Psychology students a few weeks from handing in an honors thesis were asked when they would submit it. On average they said 33.9 days. They took 55.5.

Then they were asked to predict again, assuming everything went as badly as it possibly could. They said 48.6 days. They still took 55.5, and fewer than half of them made even their own worst case.[3]

Those were people with one project, one deadline, and years of experience watching their own work run late. Now think about how quickly you could tell me when you do your best thinking.

I held my own answer for years on nothing sturdier than the feeling of having been right about it. Mornings were mine, so mornings got the hard work, and everything past two o'clock got whatever was left over. It never occurred to me that I had built a schedule on a guess and then used the schedule as proof of the guess.

The mental energy guide names the method that fixes this. Cognitive Budgeting matches your most demanding work to the conditions where you tend to do it best. This page runs it properly: the tracking week, the sorting test, the placement rules, and the part nobody writes about, which is how any of this survives a calendar that belongs half to other people.

One caution before the method. Cognitive Budgeting is a planning approach rather than a measure of anything. It cannot tell you how much thinking you have left, because there is no such quantity sitting there to be read.

Your opinion about your own pattern is probably wrong

Time of day does move cognitive performance. A review of circadian rhythms in cognition found that time-of-day fluctuations affect attention, executive function, and memory, with the size and direction depending on a person's circadian preference and whether testing happened near their peak.[1]

So the windows are real. Which window is yours stays an open question, and the research does not close it for you. A 2025 systematic review of chronotype and cognition found that more than 80% of studies showed no main effect of chronotype on performance, with some support for better results when a task matched a person's preferred time.[2]

Read those two together and you get a narrow, useful conclusion. Something varies across the day. Your morning-person identity does not reliably predict what.

Which brings back the thesis students from the top of this page. Buehler, Griffin, and Ross ran that study to test a specific idea: people predict their own completion times from an imagined future rather than from what actually happened last time.[3] Fewer than a third finished inside their own best estimate.

Their predictions still carried information, which is the part I find genuinely interesting. Predicted and actual times correlated at r = 0.77, so students who expected to take longer did take longer. The ranking held even while every number in it ran short. Your instinct about your own pattern is probably pointing in a real direction and mis-stating the size of it.

Those were university projects rather than a founder's week. The bias travels well enough to justify one week of writing things down instead of another year of assuming.

What to write down, three times a day

Three entries a day. Late morning, mid afternoon, and end of day.

Each entry has four parts:

  • The time. Clock time, no rounding to the nearest tidy hour.
  • Your alertness, 1 to 9. One is fully alert. Nine is fighting to stay awake.
  • What you just finished. Four words is plenty. "Pricing call." "Wrote landing copy."
  • Whether your attention held. Steady or scattered. Nothing more graded than that.

The 1-to-9 rating is doing real work, and it is not something I invented for the exercise. It borrows the Karolinska Sleepiness Scale, a single self-rating validated against instruments. In a study of 16 healthy women aged 33 to 43, measured eight times a day across three days, higher ratings tracked slower median reaction times, more attention lapses, and higher alpha and theta activity on EEG.[4]

Sixteen women in a laboratory is a small, narrow sample, and the scale measures sleepiness rather than the fuzzier thing you mean by mental energy. Take from it the point that matters here. A one-number gut rating is not worthless data. It moves with things a machine can measure, which is more than most productivity tracking can claim.

Keep the whole thing on paper or in one note. Any tool you have to open, choose, and configure becomes a fifth task, and the tracking week dies of admin before Wednesday.

Reading a week that looks like noise

Most people expect the log to produce a clean curve. Most weeks do not.

Three shapes show up. A clear peak is the easy case: two or three hours where the ratings stay low and attention holds, usually in the same part of the day. A trough with no peak is more common, where nothing looks strong and one stretch after lunch looks distinctly bad. And plenty of weeks look like nothing at all, where the numbers move around with no relationship to the clock.

That third result is where people quit, and it is usually the most informative week of the three. When alertness does not track the time of day, it is tracking something else, and your log already recorded what.

Sort the entries by what you had just finished rather than by clock time. Ratings that cluster after back-to-back calls point at switching load rather than any circadian window. Four experiments on task switching found measurable time costs from switching, growing as the task rules got more complex and shrinking when a cue let people prepare.[7] Lab tasks cannot price the jump from a sales call to a cash-flow review. Use the finding to decide where to look in your own log, and let your own log supply the size.

If the shape is still unreadable after a week, run a second week before changing anything. Two thin weeks beat one confident misreading.

One honest complication. The act of tracking changes what you are tracking. A meta-analysis of reactivity to in-the-moment measurement found small but meaningful pooled effects on behavior simply from being measured, around 0.27 to 0.30, with most of that evidence coming from physical activity studies.[5]

Your first week is therefore not a clean baseline. You will notice the two o'clock slump more because you agreed to rate yourself at two o'clock. That contamination is worth accepting, since the goal is a working schedule rather than a publishable measurement.

Which work actually deserves the strong window

Most attempts collapse right here. People finish the tracking week, find their strong hours, and then fill them with whatever felt urgent on Sunday night.

Urgency and cognitive demand are different properties. A vendor chasing an invoice creates urgency and asks almost nothing of your thinking. A pricing decision you have postponed twice asks a great deal and generates no pressure at all until the day it goes wrong.

Sort a task into the demanding pile when it meets most of this:

  • It requires holding several unrelated things in mind at once.
  • Getting it wrong is expensive or slow to undo.
  • It needs original output rather than a response to someone else's.
  • You would want to be at your sharpest if a competent person were watching you do it.

Pricing, positioning, hiring calls, financial review, and first-draft writing usually qualify. Approvals, scheduling, most inbox work, CRM cleanup, and routine check-ins usually do not, whatever their deadline says.

Run the sort once on a full week of your real calendar. Nearly everyone finds at least one item that had been eating a strong window for months on the strength of feeling important.

One demanding task, one empty window

The placement rule is deliberately blunt. One demanding task per strong window, and one strong window left empty.

The single-task limit exists because stacking two hard tasks into one window means paying a switch between them at the exact point you have the most to lose. Grouping similar work and cueing the transition in advance recovers part of that cost.[7]

Give the entry a written trigger rather than an intention. Plans that state when, where, and how you will act had a positive effect on goal attainment across 94 tests.[8]

When I close my laptop after the 8:45 standup, I will open the pricing model and write the three constraints the number has to respect.

Those studies covered goal pursuit broadly and did not test Cognitive Budgeting, which no study has. The mechanism worth borrowing is narrow: the plan removes one decision from the moment you are least equipped to make decisions.

The empty window is the part people cut first, and cutting it is what breaks the system by Wednesday. A week with every strong hour committed has nowhere to put the call that runs long, the client emergency, or the recovery that follows a hard conversation. Protecting one unbooked window is what lets the other four survive contact with a real week.

Every framework sounds reasonable until you try to write it down, so here is the artifact. This is an illustration of the finished output rather than anyone's tracked data.

| | Strong window | Middle | Low window | |---|---|---|---| | Pattern from week one | 8:30 to 11:00, steady | 11:00 to 2:00, variable | 2:30 to 4:30, scattered | | Monday | Pricing model rebuild | Client calls | Invoices, expense approvals | | Tuesday | Q3 strategy draft | Team 1:1s | Inbox, scheduling | | Wednesday | Protected, nothing booked | Sales calls | Prep tomorrow's inputs | | Thursday | Hiring decision | Vendor review | CRM cleanup | | Friday | Financial review | Buffer | Plan next week |

The strongest window holds exactly one demanding task per day. Wednesday's is deliberately empty, because a week with no slack breaks the first time a meeting runs long. The low window is full of real work, just work that tolerates a scattered mind.

That grid assumes you can move things. A realistic version looks like this. A four-person agency owner tracks a week and finds their steadiest attention between 9:00 and 11:30. That block already holds four client calls, because clients ask for mornings and the answer has always been yes. Clearing the mornings is not on the table. Moving two of the four calls to Tuesday and Thursday afternoons is, which leaves Monday and Wednesday mornings for proposal and pricing work and keeps Friday morning unbooked. Two protected windows out of five is the version that survives.

When you do not own your morning

Every version of this advice assumes you control your calendar. Founders with clients, children, staff, or a timezone problem control roughly half of it, and that is the case worth planning for.

Protect one window a week to start. One demanding task landing in a strong window every week beats a five-window system you abandon by the ninth day.

Trade rather than clear. Moving two of five recurring calls out of your strong hours is achievable. Announcing that you no longer take morning calls usually is not.

Split the decision from the execution. When the strong window is only 20 minutes long, use it to make the call rather than do the work. Deciding the pricing structure needs your best thinking. Building the spreadsheet that implements it does not.

Use the second-best window on purpose. A middling window you actually control beats a strong window you have to fight for every week. The pattern is a planning input rather than an instruction.

Let a fixed early start reshape the map. If a 6:00 a.m. school run puts your strong hours somewhere unusual, those unusual hours are your pattern. The method never required mornings.

Sometimes the constraint has nothing to do with the calendar. When a stirred-up mind is what you keep bringing to demanding work, the repair is the 10-minute stress reset before the window opens. Rearranging the week will not reach that problem.

When the budget stops working

Placement drifts. A recurring meeting lands at 9:15, then a second one, and within two months the protected window is gone without anyone deciding to remove it.

Re-run the tracking week when the schedule changes materially, when a season changes, or roughly quarterly. It costs a week and it is the only way to catch drift, because drift never announces itself.

Two failure modes deserve naming. The first is treating a bad week as a failed method. Illness, a launch, a bad stretch of sleep, or travel will flatten the pattern, and the correct response is to run the week again later rather than conclude that you have no pattern.

The second is using the framework against yourself. A log showing scattered attention at 3 p.m. describes a Tuesday. It does not describe your capability, and a planning tool that turns into a scorecard gets abandoned for good reason.

Cognitive Budgeting also has a hard boundary. It arranges work around fatigue and does nothing to treat fatigue. A systematic review of state mental fatigue found that sustained cognitive work may produce both weariness and reduced performance, without identifying a single cause or a fixed quantity to be topped up.[6] Breaks help less than people assume: a meta-analysis of micro-breaks found small gains in vigor and lower fatigue without a reliable overall gain in performance.[9]

Run it for one week

  1. Log three entries a day for five working days. Time, a 1-to-9 alertness rating, what you just finished, and steady or scattered. Change nothing else this week.
  2. Sort next week's calendar into demanding and routine. Use the four criteria above rather than what feels urgent.
  3. Place one demanding task in your strongest window and leave one strong window empty. Write the if-then trigger for the task you placed.

Every part of that runs on paper. If a repeatable audio cue helps you mark the start of a protected window, Awakened Mind includes programs built for that use. It is a general wellness tool, results vary, and the method above works without it. Try Awakened Mind on the Apple App Store.

Once the window is placed, the remaining problem is entering the work quickly, which is a separate process with its own guide: the Deep Work Runway. If the task keeps stalling because the decision underneath it is unresolved, start with decision fatigue instead. Cognitive Budgeting sits inside the broader practice of mental state management.

You already know how to protect time. This is how you find out what the time was worth.

References

  1. A time to think: circadian rhythms in human cognition. Cognitive Neuropsychology, 2007. Link Limit: a review of laboratory studies. Establishes that time-of-day effects exist without identifying any individual's window.
  2. Chronotype and synchrony effects in human cognitive performance: a systematic review. Chronobiology International, 2025. Link Limit: more than 80% of studies showed no main effect of chronotype. Findings were mixed.
  3. Exploring the "planning fallacy": why people underestimate their task completion times. Journal of Personality and Social Psychology, 1994. Link Limit: university students predicting academic projects. Covers duration estimates rather than judgments about cognitive demand.
  4. Validation of the Karolinska sleepiness scale against performance and EEG variables. Clinical Neurophysiology, 2006. Link Limit: 16 healthy women aged 33 to 43 in a laboratory. Validates a sleepiness rating, not a measure of mental energy or work quality.
  5. A systematic review and meta-analysis of studies of reactivity to digital in-the-moment measurement of health behaviour. Health Psychology Review, 2022. Link Limit: 81% of included studies measured physical activity. Effects on other behaviors were mixed and based on few studies.
  6. Neural mechanisms underlying state mental fatigue: a systematic review and activation likelihood estimation meta-analysis. Reviews in the Neurosciences, 2022. Link Limit: different tasks and methods across included studies. No single cause, brain marker, or fixed quantity of mental energy identified.
  7. Executive control of cognitive processes in task switching. Journal of Experimental Psychology: Human Perception and Performance, 2001. Link Limit: simple laboratory tasks. Cannot price real switches such as a sales call to a financial review.
  8. Implementation intentions and goal achievement: a meta-analysis of effects and processes. Advances in Experimental Social Psychology, 2006. Link Limit: 94 independent tests across varied goals. None tested scheduling by tracked alertness.
  9. "Give me a break!" A systematic review and meta-analysis on the efficacy of micro-breaks. PLOS ONE, 2022. Link Limit: no reliable overall performance gain. Demanding mental tasks may need longer recovery.
  10. Fatigue and Work. National Institute for Occupational Safety and Health, 2026. Link Limit: broad occupational guidance. Does not identify which factor caused an individual slump.