The measurement
What happens in the twelve minutes, what is computed from it, and how it leads to one decision.
What is recorded
Your chest strap records RR intervals — the exact time between one heartbeat and the next, in milliseconds. It detects the electrical activity of each heartbeat and reports the interval between successive beats. Everything Today's Load tells you is computed from those intervals and nothing else.
Heart rate, accelerometer data, Apple Health, your notes and your logged training are context and display only. None of them reaches the engine that makes the decision.
The four phases
Three minutes each, twelve in total. Two positions, two breathing conditions, crossed.
| Phase | Condition | What it probes |
|---|---|---|
| Resting baseline | Lying, natural breathing | Your resting autonomic state, with nothing asked of it. |
| After standing | Standing, natural breathing | The orthostatic challenge — the calm branch steps back and the system has to respond to the changed demand of standing. |
| Settling | Lying, paced at five breaths a minute | Whether the calm branch re-engages once the challenge is removed, with breathing held at a fixed rate so the comparison is fair. |
| Standing again | Standing, paced at five breaths a minute | Whether that recovery survives a second challenge, or frays. |
Why paced breathing, and why five a minute
Breathing drives a large part of beat-to-beat variability, so an unpaced comparison partly measures how you happened to breathe. Fixing the rate removes that. Five breaths a minute sits near the rate at which the cardiovascular system's own control loops resonate, which makes the response large enough to read reliably.
Three minutes of real beats, not three minutes on a clock
A phase is complete when it holds three minutes of actual heartbeats, not when three minutes have passed on a clock. If the connection drops, the recording gets shorter rather than quietly becoming something else. Every phase that gets scored contains a real three minutes, or the morning is refused.
The three transitions
The phases are the evidence; the engine also reads each boundary between them as an event in its own right.
Standing up. Did the system answer the challenge? Standing changes the cardiovascular demand within seconds. The measurement reads how the beat-to-beat rhythm changes in response.
Lying back down. Whether the calm branch comes back once you are horizontal and breathing on the pacer.
Standing again. Whether the recovery you just demonstrated holds when the same challenge is repeated.
The four signals
Good regulation is not only answering a disturbance. It is re-establishing an organised state once the disturbance is over, and holding that organisation when the demand comes again. Response, Rebound and Stability read those three parts in turn; Reserve reads the regulatory headroom across the whole challenge-and-return. Each is scored from the phase and transition evidence above.
Response — how effectively does your system respond to the challenge?
How clearly your body answered standing up.
When you stand up, gravity changes the cardiovascular demand within seconds. Your autonomic regulation responds, and that response appears in the changing beat-to-beat rhythm. Response reads how clearly that change is expressed when you stand.
A muted Response means the expected change was less clearly expressed in this measurement. Its meaning is interpreted together with the other signals, not in isolation.
Rebound — how strongly does your system settle back after the challenge?
How fully your calm-down system settled back after the challenge.
The challenge is only half the story. Once you're standing steady, a healthy nervous system hands control back to its calm branch — the parasympathetic side — and your heart rhythm regains its natural, breathing-linked variability. Rebound measures that handing-back: how fully your calm-down system re-engages after being pushed aside.
This is not the same as your watch's "recovery". Rebound is measured in the minutes after a challenge, this morning. A watch's recovery number is usually a countdown estimated from yesterday's training dose. Different questions.
Stability — how well does your system hold its organization when challenged again?
Whether your rhythm stayed steady through the position changes.
Later in the measurement, the system is challenged again under controlled breathing. Stability reads how consistently the beat-to-beat rhythm remains organized when demand is reintroduced.
Reserve — how much regulatory reserve does your system show today?
How much regulatory headroom the system showed across the challenge-and-return.
Reserve describes how much regulatory headroom the system shows across the measured challenge-and-return. It is not "how much energy you have" or a fuel gauge: a rested body can show modest Reserve, and Reserve can stay low for days while everything else works well.
Low Reserve alongside strong regulation is a real and informative pattern, not a malfunction.
What the signals are computed from
The raw material is the standard set of heart-rate-variability measures, the ones defined in the field decades ago and used unchanged since: RMSSD, SDNN, pNN50, the mean interval, the balance of low- and high-frequency power, and a measure of how fractal the rhythm is. What the signals read is not those values in isolation but how they move when the conditions change.
The constants that turn those movements into a score are fixed in a frozen engine. They are identical for every user, they do not adapt to you, and they change only in a versioned engine release.
How four signals become one decision
The four signals are not averaged into a score. They are read together under one fixed logic:
Not every strong signal can compensate for every weak one.
This is what separates Today's Load from a points total. In a weighted score, a very good number in one place can quietly cancel a poor one somewhere else, and the result looks fine. A body that is regulating poorly is not made well by doing something else adequately.
The logic is fixed and the same for everyone. It does not learn, and it changes only in a versioned engine release. We do not publish its thresholds. What we publish is what makes it trustworthy: the same beats always give the same decision.
What reaches the screen
The answer arrives as one of three families — TRAIN, REDUCE, RECOVER — and each family has finer grades inside it. A more conservative grade within a family means the engine itself judged the morning borderline, so a go-ahead can still come with a caution attached: train, but this is not the day to add.
Progression is the one judgement that looks beyond today. An invitation to actually add load asks for a run of recent mornings that were clear, not a single good one, so it cannot appear on your first measurement however good it is.
The morning decision is based on the current measurement against fixed limits; recent clear mornings affect progression eligibility only.
And when nothing fits, nothing is invented. An unrecognised result becomes a refusal rather than a guess.
When Today's Load refuses
A refusal is a feature, not a failure. The quality checks run before anything is scored, and any hard failure stops the scoring outright — you get NO READ, with no number attached.
Not enough signal. Too few usable beats in a phase, or a phase shorter than the protocol requires, because the connection dropped or the session was cut short.
Timing that cannot be real. Too many intervals that no heartbeat produces.
Too much noise. Beat-to-beat jumps that a real heart rhythm does not make — a slipped electrode, a dry strap, movement. The checks are built to tell those apart from ordinary breathing-linked variability, so honest physiology is not thrown out as noise.
The pacer was not followed. The paced phase is checked for the breathing rhythm it was supposed to contain. If it is not there, the comparison it exists to make is not available.
Low confidence. Softer problems do not refuse the read but mark it as limited, and that is shown as a separate neutral chip — never as a worse-looking score.
The app says so plainly when this happens: we couldn't get a clean enough read to give you a decision today — and we won't guess.
What the app does with the result
You get the decision and its family, the four signals with the question each answers and how each one came out, what shaped the morning, and an honest note when the read was of limited quality. If you ask for it, you also get a written explanation of the morning.
What you deliberately do not get is a single number on the front screen. One figure out of a hundred invites exactly the reading Today's Load exists to argue with: a score you trust instead of a reason you can check.
The explanation is free. The daily decision is never behind a paywall.