Sleep tracking without a wearable

A phone on the nightstand is not a sleep lab, and any app that implies otherwise is selling you something. What sleep tracking without a wearable can honestly do is tell you when you fell asleep, when you woke, how broken the night was, and what your bedroom sounded like while it happened. What it cannot do is measure sleep stages. Knowing exactly where that line sits is the difference between data you can act on and a graph you half believe.

What a phone on the nightstand can actually see

Two sensors, running all night, a metre from your head.

The accelerometer records movement. Not your movement specifically, but the vibration your bed and your bedside table transmit when you turn over, sit up, or get out of bed. The microphone records sound. Breathing, snoring, coughing, talking, the neighbour’s door, the heating coming on at 5am.

That is the whole input. Everything an app tells you about a phone only night is derived from those two streams. It is a smaller set of signals than a wrist device gets, and it is also a much better set than most people expect, because the two questions people actually want answered, when did I fall asleep and how broken was the night, are exactly the questions movement and sound answer well.

Movement is the backbone

Inferring sleep from movement has a name and a history. Actigraphy has been used in sleep research for decades, long before consumer wearables existed, on the simple and durable observation that sustained stillness means sleep and repeated movement means wake. Research grade actigraphs are worn on the wrist, but the underlying logic is the same one a phone applies to the nightstand.

It is good at the boundaries. The transition into stillness gives you sleep onset, the point at which you actually fell asleep rather than the point you turned the light off. The gap between those two gives you sleep latency. Stretches of movement in the middle of the night mark the awakenings you may not remember at all, which is the most useful thing tracking does for most people, since human recall of night waking is famously bad.

It has a known weakness, and it is worth stating. Lying awake perfectly still looks like sleep to an accelerometer. That is a limitation of actigraphy in general, not of phones in particular, and it is one reason sound matters.

Sound catches what movement misses

This is where a phone stops being a worse wearable and starts being a different instrument. A wrist device cannot hear your bedroom. A phone can.

Tock runs the microphone alongside the accelerometer and classifies what it hears on the device with Apple SoundAnalysis, tagging snoring, coughing and sleep talking, and saving short clips so you can actually play them back in the morning rather than trusting a bar chart. Hearing yourself snore for forty minutes at 2am is a different kind of information than seeing a number labelled snoring: 40 min. There is more about how that works, and what it deliberately does not keep, in what your phone can actually hear at night.

Sound also fills in the gaps movement leaves. Breathing changes across the night. Room noise explains a lot of awakenings that otherwise look random, which is often the answer when someone is trying to work out why they wake up at 3am.

Where the estimates stop being measurements

Here is the honest part, and the part most sleep apps blur.

Real sleep staging is done with polysomnography, which reads brain activity, eye movement and muscle tone, because stages are defined by what the brain is doing. A sleep spindle, the burst of activity that marks light NREM sleep, is an EEG event. There is no version of a microphone or an accelerometer that detects one. Not with better software, not next year.

So when a phone only night produces a hypnogram with bands of deep, light and REM, those bands are estimates. They are inferred from movement patterns, sound, and what is generally known about how a night is organised, mainly that deep sleep is concentrated early and REM lengthens toward morning. Tock labels them as estimates, and we mean it literally. Measured stages need either a wearable or data brought in through Apple Health.

That does not make estimated stages useless. The overall shape of a night, its sleep architecture at a coarse resolution, is often directionally right, and week to week changes in that shape can be informative. It does mean you should not read Tuesday’s 43 minutes of deep sleep as a fact, and you should be suspicious of any app that invites you to. We wrote about that distinction at more length in what estimated sleep stages really mean, and sleep cycles explained covers what the stages actually are.

What a phone is genuinely better at

There is one thing this setup does that a wearable typically does not, and it is the reason to have it running at all.

Because the phone is already awake, listening and sensing movement, it can act on what it finds. Tock’s smart wake watches the 30 minutes before your set alarm across both channels, movement and sound, and rings early when it finds a lighter moment. Your set time is always the fallback, so the worst case is the alarm you set anyway.

In Tock’s own replay testing across 17 nights, smart wake fired on 15 of them, a median of 14 minutes early, ranging 3 to 29 minutes. That is our own testing on a small sample, so treat it as a shape rather than a guarantee, but the shape is the point: most nights it finds something, and it is usually a modest shift rather than a dramatic one. Waking out of a lighter moment is the most direct lever anyone has on sleep inertia, which is the part of the morning most people are actually trying to fix.

Who should just buy a wearable instead

Some people should, and pretending otherwise would be dishonest.

Buy a wearable if you want heart rate, resting heart rate, HRV, blood oxygen, respiratory rate or wrist temperature, because those need skin contact and a phone on a table will never have it. Buy one if staging accuracy specifically matters to you. Buy one if you travel constantly and sleep without a bedside table, or if you and your partner keep very different hours and the room is never quiet.

Worth knowing: Tock has no Apple Watch app. If you already wear one, Tock reads its data through Apple Health, so measured stages, heart rate and the rest fill in the parts a phone cannot reach, and the sound layer sits on top. That combination is better than either alone.

Skip the wearable if you do not want to charge another device, do not want something on your wrist at night, or already know that a band on your arm keeps you awake. The reason phone tracking is worth taking seriously is not that it beats a wrist device on sensors. It is that it is the tracker you will still be using in three months.

The numbers worth watching week to week

Ignore the nightly score. Watch four things over a fortnight.

Total sleep time against time in bed tells you whether you have a sleep problem or a schedule problem. Nine hours in bed and six and a half asleep is not the same complaint as six and a half in bed and six and a half asleep, and the fix is completely different.

Sleep onset consistency tells you whether your body clock is stable, which matters more than any single night’s duration.

Awakenings tell you whether the night is fragmented, and paired with the audio, often why.

Snoring minutes are worth watching as a trend, especially against alcohol and sleeping position, and worth escalating to a doctor if the trend is bad.

Tock keeps 21 metrics in 7 groups, but four of them carry most of the signal, and none of the four depend on staging being right.

Setting your phone up so the night is worth reading

Placement first. Flat on the nightstand, face up, screen unobstructed, within roughly an arm span of your head. Not under a pillow, not in a drawer, not on the mattress, where you will record your partner turning over as reliably as yourself.

Plug it in. All night audio classification and motion sensing use real battery, and a phone that dies at 4am has recorded you three quarters of a night.

Leave the good night screen up. It dims to near black after 20 seconds untouched and stays on all night, so a glance tells you the tracking is running, and a tap wakes it without lighting up the room.

Then give it two weeks before you draw any conclusions. One night of data is noise. Ten nights is a pattern, and a pattern is the only thing here worth changing your evening over.