Real-Time Data: What Continuous Signals Can and Cannot Tell You
Continuous data fills the gap between annual blood panels, and only a few of the available streams are reliable enough to act on.
The Short Answer
Blood panels are precise and infrequent. Behaviour changes daily. That gap is what continuous data fills: a member's sleep, activity and autonomic signals between draws, which is where most of the actual variation in how someone feels and functions lives. The constraint is that only some of the available streams are validated, and the ones that get the most screen space, sleep staging and proprietary readiness scores, are among the least reliable.
What Continuous Data Adds
Frequency. An annual panel is one observation. Sleep regularity, activity volume and resting heart rate are observed daily, which makes trends visible on a timescale where behaviour can respond.
Context for a blood result. An hs-CRP drawn three days after an unaccustomed hard session, or two weeks after an infection, is misleading, and continuous data supplies the context that makes the panel interpretable.
Behavioural rather than biochemical information. Wearables measure what someone did, which is upstream of most biochemistry and is the part that can be changed.
Early signals. Resting heart rate rising with HRV falling often precedes both illness and the subjective sense of accumulating load.
Adherence. Whether a training or sleep recommendation was actually followed, which is the variable that explains most null results and is otherwise reported by self-report.
That last one is unglamorous and consequential. Most protocols that appear not to work were not followed as intended, and continuous data distinguishes a failed intervention from an unexecuted one.
Which Streams Are Reliable
| Signal | Reliability | Act on it? |
|---|---|---|
| Sleep and wake timing, regularity | Good; derived from timing rather than staging | Yes, first |
| Activity volume, steps | Good | Yes |
| Resting heart rate | Good | Yes, as a trend |
| Heart rate during steady exercise | Good with a chest strap, moderate at the wrist | Yes |
| Total sleep time | Moderate; slightly overestimated | Yes, as a trend |
| Heart rate variability | Real measurement, highly variable interpretation | Multi-week trend only |
| Estimated cardiorespiratory fitness | Moderate; trend useful, absolute value not | As a trend |
| Sleep staging | Poor against polysomnography | No |
| Readiness and recovery scores | Proprietary, undisclosed derivation, unvalidated | No |
| Stress scores | Usually an HRV derivative with an interpretive layer | No |
| Blood oxygen saturation | Variable; affected by skin tone, motion and fit | Persistent patterns only |
The pattern is consistent: the plain measurements and their trends are reliable, and the interpreted composites that occupy the most screen space are not. A readiness score combines several inputs by an undisclosed formula with no published predictive validity, and it may correlate with something useful while nobody outside the company can say what.
Continuous glucose monitoring is a separate class of device. It is genuinely informative and its interpretation in non-diabetic use is unsettled, with normal excursions frequently misread as dysfunction.
The Reaction Problem
Continuous data invites daily reaction, and daily reaction to noisy signals produces worse decisions than no data at all.
Day-to-day variation is large in HRV, sleep scores and to a lesser extent resting heart rate. A single value differing from yesterday usually reflects variation rather than change.
Reacting to noise produces protocol churn. Changing training, supplements or sleep strategy on the basis of one morning's reading makes attribution impossible and generates a stack that oscillates without direction.
The measurement can degrade the thing measured. Sleep is unusually sensitive to attention, and anxiety about scores produces exactly the poor sleep the score reports. Clinicians in sleep medicine encounter this specifically.
Score chasing optimises an unknown function. Improving a proprietary composite whose derivation is undisclosed means optimising something you cannot inspect.
Displaced attention. Time spent reviewing dashboards is not time spent training, cooking or sleeping.
The discipline that fixes it: review weekly rather than daily, read rolling averages against a baseline rather than single values, ignore proprietary composites, and stop tracking anything that is making the underlying behaviour worse.
How Continuous and Periodic Data Combine
The two kinds of data answer different questions, and their value is in combination.
Continuous data explains a periodic result. A raised hs-CRP with recent illness or hard training in the record is explained; the same value without that context invites an unnecessary investigation.
Periodic data validates a continuous signal. A fitness estimate trending upward is corroborated by heart rate at a fixed workload falling; a sleep improvement is corroborated by better glycaemic markers.
Continuous data measures adherence, periodic data measures effect. Together they distinguish an intervention that did not work from one that was not followed, which is the single most useful distinction in this whole area.
Continuous data catches the timing errors that ruin a panel. Recent illness, hard exercise and poor sleep in the days before a draw are all visible.
Neither substitutes for the other. Wearables are blind to biochemistry: no device measures apolipoprotein B, HbA1c, inflammatory or nutrient status, and those inform long-term trajectory more than anything on a wrist. A person tracking only continuous data has a detailed record of behaviour and none of physiology.
What Would Make This Better
The realistic near-term improvements are in inputs rather than in models, which is worth knowing for anyone deciding what to buy.
Validated rather than proprietary metrics. Published derivations and independent validation would convert several current composites from entertainment into signals.
Better sleep staging, which remains the largest accuracy gap between what devices report and what they can measure.
Non-invasive blood pressure, which would be transformative given how much blood pressure predicts and how rarely it is measured properly.
Cheaper and more frequent biochemistry, which is the real gap. A quarterly apoB and HbA1c would add more than any wearable improvement.
Interoperability, so a member's record is not split across incompatible platforms.
What would add least: more metrics. The constraint is not the number of signals available, it is that few of them are validated and fewer still change a decision. A modest set of reliable inputs read as trends outperforms a dashboard of forty numbers, most of which are estimates of estimates.
A Practical Standard
For anyone assembling their own version of this, the criteria are the same ones a platform should apply.
Use signals with published validation. Timing, activity, heart rate. Not composites.
Read trends, not values. Rolling averages against a personal baseline established over weeks.
Record context. Alcohol, illness, travel, training and stress explain most deviations.
Review weekly. Frequent enough to catch drift, infrequent enough to avoid reacting to variation.
Pair with periodic biochemistry. Annual panel at minimum, since behaviour and physiology are different records.
Change one thing at a time, and use the continuous data to confirm the change was actually made.
Stop tracking what is not helping. A legitimate outcome rather than a failure, particularly for sleep.
Used this way, continuous data is genuinely useful and it is a modest instrument rather than a health platform. The reliable signals are few, unglamorous and trend-based, which is a fair description of most things that work in this field.
The AEONNN Perspective
The Real-Time User layer exists to fill the gap between annual panels, and AEONNN reads a deliberately narrow set of streams: sleep and wake timing with regularity, activity volume, resting heart rate, heart rate at a fixed workload, total sleep time as a trend, and HRV only as a multi-week trend.
The Quality layer excludes what occupies the most screen space. Sleep staging agrees poorly with polysomnography, and readiness, recovery and stress scores are proprietary composites with undisclosed derivations and no independent validation of predictive value. The raw signals underneath them are more interpretable than the scores built on them.
The most useful thing this data does is unglamorous: it distinguishes an intervention that did not work from one that was not followed, which explains most null results. It also supplies the context that makes a blood panel interpretable, since recent illness, hard training and poor sleep before a draw are all visible. The platform reviews weekly rather than daily, because daily reaction to noisy signals produces protocol churn and, in the case of sleep, can degrade the thing being measured.
Pillar Matrix mapping
Longevity and Biological Age, Sleep and Circadian Regulation
Database Matrix layers
- Real-Time User Layer (wearable and adherence signals)
- Quality / Formulation Layer (ConsumerLab, Labdoor)
- Evidence Layer (PubMed, Cochrane, ClinicalTrials.gov)
- Meta / Consensus Layer (JAMA, BMJ, specialty society positions)
Frequently Asked
What does continuous data add over blood tests?
Frequency, behavioural rather than biochemical information, early signals, context that makes a blood panel interpretable, and a measure of whether a recommendation was actually followed.
Which wearable signals are reliable?
Sleep and wake timing with regularity, activity volume, resting heart rate, heart rate during steady exercise, and total sleep time as a trend. HRV is reliable as a multi-week trend only.
Why ignore readiness and recovery scores?
They are proprietary composites with undisclosed derivations and no independent validation of predictive value. The raw signals underneath them are more interpretable.
Why not review the data daily?
Day-to-day variation is large, so a single value usually reflects noise. Reacting to it produces protocol churn, and in the case of sleep the measurement can degrade what it measures.
What is the most useful thing continuous data does?
It distinguishes an intervention that did not work from one that was not followed, which explains most null results and is otherwise left to self-report.
Can wearables replace blood tests?
No. No device measures apolipoprotein B, HbA1c, inflammatory or nutrient status, and those inform long-term trajectory more than anything a wrist device reports.
What would improve this most?
Better inputs rather than better models: validated rather than proprietary metrics, better sleep staging, non-invasive blood pressure, and cheaper more frequent biochemistry.
Evidence and review
Any dosage ranges cited here reflect the ranges used in published human trials, not personal recommendations. Evidence in this field moves, so this article is reviewed quarterly and carries its last-updated date above. Nothing here is intended as medical advice, and supplementation should be discussed with a qualified clinician, particularly alongside prescribed medication or an existing condition.