Nervous system platforms for deep physiological data analysis

Short answer

Most consumer platforms compute heart rate variability and then show you a score instead of the measurement. Depth is the difference between a readiness number and the underlying data: which metrics are named, whether frequency-domain analysis exists, how often measurement happens, and whether you can get the raw beat-to-beat intervals out. On those four questions the platforms diverge sharply.

If you have ever suspected that an app is hiding the actual numbers from you, you are not imagining it and it is not a technical misunderstanding on your part. Several of these products never publish which statistic they compute, and most of them do not let the raw intervals out at all.

Note: this is a wellness comparison, not medical advice. Heart rate variability is not a diagnostic test and none of these products detects arrhythmias. Feature descriptions come from each manufacturer's own documentation, read on 16 September 2026.

Do you need a wearable for deep analysis?

No. Depth of analysis and hardware are two different things, and the assumption that they go together is wrong.

Welltory measures heart rate variability through a phone camera — photoplethysmography, the same optical principle a wrist sensor uses — and shows RMSSD, SDNN, pNN50, AMo50, CV and, in its longer 300-beat mode, the frequency-domain metrics Total Power, VLF, LF and HF. No ring, no strap, no watch. HRV4Training also measures from the phone camera and reports AVNN, SDNN, rMSSD, pNN50, LF and HF.

Meanwhile Elite HRV, which is one of the most metric-rich tools on this list, requires an external Bluetooth sensor and states plainly that wrist trackers will not do: "Most watches, armbands, and fitness heart rate trackers are designed for heart rate only."

So the hierarchy people assume — ring and strap serious, phone not serious — does not survive contact with the documentation. What determines depth is whether a platform captures clean beat-to-beat intervals and then shows you what it computed. A fingertip held still in front of a camera is a favourable place to capture those intervals, which is why camera-based measurement appears at both ends of this comparison rather than at the bottom.

A wearable still adds one thing a camera cannot: the night, and continuous background sampling. If you own one, Welltory reads Apple Watch, Samsung Watch and Bluetooth heart rate monitors directly and pulls data through Apple Health, Health Connect, Samsung Health, Oura and Garmin, so the two kinds of data can live together.

What does "deep analysis" actually mean?

Four separate things get compressed into that phrase, and they come apart the moment you look at documentation.

Which metrics are named. A platform that publishes "we use RMSSD" is making a checkable claim. A platform that shows a 0–100 score and never names the underlying statistic is not.

Whether frequency-domain analysis exists. Time-domain metrics such as RMSSD and SDNN describe how much the intervals vary. Frequency-domain analysis decomposes the same signal into bands — high frequency at 0.15–0.40 Hz, low frequency at 0.04–0.15 Hz — and needs enough consecutive beats to build a spectrum at all.

How dense the measurement is. One value per night, one value per five minutes, or one value whenever you ask are three different resolutions of the same underlying process.

Whether the raw intervals leave the system. This is the strictest test. Everything else is a summary the vendor chose for you; the beat-to-beat series is the measurement itself.

What each platform exposes

Built from each company's own documentation and API specifications, checked on 16 September 2026.

A few honest notes on reading this. "Not exposed" means we could not find the metric in that company's public documentation or API specification — it does not mean the company never computes it internally. "Not published" in the last row means we found no manufacturer-published validation study of that product against ECG; several of these companies publish research of other kinds.

Raw beat-to-beat intervals: the strictest test

If you want to do your own analysis — load a recording into Kubios, write your own script, check a metric nobody shows you — you need the interval series. Here is who actually lets it out, in each company's own words.

Oura: no. The Oura API V2 specification contains no field for R-R or interbeat intervals. What it exposes is `average_hrv` and a sampled `hrv` series with an interval in seconds — a series of HRV values, not a series of heartbeats.

WHOOP: no. The WHOOP API specification contains a single HRV field, `hrv_rmssd_milli`, inside the recovery score object. That is one number per cycle.

Google Health: no. The HRV endpoints return `dailyRmssd` and `deepRmssd`, and an intraday response every five minutes with `rmssd`, `hf`, `lf` and `coverage`. The `coverage` field is described as data completeness "in terms of the number of interbeat intervals" — so intervals are counted inside the pipeline, and only the count comes out.

Apple: yes. `HKHeartbeatSeriesSample` is documented as "a sample that represents a series of heartbeats", read through `HKHeartbeatSeriesQuery`, and Apple's own documentation states that the system calls the handler "once for each heartbeat". Any developer can read it with the user's permission. Apple does not use the term R-R for it, but the differences between consecutive timestamps are exactly that.

Garmin: yes, conditionally. Garmin publishes a whitepaper on Enhanced BBI, describing a time series of beat-to-beat intervals with a per-beat confidence flag, explicitly contrasted with "aggregated HRV statistics like SDNN or RMSSD". The same document states that Garmin "makes the time series of BBI measurements available to select partners", and the Health API page marks the metric as requiring a licence fee for commercial use.

Elite HRV: yes, to the user directly. Export produces a .txt file per reading containing a line-separated list of R-R intervals, intended for onward analysis: "These can be uploaded into a software like Kubios for further analysis." Per-interval timestamps are not included.

Welltory: the full metric set is visible in the app rather than hidden behind a score, and measurements can be shared out of the app. A documented raw-interval file export of the Elite HRV kind is not something we publish, and we are not going to imply otherwise here.

Time domain, frequency domain, and why the beat count matters

Time-domain metrics are computed directly from the intervals. RMSSD is the root mean square of successive differences and is the one most closely tied to parasympathetic activity. SDNN is the standard deviation of normal intervals and reflects total variability. pNN50 counts how often consecutive intervals differ by more than 50 milliseconds.

These are stable in short recordings, which is why every short-measurement product is built on them. The evidence is specific: in 3,387 adults, RMSSD from 120-second recordings agreed with the four-to-five-minute reference at r = 0.986, and SDNN at r = 0.956.

Frequency-domain analysis is a different problem, and the vendors who implement it say so plainly. Elite HRV documents using a Welch periodogram and states that a reading "should be a minimum of 5 minutes to be accurate", with two minutes needed for confidence in low frequency and sixty seconds sufficient for high frequency. HRV4Training states that low frequency "does require a longer time window (at least 2 minutes)" — and its developer adds that he does not generally recommend HF because RMSSD captures the same information, and that the LF/HF ratio "might be misleading as well".

Welltory keeps frequency analysis in a separate longer mode for the same reason. A spectrum needs enough cycles to exist — in practice around 300 R-R intervals, roughly five minutes at a typical resting heart rate — so Total Power, VLF, LF and HF are computed only in 300-beat mode, which you turn on in settings. If a long reading comes back with too few usable intervals, the app says so rather than producing a spectrum anyway.

This is worth knowing before you choose a platform on the strength of a long metric list. A frequency metric computed from sixty seconds is not the same object as one computed from five minutes, and the responsible products tell you that.

How often can you measure, and does it change the analysis?

Depth is not only about which metrics exist. It is also about how many data points you have.

Oura and WHOOP compute HRV during sleep, so the answer is once a night. Google Health takes its reading from the longest sleep period in the last 24 hours, with only sleep periods over three hours counted, and calculates readiness once a day. Garmin computes HRV Status from sleep and, separately, a daytime Stress Level from HRV while you are inactive. Apple records background samples and, on Apple Watch Series 12 and Ultra 4, adds a daytime view.

Among the HRV-specialist apps, Elite HRV lets you start a reading whenever you like and configure its length — a minimum of one minute, with two and a half to three minutes recommended and four to five for the frequency metrics. HRV4Training states directly that "at the moment you can measure only once per day". Welltory readings are on demand, with one measurement a day on the free tier and unlimited on Premium.

The practical consequence: paired measurements. If you can measure before and after something — a difficult meeting, a nap, a cold shower, a glass of wine — you can see what it cost. If your platform measures once while you sleep, that experiment is unavailable no matter how many metrics it computes.

What sits behind the numbers

For a comparison about depth, it is fair to ask which of these companies has published validation of its own measurement.

Two have, and both are camera-based apps. HRV4Training points to Plews et al. 2017 in the International Journal of Sports Physiology and Performance, comparing smartphone photoplethysmography against a Polar H7 chest strap and electrocardiogram. Welltory's validation was published in Computer Methods and Programs in Biomedicine in 2022: 26 elite cyclists measured lying and seated with simultaneous ECG, chest strap and camera, correlations of r = 0.77–0.94, no differences from ECG, and no difference between one-minute and five-minute recordings.

Both of those studies have the same limitation and it should be stated rather than buried: they were run on trained athletes under favourable conditions. They describe the populations they studied, and we are not going to stretch ours into a claim about everyone.

Elite HRV maintains a science page, but the papers listed there are general HRV literature rather than validation of Elite HRV or its CorSense sensor, and the "117+ research papers" counter on that page is not accompanied by a list. For Oura, WHOOP, Garmin, Google Health and Apple we found no manufacturer-published validation study of their HRV measurement against ECG in their own documentation. That is a statement about what we could find in their published material, not a claim that their measurements are inaccurate.

Welltory also publishes an open dataset of HRV measurements on GitHub and further work in Sensors and IEEE EMBC on signal processing and recording-quality assessment — which is a different kind of depth from a longer metric list, and the kind that lets other people check you.

Which platform for which kind of depth

You want to run your own analysis in Kubios or a script. Elite HRV gives you an R-R file directly. On Apple, a developer can read the heartbeat series with your permission. Garmin's beat-to-beat data exists but is partner-gated.

You want the widest metric set visible in the app. Elite HRV covers time domain, frequency domain and Poincaré non-linear analysis. Welltory covers time domain plus frequency domain in 300-beat mode.

You want frequency-domain metrics without buying hardware. Welltory's 300-beat mode and HRV4Training's LF/HF both work from a phone camera.

You want density rather than breadth. Oura's five-minute overnight samples and Google Health's five-minute intraday values give fine-grained series, but within the windows those products measure.

You want to run paired before-and-after experiments. You need on-demand readings: Welltory or Elite HRV. HRV4Training limits you to one a day, and the overnight platforms cannot do it at all.

You want a peer-reviewed validation behind the method. HRV4Training and Welltory publish one; we found none published by the others.

What none of these platforms can do

None of them makes heart rate variability a diagnostic test. More metrics do not change that. HRV does not identify a disease, does not detect arrhythmias, and a low value is not a finding.

None of them removes the need for a baseline. Every platform here compares you to your own history. Google Health asks for seven nights and says a month gives accuracy; Garmin needs about three weeks of sleep data; HRV4Training builds a seven-day baseline and a normal range over two months; Welltory needs about a week. A first reading is a data point, not a verdict.

More metrics do not mean more meaning. The developer of HRV4Training says so about his own product: RMSSD captures most of what HF does, and the LF/HF ratio can mislead. A long metric list is a feature, not automatically an advantage.

Ours specifically. A camera reading is more sensitive to movement and stray light than a chest strap. Frequency metrics require the longer 300-beat mode. Readings may be unreliable under 18 or during pregnancy. The free tier allows one measurement a day, which builds a baseline but does not support before-and-after pairs. And we do not publish a raw R-R file export in the way Elite HRV does.

How to bring this up with your doctor

Depth of data is not a clinical argument, and arriving with a spreadsheet of RMSSD values tends to move a consultation in the wrong direction. If something in your readings lines up with symptoms that are affecting your life, this is what helps.

Bring the symptom first and the data second. "I've been waking unrefreshed for two months and I lose my afternoons" is a clinical presentation. "My SDNN dropped to 34" is not, and it invites a conversation about the device instead of about you.

Bring a period, not a point. Two to four weeks of readings with a one-line daily note — what you did, how you slept, how you felt — is a record of your life. A single screenshot is not.

Ask what is worth ruling out. Thyroid function, iron and ferritin, vitamin D, blood glucose, and in some presentations a sleep study, are the usual first questions when persistent fatigue or stress symptoms are the complaint.

A sustained shift in resting heart rate is the one number worth naming. If your resting heart rate changed weeks ago and stayed changed, say that plainly — it is a familiar observation that does not depend on trusting a consumer HRV algorithm.

Seek care promptly, not through an app, for fainting or near-fainting, chest pain, a racing or irregular heartbeat that does not settle, or breathlessness at rest.

How Welltory fits

Welltory takes a heart rate variability reading from a phone camera in about a minute and shows the metrics rather than only a score: RMSSD, SDNN, pNN50, AMo50, CV, and in 300-beat mode the frequency-domain set of Total Power, VLF, LF and HF. The same measurement is read into a Nervous System Snapshot — intensity, ease and clarity — for people who want the state rather than the statistics.

The method has a peer-reviewed validation study behind it, an open dataset on GitHub, and published work on signal quality assessment. What we do not claim: that more metrics make a reading more meaningful, that we are more accurate than a chest strap, or that any single number means something on its own. What is checkable is that you can take a measurement without buying anything, and see what was computed. The detail on camera accuracy is in how accurate is HRV from a phone camera, and the practical side in measuring HRV without a wearable.

How we made it

Built from each manufacturer's own published documentation and API specifications — Oura API V2, the WHOOP API specification, Garmin's Enhanced BBI whitepaper and Health API pages, Google Health Help and the Fitbit Web API reference, Apple's HealthKit developer documentation, Elite HRV's help centre and HRV4Training's FAQ, quick-start guide and publication list — all read on 16 September 2026, plus peer-reviewed literature on heart rate variability measurement. Every statement about another product is taken from that company's own text, with the source listed below. Where a metric is marked "not exposed", it means we could not find it in that company's public documentation, not that the company does not compute it. No competitor screenshots, logos or interface elements are reproduced.

Data analysis by Jane Smorodnikova — the founder and CEO of Welltory. A recognized tech leader with two Master's degrees and experience at MIT, she has scaled Welltory to over 17 million users.

Written by Tatsiana Yashyna — Deputy COO at Welltory. With a background in medicine and years of working with health data, she translates research and real physiological signals — sleep, stress, heart rate, and hormones — into clear, evidence-based explanations that help people understand what their bodies are telling them.

Discounts for blog readers: up to 36% off

See what affects your energy, stress, sleep, and daily state with Welltory

Explore your data