How does Base calculate my maintenance calories?

By back-calculating from the scale. It takes your average logged intake over a window, subtracts the energy implied by your trend weight change at 3,500 calories per pound, and what remains is what you actually expended. That is a measurement of you rather than a prediction from a formula.

How long before maintenance is accurate?

About two weeks of consistent logging and regular weigh-ins. Below that data floor Base falls back to a Mifflin-St Jeor prediction and explicitly labels it as an estimate, because a confident wrong number is worse than an admitted guess.

When should I recalibrate?

When the scale and the prediction disagree persistently. Base flags this itself with a scale check — if your trend rate is meaningfully off what your surplus predicts for a sustained period, the maintenance figure behind it is stale. Also after a big change in training volume, since that moves expenditure directly.

Why is my maintenance lower than an online calculator says?

Usually under-logging rather than metabolism. The calculation reads your logged intake, so anything you consistently miss — cooking oil, sauces, drinks — is subtracted from your apparent maintenance permanently. A calculator predicts from your size; this measures what your scale did against what you recorded.

What is the flux range behind the expenditure line?

One standard deviation of the recent daily estimates, drawn as a shaded envelope. It is an honest measure of how much the estimate is currently moving around, which is what deciding whether to trust today's number actually requires. A fixed plus-or-minus five percent band would look the same and mean nothing.

Recalibrate Your Maintenance Calories in Base

The arithmetic

energyBalance = trend weight change × 3500 ÷ days
maintenance   = average calories logged − energyBalance

Example: 3,000 kcal/day over 14 days, trend up 1.8 lb
         energyBalance = 1.8 × 3500 ÷ 14 = 450
         maintenance   = 3000 − 450      = 2,550 kcal

This is what lets the app self-correct, and the direction is the part worth internalising: gaining 0.9 lb a week on 3,000 calories means maintenance was underestimated, so the target should come down, not stay wrong.

A formula-based app cannot do this. It predicts expenditure from your height, weight, age and an activity multiplier, and it is wrong by however much you differ from the population — permanently, because nothing feeds back.

The data floor

Below a minimum number of logged days and weigh-ins, Base does not compute this. It falls back to Mifflin-St Jeor and labels it as an estimate.

That labelling is the point. A confident wrong number is worse than an admitted guess, because you act on the first and check the second.

The maintenance screen, answering five questions in order: what maintenance is, how confident that is, what you are adding on top, where those extra calories go, and whether the scale agrees with any of it.

Reading the screen

The screen is one argument top to bottom.

Section	Answers
The hero number	What is my maintenance, and how confident is that
Your target	Maintenance + surplus = eat this much
Where the extra goes	What can become muscle versus what can only become fat
What you actually averaged	Your real 7- and 14-day surplus, not the intended one
Recalibration	Whether the figure is stale

What you actually averaged

Worth dwelling on. Your intended surplus is what you set. Your achieved surplus is what you ate, and they are rarely the same.

A 300-calorie surplus intended and 90 achieved explains a stalled bulk completely, and it is invisible on any screen that only shows the target.

Where the extra goes

The surplus is split into what can plausibly become lean tissue and what can only become fat, using the energy costs of each:

a pound of fat  ≈ 3,500 kcal
a pound of lean ≈ 2,500 kcal   (lean tissue is ~70% water, which is free)

Past the rate at which lean tissue can actually be built, the extra energy has nowhere to go but fat. That is the arithmetic behind why a large surplus is pointless rather than merely inefficient, and it is what the partition bar draws.

When to recalibrate

When the scale check flags it. The screen compares your actual trend rate against what your surplus predicts. A persistent gap means the maintenance figure underneath is stale.

After a change in training volume. Adding six hours of endurance a week moves expenditure by a lot, and the estimate will lag it.

Not weekly. The estimate is noisy day to day — that is what the flux band shows — and recalibrating constantly chases noise.

The flux range

The shaded envelope behind the expenditure line is one standard deviation of the recent daily estimates.

It is an honest uncertainty band rather than decoration. The estimate is a back-calculation from a noisy scale against imperfect logging, and the spread of recent estimates is exactly what "how much should I trust today's number" means. A fixed ±5% band would look identical and mean nothing.

A wide band means recent data is inconsistent — usually irregular weigh-ins or gaps in logging.

Common mistakes

Recalibrating every week. Chasing noise.

Blaming metabolism for a low figure. The calculation reads logged intake. Consistent under-logging shifts measured maintenance down by exactly the amount you miss, and it never averages out.

Comparing against an online calculator. Those predict from your size. This measures what happened.

Recalibrating during the first two weeks of a bulk. Early weight gain is substantially glycogen and water, which the 3,500-per-pound conversion treats as fat. The estimate is unreliable there and Base's phase logic knows it.

Ignoring the achieved-surplus rows. They explain most stalled bulks and most faster-than-expected cuts.

Logging inconsistently and expecting a stable number. The flux band will show you exactly how unstable it is.

What to do next

Open the maintenance screen and compare your intended surplus against the fourteen-day achieved figure. If they differ by more than about 100 calories, the problem is logging consistency rather than the target — fix that before recalibrating anything.