Vitals — an Apple Health dashboard app that explains what changed
Apple Health quietly collects years of numbers and never tells you what any of them mean. This reads them on the device and surfaces only the changes worth knowing about.
You have the data. You do not have the reading.
Open the Health app and you are looking at charts. Steps, resting heart rate, sleep, heart rate variability — all accurate, all present, and none of them answering the question you actually have, which is whether anything has changed and whether it matters. Doing that yourself means eyeballing a chart and guessing, which is exactly the task humans are worst at: real trends and random noise look identical without a threshold.
Vitals applies the threshold. It compares this week against last week and against a rolling thirty-day baseline for each metric, and it only says something when the change clears a significance bar.
What the Apple Health dashboard app reads
The whole engine runs on the phone and follows one rule: say less, and mean it.
- Eleven Apple Health metrics. Steps, heart rate, sleep, heart rate variability (SDNN), exercise minutes, active energy, walking and running distance, flights climbed, walking speed, weight and mindfulness minutes. Each gets a card on the dashboard with a sparkline and a trend indicator.
- A significance threshold, not a hot take. A change has to clear a relative threshold of 12% and stand outside normal day-to-day variation, measured as a z-score against a rolling 30-day baseline, before it is reported. There are also minimum data requirements per side of a comparison — an insight from two data points is worse than no insight.
- Deliberate subtraction. The engine caps output: at most one insight per metric, and a hard ceiling on the total. A dashboard that flags fifteen things flags nothing, because you stop reading it by the third.
- Recent extremes and streaks. Alongside week-on-week comparison, the engine looks for extremes in the last seven days and for streaks long enough to be worth mentioning.
- An insight timeline and weekly review. Insights accumulate into a timeline you can scroll back through, and a weekly review card summarises the period — which is when most people actually want a read on how they are doing.
- Metric detail with sleep stages. Tapping a card opens a detail view with a trend chart, and sleep is broken down by stage rather than reported as a single duration.
- Ten languages. The interface and the insight wording are localised into English, Chinese, Japanese, Korean, French, German, Spanish, Portuguese, Hindi and Arabic.
No AI service, no upload, no account
The insight engine is explicitly rule-based. There is no model call, no server and no network request in the analysis path — the rules are fixed, inspectable code that runs against data already on your phone. That is a deliberate trade: a language model could write prettier sentences, but it would mean sending your heart rate and sleep history to somebody else's computer.
The wording is observational and non-clinical by design. The app tells you that your resting heart rate is up eight per cent against your baseline; it does not tell you what that means for your health, because that is a question for a doctor and not for an app.
Who it is for
People who wear a watch every day and have accumulated a large amount of health data they have never once interrogated. The value is not in more tracking — the tracking is already happening — it is in someone reading it back.
It also suits anyone making a deliberate change: starting a training block, fixing a sleep schedule, recovering from illness. Those are the periods when a week-on-week comparison against your own baseline is genuinely informative, and when a metric moving in the wrong direction is worth catching early.
Data, privacy and limits
All data comes from Apple Health with your permission, and it is read on the device. Nothing is uploaded, there is no account, and the analysis engine makes no network calls at all.
This is an informational wellness tool. It does not diagnose, treat or monitor any medical condition, and its insights are descriptions of your own data rather than medical guidance.
Frequently asked questions
Where does the data come from? From Apple Health on your device, with your permission. It reads the metrics you allow and analyses them locally.
Does it use AI or send my health data anywhere? No. The insight engine is rule-based code running on the phone — no model call, no server, no upload.
Why does it show so few insights? By design. A change must clear a 12% relative threshold and sit outside your normal 30-day variation before it is reported, and the engine allows at most one insight per metric. Fewer, meaningful statements beat a wall of noise.
Which metrics does it cover? Steps, heart rate, sleep, heart rate variability (SDNN), exercise minutes, active energy, walking and running distance, flights climbed, walking speed, weight and mindfulness minutes.
Is it medical advice? No. It describes changes in your own data in observational, non-clinical language. Anything that worries you belongs with a doctor.