There is a device on a lot of wrists right now that measures heart rate every few seconds, estimates sleep stages, tracks movement, and in some cases takes an electrocardiogram. Twenty years ago that was hospital equipment.

And for most owners, it produces exactly one outcome: a mild sense of failure every morning when the sleep score is yellow.

That’s a waste, and it isn’t the wearer’s fault. These devices are sold on the promise of data and then hand you a wall of it with no instruction on which parts to care about. So let’s fix that.

Why most people quit within a month

Three predictable reasons, and understanding them is most of the battle.

They watch daily numbers. A single night’s sleep score is close to meaningless. It bounces around based on the alcohol, the room temperature, the argument at 9pm, whether the strap was tight. Watching it daily is like judging your business on one afternoon’s takings.

They collect without deciding. Data with no decision attached is just noise you feel bad about. If a number is never going to change what you do, stop looking at it.

They chase the score, not the outcome. The score is a proxy. The outcome is whether you feel decent at 3pm. When those two disagree, the way you feel is the more reliable instrument.

The purpose of the data is not to be measured. It is to catch you being wrong about yourself — and you are, in at least one predictable way.

The three numbers that matter

1. Sleep duration — not sleep score

Ignore the composite score. Ignore the sleep stage breakdown, which consumer devices estimate rather than measure and get wrong more often than the pretty graph suggests.

Watch one thing: hours actually asleep, averaged over a week.

That single number is where nearly all the value is. Most people who feel persistently flat are not suffering a mysterious deficiency. They are averaging six hours and telling themselves they’re fine on six hours.

What to do with it: look at the weekly average, once a week. If it’s below seven, the intervention isn’t a supplement or a new mattress. It’s an earlier bedtime, which is boring, free, and the thing nobody wants to hear.

2. Resting heart rate

Your heart rate while genuinely at rest, usually captured overnight. The absolute number matters far less than your normal and how it moves.

It is one of the more honest signals a consumer device produces. It tends to drift down as fitness improves over months, and it spikes when your body is dealing with something — illness coming on, alcohol, poor sleep, stress, overtraining.

What to do with it: know your baseline. When it’s sitting five or more beats above that for a few days without an obvious cause, treat it as a nudge to back off, sleep more, and not schedule anything heroic.

3. Time on your feet

Not steps, exactly — the ten-thousand figure was originally a marketing number, not a clinical one. What matters more is how much of your day is spent not sitting.

What to do with it: don’t chase a target. Look at your worst days and ask what they have in common. For most people it’s a specific recurring thing — a long meeting block, a commute, a particular kind of workday — and once you can name it, you can break it up.

The numbers that are mostly noise

Not because they’re fake, but because consumer-grade accuracy plus daily variability means the signal is buried.

• Sleep stages. Estimated from movement and heart rate. Interesting; not actionable.
• Calories burned. Estimates that can differ substantially between devices on the same person on the same day.
• Stress scores. Usually derived from heart rate variability. HRV is a real and useful signal, but it’s so individual and so noisy day-to-day that the daily number tells you little you didn’t already know from how you feel.
• Blood oxygen, in healthy people. Genuinely useful in specific clinical contexts. For a healthy adult glancing at it each morning, it mostly generates anxiety.

Run an experiment on yourself

This is where wearables genuinely earn their keep — not as monitors, but as measuring instruments for a deliberate test. Most people never do this, and it’s the whole point.

The method:

1. Get a baseline (7 days). Change nothing. Live normally. Note the weekly average of your three numbers.
2. Change exactly one thing (7 days). One. Not three. Examples: no caffeine after midday. Lights down an hour before bed. Phone charging outside the bedroom. A twenty-minute walk after lunch. In bed thirty minutes earlier.
3. Compare the weekly averages. Not individual days. Averages.
4. Add the subjective column. Each evening, rate your energy out of five in a note. This is not soft data — when it disagrees with the device, trust yourself.
5. Keep what worked. Discard what didn’t. Then test the next thing.

Two months of this and you’ll know more about what actually affects your energy than any article, including this one, could tell you. Everybody responds differently, and the only relevant dataset is yours.

Almost everyone who does this properly finds one surprise — something they were sure didn’t affect them, that clearly does. The afternoon coffee. The 10pm glass of wine. The Sunday lie-in that wrecks Monday.

The unglamorous one nobody mentions

The most clinically useful thing most adults can measure at home costs a fraction of a smartwatch, isn’t connected to anything, and has no app: a blood pressure cuff.

High blood pressure typically has no symptoms at all until it has caused damage. It is common, it is largely manageable, and enormous numbers of people have it without knowing. A validated upper-arm cuff and five minutes a week is a genuinely high-value habit — considerably more so than most of what your watch is showing you.

How to take a reading that means something:

• Sit quietly for five minutes first. No phone.
• Feet flat, back supported, arm resting at roughly heart height.
• Upper arm, not wrist. Wrist devices are far more position-sensitive.
• Take two readings a minute apart and use the second.
• Same time of day, roughly. Morning before coffee is easiest to keep consistent.

Write them down. A month of readings taken to a doctor is worth vastly more than one reading taken in a waiting room while you’re slightly anxious about being in a waiting room.

Using AI to prepare for the doctor

This is one of the most valuable and most misunderstood uses of AI in health. Used one way it’s genuinely excellent. Used another it’s a fast route to unnecessary panic.

The good use: organising, not diagnosing

You get roughly fifteen minutes with a doctor. Most of it gets spent on you trying to remember when the thing started and whether it was worse in the mornings. Preparation makes that quarter-hour dramatically more productive.

I have a doctor's appointment about [symptom]. Here's what
I've noticed: [when it started, how often, what makes it
better or worse, anything that changed around the same time].

Help me organise this into a clear, factual summary I can
read out in two minutes. Then suggest the questions I should
ask so I don't leave and realise I forgot something.

Do not diagnose me or suggest what it might be.

That last line matters. The value is in structure and recall, not in a machine guessing.

The other good use: translating afterwards

Test results, discharge notes, a medication leaflet written for pharmacists. Asking for a plain-English explanation of terminology you were given is a legitimate and genuinely useful thing to do — it helps you understand what you were told, which is your right.

The bad use

Describing symptoms and asking what you have. These tools will produce a confident list, that list will include something frightening, and you will not be able to un-read it. They cannot examine you, cannot see your history, and cannot order a test. The output is pattern-matching, not diagnosis.

Who else can see your body

Health data is among the most sensitive information about you, and it deserves a moment of thought.

• Read what you agreed to. Most wearable companies publish a readable privacy summary. Look specifically for what is shared with third parties and whether anonymised data is sold.
• Be careful with insurance-linked programmes. The discounts can be real. So is the fact that you are handing continuous behavioural data to a company whose business model involves pricing your risk. That may still be a good trade for you — make it deliberately.
• Don’t paste identifiable medical detail into AI tools. Describe symptoms without names, ID numbers or dates of birth. You’ll get identical help.
• Check what your apps share with each other. Health platforms connect enthusiastically. Review which apps have access every few months and revoke the ones you’ve stopped using.

Where to start

If you already own a wearable: turn off every notification except one, look at your three numbers once a week, and run one experiment.

If you don’t own one: you do not need to buy anything to begin. Your phone already counts your movement. A cheap cuff covers the highest-value measurement. And the two changes with the largest effect on how you feel — going to bed earlier and walking more — require no device at all.

The technology is not the point. The technology is a mirror, and mirrors are only useful if you look at them occasionally and then actually do something about what you see.

Start with sleep hours. Weekly average. One week from today, look at the number and be honest about it. Everything else follows from that.