What do I need to set up on my first day with Base?

Four things: your profile facts (sex, birth year, height), your current bodyweight, your health data connection, and one day of food logged. That is enough for the app to produce a calorie target and start a trend line. Your training split already exists — a default six-day split is installed on a fresh account, so you can lift on day one without designing anything.

How long before Base's numbers are accurate?

About two weeks. Adaptive maintenance is back-calculated from your trend weight against your logged intake, and it needs enough weigh-ins and enough logged days to separate a real change from scale noise. Until then the app uses a Mifflin-St Jeor estimate and labels it as an estimate rather than presenting it as a measurement.

Do I have to log every single day for Base to work?

No, but the trend weight and the adaptive maintenance figure both get better the more days they have. A missed day is a gap the maths carries forward rather than a zero. What breaks the numbers is logging some days and not others in a pattern — logging weekdays but not weekends makes your average intake look lower than it is, which makes maintenance look lower than it is.

Should I set up my training split before I start logging?

No. A fresh account already runs a default six-day split, so you can start lifting immediately. Designing a custom split is worth doing once you have trained a week or two and know what you actually want to change. The builder lives on the Workout tab and never touches the program you are currently running until you finish it.

Does Base work without an Apple Watch or a fitness tracker?

Yes. Sleep, recovery and heart data need a wearable, and those screens will honestly say there is no data rather than inventing one. Everything else — food, bodyweight, lifting, trend weight, adaptive maintenance, progress photos — works from what you log yourself.

Getting Started With Base: Your First Week

Why the order matters

Base derives almost everything from your own data rather than from a lookup table. That is the point of it, and it has one consequence worth understanding before you start: the app is least accurate on the day you install it, and gets better roughly linearly for a fortnight.

On day one there is no trend weight, so there is no measured maintenance. The app falls back to a Mifflin-St Jeor prediction and says so. Two weeks later it has back-calculated your actual expenditure from what the scale did against what you ate, and that number is yours rather than a population average.

So the setup order is chosen to start the slowest-converging measurements first. Weight before food, food before training design, training design last.

The four things

Your profile facts

Three numbers Base cannot measure and will not guess: sex, birth year, height.

Each is load-bearing somewhere specific. Sex and height feed the Mifflin-St Jeor constant. Birth year sets the age band used for heart-rate variability, resting heart rate and body-fat comparisons, and it is the only honest basis for a maximum heart rate. Without a birth year you get no heart-rate zones rather than a made-up 190.

The three profile facts, each labelled with what it is used for. Every field saves as you type — there is no Save button, because a half-entered height is simply a height that has not been entered yet.

Your bodyweight

Weigh yourself and log it. This is the single highest-value thirty seconds of setup, because the trend weight is an exponentially weighted average and it needs a starting point.

If you entered a weight during onboarding, that already counted as your first weigh-in and you can skip this.

Daily weigh-ins are better than weekly, and the reason is arithmetic rather than diligence. Day-to-day scale noise from water, food volume and glycogen is routinely two to three pounds. A weekly weigh-in gives you one noisy sample per week, so a genuine change of half a pound is invisible underneath it. Seven noisy samples smooth into a usable signal. Read how trend weight turns a noisy scale into a signal for the actual smoothing.

Your health connection

Base reads from Apple Health on iOS and Health Connect on Android. Grant it access and active energy, resting energy, sleep, heart rate and workouts flow in.

This is what makes the calorie target move when you move. Without it the target falls back to your committed maintenance plus your surplus, which is a fixed number and cannot know that Saturday had a two-hour ride in it. Full detail in connecting your health data.

One complete day of food

Log a whole day, start to finish, including the things you would normally not bother with.

Completeness matters more than precision here. A day where you logged four meals accurately and forgot the milk in three coffees is worse than a day where you estimated everything roughly but got all of it, because adaptive maintenance is computed from your average logged intake — and a consistent underlog does not cancel out. It shifts your measured maintenance down by exactly the amount you consistently miss.

A logged day on the Food Log. The line under the macros spells the target out — maintenance plus surplus, and what is left — so the number is never unexplained.

What can wait

Task	When	Why not day one
Custom training split	Week 2+	A default six-day split is already installed and running. Design a new one when you know what you want to change about it.
Endurance thresholds	Week 1, any day	Needs your run and ride history, which the app scans from your health store. Takes minutes and does not depend on anything else you set up.
Progress photos	Day one if convenient	The first set is the baseline everything later is measured against, so earlier is better — but it costs nothing to start in week two.
Micronutrients	Week 2+	Only meaningful once several days of food are logged, since coverage is computed from the share of your calories that carried a nutrient profile.
Supplements	Anytime	Independent of everything else.

The arithmetic you should sanity-check in week one

After about seven days, open the maintenance screen and check the number against a rough hand calculation. If they disagree wildly, something is wrong with your logging rather than with the app.

maintenance ≈ average daily calories logged − (trend weight change × 3500 ÷ days)

Example: 2,900 kcal/day average over 14 days, trend up 1.2 lb
         2900 − (1.2 × 3500 ÷ 14)
       = 2900 − 300
       = 2,600 kcal/day maintenance

If that lands 600 calories below what you would expect for your size and activity, the usual cause is under-logging rather than a slow metabolism. Cooking oil, sauces and drinks are where it hides.

Common mistakes

Designing a training split before training once. The builder asks what you want on each day, and on day one you do not have an opinion worth acting on. Run the default, notice what annoys you, then change that.

Logging only weekdays. This is the single most damaging pattern, because it biases your measured maintenance downward by however much the unlogged days differ. Either log everything or log nothing; a partial week is worse than no week.

Weighing in at different times of day. Trend weight is robust to noise but not to a systematic shift. Pick a time — first thing, after the bathroom, before food — and keep it.

Chasing the calorie target on day one. It is a Mifflin-St Jeor estimate until there is data. Treat the first fortnight as measurement rather than execution.

Turning on every screen at once. Micronutrient coverage, weekly volume per muscle, ramp rate and recovery scoring are all genuinely useful and all meaningless with two days of data behind them.

Skipping the RIR rating on logged sets. Weekly volume only counts sets it knows were close enough to failure. Sets logged without a rating are reported separately rather than assumed, so an unrated week reads as a light week.

What to do next

Enter your three profile facts and weigh yourself today, then log one complete day of food — including the coffees. Come back in a week and open the maintenance screen to see the first real estimate of what your body actually costs to run.