Three Years of Health Data and Nothing Has Changed
Asier G. Morato
·
·
7
min read
Cover image generated with AI — © Chubby Studio S.L.
KEY TAKEAWAY
Collecting data is not an intervention. Tracking only changes behaviour when a number is tied to one specific decision you agreed to make in advance, so pick one metric, write one if-then rule, and review it weekly. More data does not close the gap between intending to train and actually training.
You have three years of resting heart rate, a sleep chart for every night since 2023, and a VO2 max estimate you check more often than your bank balance. Your training has not changed much. Neither has your sleep.
Why does more data not change anything?
The short version: collecting data is not an intervention. Measuring and changing are two separate activities, and doing the first one well can feel a great deal like doing the second. The evidence on this is unusually blunt, and the fix is smaller than most people expect: one metric, one decision rule, one weekly review.
The most useful study here is one that set out to prove the opposite. The IDEA trial at the University of Pittsburgh put 470 adults through a two-year weight loss programme. Half received the standard behavioural programme. Half received the same programme plus a wearable device that tracked their activity and fed the numbers back to them.
The group with the device lost less weight. At 24 months the standard group was down 5.9 kg and the technology group 3.5 kg, and the authors concluded that adding a device to an intervention that already worked offered no advantage over the intervention alone. The extra information did not produce extra behaviour.
That trial ran on 2010s hardware with a feedback experience nothing like a modern watch, so it is not a verdict on tracking as such. But the mechanism it exposed has not aged. A wearable is excellent at answering "what happened". It cannot answer "what will I do differently on Thursday", and when nobody answers the second question, the first one becomes a very well documented account of standing still.
So when does monitoring actually work?
We know this fairly precisely. Harkin and colleagues pooled 138 experiments covering 19,951 people to test whether prompting someone to monitor their progress helps them reach a goal. It does, with an effect size of d = 0.40. Modest, real, and worth having.
The conditions are the interesting part. The effect was stronger when progress was physically recorded rather than simply noticed, and stronger again when it was reported to someone else. The benefit does not come from the data existing. It comes from confronting it on a schedule, against a goal you actually named.
Automation has quietly removed the part that worked. Your watch records everything without being asked, which strips out the friction and the confrontation in one go. No goal gets set. Nothing gets looked at deliberately. Three years accumulate the way receipts accumulate in a drawer.
What sits between knowing and doing?
A 2025 meta-analysis of 92 studies and 47,548 participants put the pooled correlation between exercise intention and exercise behaviour at r = 0.41. Intending to train is genuinely related to training, but it leaves most of the variation in what people actually do unaccounted for. Seeing that your HRV is down and intending to go easier are both upstream of the only thing that changes an outcome, which is what you do at six in the evening.
That same analysis found the closest predictor of behaviour was not intention, and not even planning, but action control: the ongoing business of checking yourself against a plan and correcting course. Your data is genuinely good at that, provided there is a plan for it to check against. Without one it has nothing to compare you to except yesterday.
Turn your data into one decision a day
FitWoody reads what your Apple Watch already collects and tells you whether today is a push day or a back-off day, judged against your own baseline instead of a generic range. 7-day free trial of Plus.
Download the app
★★★★★
4.5 · 732 ratings on the App Store
How do you pick the one number to watch?
One number, not a dashboard. The constraint is the entire point: choosing one forces you to say in advance what it will change.
Pick it by the decision it feeds, not by how interesting it looks. If you keep training through fatigue and then losing a week to it, a resting heart rate trend is the metric that tends to warn you first, and the decision it feeds is whether tomorrow is easy or hard. If your problem is that you cannot tell a bad night from a bad fortnight, HRV read as a trend rather than as a single morning is the better choice. If you simply do not train often enough, the honest metric is sessions per week, and no physiological number will substitute for it.
Whichever you choose, it has to be judged against your own history. A number held up against a population normal range tells you where you sit among strangers, which is not something you can act on. A number held up against your own last thirty days tells you whether something changed, which is.
How do you turn a number into a decision?
Write the rule before you need it, in the form "if X, then Y". That phrasing is not a stylistic preference. Specific if-then plans are among the better tested tools in behavioural science, and a 2022 meta-analysis of 39 trials found that making a specific plan produced a small but consistent improvement in physical activity, in exactly the population that struggles most to act on good intentions.
Small effects compound when the alternative is nothing at all. Three rules that survive contact with a real week:
If my resting heart rate sits above my baseline two mornings running, then tomorrow's session becomes an easy one rather than a skipped one. If I have not trained by Wednesday, then Thursday is a twenty minute walk and I stop negotiating with myself about it. If my sleep window has drifted by more than an hour, then the thing I fix this week is the wake time, not the bedtime.
Notice what those have in common. Every rule ends in a smaller action, never in a skipped one, which is the whole difference between adjusting and quitting. The temptation with data is to treat a bad number as permission to write off the week, and that is the all or nothing pattern wearing a scientific costume.
What does a useful weekly review look like?
Ten minutes, same day each week. Three questions, in this order: did the number move outside my normal range, did I follow the rule I wrote, and does the rule need changing?
The third question is the one everyone skips. A rule you break every single week is not a discipline problem, it is a badly designed rule, and the correct response is to make it smaller rather than to feel worse about yourself. A rule you have kept for six weeks straight has probably stopped teaching you anything and can be replaced with a harder one.
Write the answers somewhere you will see them again. This is the part Harkin's data actually rewards, and it is the part no app can do on your behalf: the recording is the intervention, not the measuring.
What if the numbers say nothing is wrong?
Then you have learned something worth knowing, and the answer is not to go hunting for a metric that finally shows a problem. Stable baselines with no trend is what healthy looks like on a chart, and searching your data until something looks alarming is how tracking turns into anxiety instead of information.
If the numbers are steady and you still feel flat, the useful move is to stop interrogating the data and change one input for a fortnight. An earlier alarm. One more easy session. One fewer late night. Then look. That way you are running an experiment on yourself rather than reading a report about yourself, and only one of those two produces a different outcome.
Where FitWoody fits
This is the problem FitWoody was built around. It reads what your Apple Watch already collects and turns it into a daily read on whether you are above, below or in your own range, using your HRV, resting heart rate, temperature, blood oxygen and respiratory rate rather than a generic table of normal values. Your daily goal then adapts to that, so a taxed day asks less of you and a recovered one asks more.
The point was never to give you more numbers. It is that the number arrives already attached to a decision, which is the step most people never reach on their own.
Frequently asked questions
How much health data do you actually need before it is useful?
Enough to establish your own baseline, which for most metrics means roughly two to four weeks of consistent wear. After that, more history sharpens the baseline slightly but adds nothing to the decision. Three years of data is not more actionable than three months.
Should I stop tracking if nothing has changed?
Not necessarily, but change what you do with it. Pick one metric, write one if-then rule for it, and review it weekly. If after two months you still take no decision from it, that metric is a hobby rather than a tool, and dropping it costs you nothing.
Can an app make me act on my data?
It can remove the two steps people skip, which are setting a goal and confronting progress on a schedule. It cannot make the decision for you. Trial evidence is clear that adding a device to a plan helps far less than the plan itself does.