The 168 Hours Time Audit and Calibration Training

What each one is actually for

168 Hours Time Audit

The 168 Hours Time Audit works by making a week countable. You write down what you're doing in every half-hour or hour block of a full week, in real time or as close to it as you can manage, rather than reconstructing the week from memory afterward. The number 168 is simply 24 times 7, made explicit so a week stops feeling like an unlimited, fuzzy resource and becomes a fixed, countable total you can actually account for, the same way a budget makes a paycheck countable instead of just 'money that comes and goes.' The point isn't the log itself — it's the comparison you run once the week is over, between what the log actually shows and what you would have guessed it showed before you started. Most people are confident about that guess right up until they see the real numbers, and the gap between the confident guess and the recorded fact is usually bigger, in a specific direction, than they expected. Picture writing 'email' in six separate half-hour blocks across a Tuesday you would have sworn, beforehand, had maybe one email-heavy hour in it.

Superforecasting

You would have said twelve hours, maybe fourteen. The log says twenty-six, and that gap is not a moral failure — it is a measurement of how far your sense of your own week sits from the week you actually had. Calibration training takes that same gap and extends it to anything you believe about what is coming. Instead of saying a project will 'probably' land on time, you commit to a figure: sixty-five percent it ships by the thirtieth. The number goes down before the outcome is known, which is the entire point, because a hedge word can be quietly reinterpreted afterwards into having been right and a number cannot. Then you keep score across dozens of these, spread over unrelated questions, and read the pattern the scoring gives you: of everything you called seventy percent, did roughly seven in ten happen? If many more did, you are underselling what you already know. If far fewer did, your confidence is running ahead of your accuracy, and that has just become a specific, correctable fact rather than a vague suspicion. The research behind it — Philip Tetlock's work with people forecasting real geopolitical and economic questions — found that those who kept score and adjusted toward what the score showed became measurably more accurate over time, not by knowing more, but by learning what their own confidence was worth.

What you get out of pairing them

Most personal-productivity advice quietly assumes you already have an accurate account of where your week actually goes, rather than the flattering or self-critical estimate sitting in your head — the time audit is what actually produces that account. You cannot fix a time allocation you've never measured, and the measurement itself is often the useful part: before deciding to spend less time on something, you need to know, correctly, how much time it's even taking. Calibration training does the structurally identical thing for judgment instead of hours: it produces an accurate sense of how much your own confidence is actually worth. Most people's gut sense of their own certainty doesn't correlate well with how often they're right, and without a tracked record there's no way to find that out — you just keep feeling confident, correctly or not, indefinitely. Both practices trade a comfortable but unverified inner sense for an outer one that can actually be checked, and checked is the operative difference.

What 168 Hours Time Audit rests on

• Externalizing makes pattern visible randomised trials
  Putting an internal pattern into an external, structured form makes it visible and available for examination instead of remaining implicit and merely enacted.
• Measurement precedes optimization plausible mechanism
  Accurately measuring a behavior before trying to change it, rather than estimating it, is a precondition for effectively reallocating it toward stated priorities.
• Self estimation of time use is unreliable observational studies
  A person's subjective sense of how they spend time diverges substantially from a detailed log, so self-report alone is an unreliable basis for time-management decisions.

What Superforecasting rests on

• Decomposition improves estimate accuracy observational studies
  Breaking a complex uncertain question into smaller more tractable sub-questions produces more accurate estimates than judging the whole question at once.
• Explicit probability estimates improve calibration randomised trials
  Forcing a numeric probability rather than a vague verbal hedge makes a prediction falsifiable and trainable.
• Outcome feedback corrects judgment bias randomised trials
  Regularly scoring predictions against actual outcomes and reviewing the misses corrects systematic biases in judgment over time.

Where they actually meet

Both practices replace a private impression with a public, checkable record, and then hold the record up against reality. The time audit does this to a week: instead of trusting how the week felt in retrospect, you write down what actually happened, hour by hour, and compare the total against your prior guess. Calibration training does this to a belief: instead of trusting how confident you feel, you write down an actual number before the outcome is known and compare it against what actually happens. In both cases, the record has to be made before the comparison — you can't audit a week you didn't log, and you can't calibrate a prediction you never wrote down — which is exactly the discipline most people skip, because the private impression already feels like enough information.

The principle underneath both

An internal sense only becomes correctable once it's turned into something concrete enough to be provably wrong later. A feeling that stays inside your head can be endlessly, comfortably redescribed to match whatever you already believe; a written hour-by-hour log or a written seventy percent cannot, because it's sitting in a form a later fact can directly contradict. Neither practice works by making you try harder to be accurate in the moment. Both work by making the estimate specific enough, in advance, that the gap between the estimate and the fact becomes visible afterward — and it's that visible gap, not effort or good intentions, that actually does the correcting.

What that makes the real goal

Closing the distance between how confident or informed you feel and how confident or informed you actually are is what both practices are doing underneath their different subject matter — replacing a subjective sense of accuracy with a measured one. Any place where a felt sense of 'I know how this usually goes' is quietly substituting for an actual record is a candidate for the identical fix, whether or not it involves a calendar or a percentage. The point isn't 'log more' or 'predict better' as ends in themselves; it's making sure the decisions resting on a self-assessment are resting on a real number instead of a comfortable guess.

Other practices that serve the same goal

Food diaries in behavioral weight research pursue the identical objective for eating instead of time or beliefs — people are notoriously bad at estimating what and how much they've eaten, and a written log, checked against the estimate, reliably closes a gap most people didn't know was there. Expense tracking does the same for spending: the gap between what someone guesses they spent on a category and what a tracked log actually shows is usually the whole reason budgeting apps exist. In organizational decision-making, keeping a written decision log — recording a prediction about an outcome before a project starts, then checking it against what actually happened — applies calibration training's exact structure to business bets instead of geopolitical questions, and for the same reason: a team's felt sense of 'we're usually right about this' is unverifiable without a record to check it against.

Using the time audit's habit of writing to sharpen calibration

If you already keep a 168-hour log, you already have the harder half of calibration training in place: the willingness to write something down that might turn out to be wrong, rather than skipping it out of pride or vagueness. Extend the same log with one more column. Before a block of time you're about to spend on a task with an uncertain outcome — a job application, a hard conversation, a piece of work you're not sure will land — write your actual percentage estimate right there next to the hour: '30% chance this gets a callback,' '80% chance this finishes today.' You already have the discipline of writing down something specific before you know the outcome; calibration training just asks you to write down a number instead of a task name. At the end of the week, when you total up your hours the way you already do, also glance back at the predictions and check them against what actually happened. The habit that makes the time audit work — recording before you know how things turn out — is the exact habit calibration training needs, and you've already built it for a different purpose. You are not starting a second practice from nothing; you are pointing an already-working muscle at a second kind of number.

Using calibration training to make your time audit's totals more useful

A time audit is most jarring, and most useful, at the moment you compare your prior guess to the recorded total — but most people run that comparison loosely, thinking 'I figured work took up most of my week' rather than committing to an actual number beforehand. Calibration training's habit fixes exactly this: before you start logging a given week, write down a specific percentage guess for a few categories that matter to you — '60% of my waking hours will go to work-related tasks,' '10% to exercise' — the same way you'd write a forecast before knowing the outcome. Then log the week as usual. When you compare the log to your written guesses instead of a vague prior impression, the gap becomes a number you can actually track over multiple weeks, the same way a forecaster tracks whether their 70%s land 70% of the time. Doing this for a few months turns your sense of your own time use from a single, one-off surprise, 'I had no idea I spent that much time on my phone,' into a trackable pattern of exactly how far off your gut estimate tends to run, and in which direction. That pattern is more useful than any single week's number, because it tells you how much to trust your own guess the next time you make one.

Sunday, comparing the number you wrote down Monday morning to what actually happened

Monday morning, before the week starts, you write a guess in the margin of your log: '20% of this week goes to the side project.' You believe it when you write it — you've been meaning to put real time into it, and twenty percent, a bit under a day and a half spread over the week, feels reasonable and even a little conservative. By Sunday night, totaling the actual hours, the side project comes to just under three percent — a single ninety-minute stretch on Wednesday. The gap isn't really about the side project; it's about how confidently you wrote twenty percent on Monday, with no actual evidence behind that number beyond how the week felt like it should go. You don't get to argue with the log the way you could argue with a vague memory — the hours are written down, block by block, and they don't move. What calibration training adds here isn't a better guess for next Monday; it's the plain fact, now visible, that your Monday-morning percentages have been running wildly overconfident, and that's worth knowing before you make any plan that depends on one again. Next Monday's guess starts from that fact, not from how motivated you happen to feel at the time.

When this is worth reaching for

You are deciding whether to take on a new commitment, and the real question splits in two: do you actually have the hours, and how sure are you that you do. That is when this pairing matters — not while you are still just trying to make time-logging itself a habit; add the prediction layer once tracking a week feels ordinary, not before.

Where this does not apply

If you're logging your time for the first time, skip the prediction layer entirely — learning to record accurately is already a real behavior change, and asking yourself to also generate and later score numeric forecasts on top of it is likely to make you abandon both rather than master either. And calibration training in the strict sense, tracking dozens of predictions to check a percentage against an outcome rate, needs enough predictions to mean anything statistically; running it on three or four guesses about your own week produces a number that feels precise but isn't, and treating a single miss as proof you're 'bad at estimating' is exactly the overconfident leap the practice is meant to prevent.

What the evidence does and does not support

Time-use research backs the time audit's core claim solidly: people's memory of how they spent a week reliably diverges from a contemporaneous log, and that finding replicates across a lot of self-monitoring research beyond time specifically. Calibration training's backing comes mainly from Philip Tetlock's forecasting tournaments, which found that some individuals' accuracy improved measurably when they tracked numeric predictions against outcomes over time — a real, if narrower, result about professional-style forecasting rather than everyday personal guesses. Applying that finding to a private guess about your own weekly hours is a reasonable stretch of the same logic, not a tested claim in its own right; the tournaments never measured anything like a time-use prediction. Treat the combination here as two well-supported ideas borrowed toward a shared purpose, not as a single studied method.