer notice. Useful analytics answer three questions: did the cue still fire, how often across a month, and which rows are quietly dying.

Harkin and colleagues (2016) showed that progress monitoring helps when it is recorded and, in many cases, when it is reviewed. Carver and Scheier's feedback-loop account of self-regulation needs a comparison between current state and a reference. A trophy does not make that comparison. A 28-day completion rate does. So does "you skip this one every Thursday."

Phillippa Lally's 2010 study is the other half. Automaticity grew with repetition, not with a highlight reel, and a missed day did not, on average, erase the learning. If your charts treat a miss as a cliff, they are fighting the evidence. Streaks versus consistency percentage is the metric argument. This post is about which pictures to trust once you have numbers.

Windows beat days, rates beat streaks

One day is noise. One week is a mood. Twenty-eight days is long enough to see a weekly pattern (always skip Sunday, always hit Tuesday) and short enough that you can still change the cue. Wood's context work predicts those weekly patterns: Sunday is a different room, a different clock, a different set of people. If the chart cannot slice by day of week, you will moralise a context problem.

Heat maps are useful when empty cells are allowed to stay empty. They lie when the app paints a miss as a moral event, or when you only look at the current month because the last three were ugly. Calendar chains and heat maps that do not lie is the visual version. Here: zoom out until a single red day is a speck. If you cannot bear to zoom out, the chart is running your mood, not your practice.

Rankings help when they are cruel in a specific way: this habit is at 40 percent, that one is at 85. They fail when they become a game of deleting the 40 percent row to raise the average. Deleting the hard habit is not improvement. It is all-or-nothing thinking with a pie chart.

Charles Duhigg popularised the loop so you could inspect a cue. Analytics should inspect the loop, not hide it. "Most missed" is a cue report. "Best day of week" is a context report. "Longest streak" is a story. Keep the first two on the home screen of Charts. Put the third on a shelf.

What not to optimise

Do not optimise for a 100 percent week. Fogg's tiny-habits work would rather you keep the behaviour alive at a small dose than burn it for a perfect grid. Do not optimise for identity labels the app invented ("legend," "diamond"). Those are gamified streaks, which can help as milestones and can also become the goal. Do not optimise sleep or mood scores as if they were KPIs. How to track mood without grading your day is the warning. A 1-to-5 is a snapshot. An average of snapshots is not a self.

Quantity charts are useful when the metric was chosen on purpose. A water graph that climbs because you doubled the target to impress yourself is choosing one metric gone wrong. Look at variance, not only the mean. A meditation timer that is 20, 20, 4, 20, 0 is a different practice from a flat 12. The zero is data. The 4 might be the version that survives travel.

Gollwitzer's if-then plans can be rewritten from a chart. If Thursday is always empty, the sentence is not "try harder on Thursday." It is "after the Thursday stand-up, I do the two-minute version before lunch." Analytics that do not change a sentence are entertainment.

Chart	Trust it for	Ignore it as
28-day completion rate	Whether the cue is still alive	A grade on your character
Day-of-week breakdown	Context problems	Proof that you are "inconsistent"
Heat map over months	Long shape, travel holes, illness	A stained-glass identity
Longest streak	Almost nothing operational	The thing you protect at all costs
Most missed / ranking	Which sentence to rewrite	A reason to delete the hard row

Export sits next to charts. If the only place the truth lives is a locked screen inside a subscription, you do not own the monitoring Harkin is talking about. Exporting your habit history is the portability rule. A CSV you can open in a spreadsheet will outlast the app's colour palette.

Habit AI is a useful example of charts aimed at the operational questions: active days, success rates, 28-day windows, heat maps, and rankings of which habits stick versus which get skipped, plus an export so the record is not trapped. Look at those, not at a trophy case. The product is optional. The questions are not.

Once a month, pick the worst row and change only the cue or the size. Do not rebuild the dashboard. Analytics that produce one sentence of change are doing the job. Analytics that produce a new identity are vanity with axes.

Read next

Streaks versus consistency percentage is the metric underneath the pictures. Calendar chains and heat maps that do not lie is the visual cousin. For trends you can actually use, see how Habit AI works.