A tracker that rewards finishing the day outlasts one that scores the deficit#
The Finish Today redesign shipped on September 1, 2026, and its follow-up, the receipt that replaced the review sheet's confirm button, on September 11. Between them they change what the app asks of you in the evening. The home screen now leads with what is still open in the day rather than what has been spent; every estimate logs itself and hands you a receipt with an undo; and the day ends when the day ends, with no balance carried forward. The design brief was one sentence: reward the close, not the deficit.
That brief comes from watching the logging habit fail at a specific hour. Daily logging holds through breakfast and lunch, hits the protein target by mid-afternoon, and then loosens, because once the target is met the remaining entries change no decision and start to feel like a confession. The evidence on habit formation, and on what guilt does to eating, says that is exactly where a tracker should stop asking for penance. The wider argument that a log is an act of attention rather than an audit is the pillar on mindful eating with numbers; this post is about the evening.
Habits are built by repetition in a stable context, not by the size of the win#
The habit literature is unusually consistent on mechanism. A habit forms as a person repeats the same response in the same context while pursuing a goal; once formed, it runs as the efficient default mode of action, and people tend to infer from how often they do something that they must have intended it1. Nothing in that account depends on the outcome being large; the reward that builds the loop is the completed repetition, and the cue is the context, not a number on a screen.
The timescale matters for product design because it is long and forgiving. When people adopted a new daily behavior and rated how automatic it felt, automaticity plateaued on average around 66 days after the first performance, with wide variation between people and behaviors, and missing a single day did not seriously impair the process; gains resumed after the missed performance (Gardner et al., 2012, summarizing Lally et al., 2010). The tracking-habit guide works through those numbers in full. The design consequence is narrower: a tracker that treats one skipped evening as a broken streak is punishing something the data says is harmless, and a tracker that makes the evening entry trivially completable is protecting the only thing that builds the loop, which is the repetition itself.
The habit is the closed day, repeated. It is not the deficit, and it is not the number.
Guilt predicts disinhibition, not restraint#
The second half of the brief comes from a smaller and older body of work on what happens after a person feels they have overdone it. In the original restraint experiment, normal-weight participants were split into high- and low-restraint groups and given a preload before a taste test. Low-restraint participants ate less after a larger preload, as appetite would predict. High-restraint participants did the opposite: their intake varied directly with the size of the preload4. Having already "broken" the diet, the restrained eaters ate more, not less. That counter-regulation is the mechanism behind the what-the-hell evening, and it is the strongest argument against a surface that renders an over day in red.
Guilt itself has been tested as a motivator and failed the test. Asked whether chocolate cake made them think of "guilt" or "celebration", people who chose guilt did not report stronger intentions to eat healthily or more positive attitudes toward it. They reported less perceived control over their eating, were less successful at maintaining their weight over 18 months, and, in the subsample with a weight-loss goal, lost less over three months than the people who had chosen "celebration"5. The authors found no evidence for any adaptive or motivational property of the guilt association at all. This is an observational design, not a trial, so it shows that guilt travels with worse outcomes rather than causing them. It still leaves no reason to build guilt in on purpose.
| Evidence | What it found | What the redesign does with it |
|---|---|---|
| Wood & Rünger 2016 | Habits form from repeated responses in a stable context; outcome size is not the driver | The evening ask is one tap or none; the reward is the closed day |
| Lally 2010 via Gardner 2012 | ~66 days to automaticity on average; a missed day does not derail it | No streak that breaks; a skipped evening is just a day with fewer entries |
| Herman & Mack 1975 | Restrained eaters ate more after a larger preload | An over day is shown as a number, never a verdict |
| Kuijer & Boyce 2014 | Guilt association tracked lower control and less weight loss | No "over budget" state; a treat that fits is called a treat that fits |
What changed on the screen#
The hero leads with what is open. The top of Home is now a single "calories left" figure over a bar with three segments: solid for eaten, hatched for planned, dotted for open, with the caption reading "610 eaten · 320 planned · 290 open". Before the redesign two different "left" numbers lived on the same screen; there is now one, and it already accounts for the dinner you have booked.
Macros as grams left, not percentages spent. The nutrient breakdown became a "Left to eat" bar: 20 g protein, 45 g carbs, 12 g fat, each segment sized by its calorie share. It answers the evening question directly, which is "what would finish this well", instead of the morning question of how the day has split so far.
Plan the rest of today. When the day has calories open and nothing planned against them, a card says so in plain words: "290 kcal · 20 g protein still open — choose from your usual food now, log later with one tap." It opens a sheet that stages a few meals from history, recipes, or a typed sentence ("two eggs and toast"), with a slider on each, and books them. The booked meals count immediately and log themselves at day close. This is the same machinery Plan a Day later extended to the whole day and the whole week; on September 1 it was the evening repair, and it still is.
A receipt, not a gate. Since September 11, a meal logs the moment its estimate resolves. The review sheet that used to end in "Log it" now opens with "Logged to today · 2:36 PM · 4 items", every ingredient row in the same shape with a minus-gram-plus stepper, a cooked/raw chip where it matters, the transcript or photo that produced the estimate shown as evidence, and two buttons: Undo and Done. Offline, it reads "Saving… will sync" rather than failing quietly. The reasoning is the habit data above: the act to be repeated is the sentence, and a confirmation step between the sentence and the record was a second act with no reward attached.
Voice that stays in one place. Dictation morphs into the review in the same sheet, with the final transcript as a quote at the top, so what you said and what it became are on one screen. An "Add a photo — tightens the range" row is there for when a picture would narrow the estimate.
A day you can recognize. Meal cards carry a photo or a generated plate tagged "Auto", and the source is written out: "Voice · 9:12", "Scale · Kitchen · 182 g". The picture-first idea grew into the meal grid release, which carries the memory science behind it.
The dock. Barcode, camera, and microphone glyphs; a focused composer gains +50, +100, +200 kcal and a Custom chip for the times a rough addition is all the evening deserves.
Closing the day, and why there is no weekly debt#
The app has been daily-only since the summer, when its weekly-budget interface was retired. That decision is the quiet half of this release. A weekly budget turns a large Tuesday into a debt that Wednesday through Sunday must repay, and the Herman and Mack result says that framing is precisely the one that produces counter-regulation in the people most likely to be tracking. A daily surface has no ledger to carry: at midnight the day closes, planned meals that were not touched become the record, tagged "as planned", and the next morning opens on a fresh number.
The close is the beat the redesign is named for, and it is deliberately unspectacular. There is no verdict, no streak counter, no colored grade. The reward for a closed day is that it is closed: the plan logged itself, the deviations were a sentence each, and the receipt for each one can still be undone tomorrow if it was wrong. A day that ran over is shown as a number over, in the same type as a day that ran under. Whether that is enough reward to carry a habit through the evening is an empirical question. The evidence that guilt makes things worse is stronger than the evidence that any particular reward makes them better, and the redesign is built on that asymmetry.
Streaks were deliberately left out: a counter that breaks on a skipped evening punishes exactly the thing the habit data says is harmless.
Where this could be wrong#
The guilt evidence is observational, and it is possible that people who feel guilty about cake differ in ways the design cannot fix by removing red from a screen. The habit literature is mostly about simple daily behaviors, and "log the evening" is a behavior with a variable context, which the same literature says forms more slowly. And the counter-regulation experiments are from the 1970s, in normal-weight students under laboratory preloads; the effect has been replicated many times since, but the size of it in a person's own kitchen with a phone in hand is not something those studies measured.
What the release commits to is narrower than "this will build the habit". It commits to not doing the thing the evidence says breaks it. For the reader whose plateau is not a logging problem at all, the physiology is covered in why weight-loss plateaus happen, and the case for tracking without counting is in losing weight without counting calories.
FAQ#
Why does feeling guilty about food make dieting harder?#
Because guilt tracks with less perceived control over eating rather than more. People who associated chocolate cake with guilt reported lower perceived behavioral control, were less successful at maintaining weight over 18 months, and lost less over three months when trying5. The study found no motivational upside to guilt at all.
What is the what-the-hell effect in eating?#
The pattern where a restrained eater who believes they have already broken their diet goes on to eat more, not less. In the original experiment, high-restraint participants' intake rose with the size of a preload while low-restraint participants' intake fell4. A tracker that frames an over day as failure supplies the trigger.
Does missing one day of logging break the habit?#
Not according to the habit-formation data. Automaticity plateaued around 66 days on average, and missing a single performance did not seriously impair the process; gains resumed after the missed day2. A skipped evening is a day with fewer entries, not a reset.
What does the auto-log receipt do if the estimate is wrong?#
The meal is already logged, so the receipt is where you correct it: step a gram count, flip a row to cooked or raw, tap the transcript or photo to re-estimate, or Undo to remove the meal entirely. Later edits patch the same logged meal; nothing is duplicated.
Sources#
- Wood W, Rünger D. Psychology of Habit. Annu Rev Psychol. 2016;67:289-314.
- Gardner B, Lally P, Wardle J. Making health habitual: the psychology of 'habit-formation' and general practice. Br J Gen Pract. 2012;62(605):664-666.
- Lally P, van Jaarsveld CHM, Potts HWW, Wardle J. How are habits formed: Modelling habit formation in the real world. Eur J Soc Psychol. 2010;40(6):998-1009.
- Herman CP, Mack D. Restrained and unrestrained eating. J Pers. 1975;43(4):647-660.
- Kuijer RG, Boyce JA. Chocolate cake. Guilt or celebration? Associations with healthy eating attitudes, perceived behavioural control, intentions and weight-loss. Appetite. 2014;74:48-54.
Source: BurnWeek — "Close the day, not the deficit: the Finish Today redesign", https://burnweek.fit/blog/finish-today-redesign/. Licensed CC BY 4.0: free to quote or reuse with a link to this page.
