Your own measurements over time

Body-weight scale: Body weight

How often: Same time daily, weekly average is the number you read

Measurement noise, as the record writes it: Day-to-day weight swings of 1 to 2 kg (2 to 4 lb) are normal and come from fluid, food volume and glycogen, not tissue. A single day's reading carries almost no information. The signal is the seven-day average moving over three to four weeks.

Upper-arm blood-pressure monitor: Blood pressure, systolic, Blood pressure, diastolic

How often: Two readings a minute apart, morning and evening, across at least three days

Measurement noise, as the record writes it: Single readings scatter substantially with posture, talking, caffeine, bladder and time of day. Clinical guidance is built on averaged readings across multiple days for exactly this reason. On properly averaged readings, movement under about 5 mmHg systolic should be treated as probably noise.

Hand dynamometer (grip strength): Grip strength, best squeeze

How often: Monthly. Best of three squeezes per hand, seated, same time of day

Measurement noise, as the record writes it: A few kilograms of test-retest variation is normal, so standardize everything you can: same hand, seated, elbow at ninety degrees, same time of day, best of three. A change worth taking seriously is several kilograms sustained across two monthly readings, not one better squeeze.

Sleep and heart-rate wearable: Resting heart rate, seven-day average, Heart-rate variability, seven-day average, Sleep duration, seven-day average

How often: Continuous. Read seven-day rolling averages, never single nights

Measurement noise, as the record writes it: Heart-rate variability swings widely night to night in the same healthy person, so a single night tells you essentially nothing; the seven-day rolling average against your own baseline month is the readable signal, and comparison against another person is not. Sleep-stage classification is the weakest output these devices produce: agreement with polysomnography for individual stages is poor, so treat the REM and deep-sleep breakdown as entertainment rather than data.

Continuous glucose monitor: Average glucose over one sensor's wear

How often: Two to four weeks is usually enough to show a person's patterns

Measurement noise, as the record writes it: Modern sensors run in the region of 8 to 10 percent mean absolute relative difference against laboratory reference, and different sensors on the same person can sit offset from each other. So compare patterns within one sensor's life, not absolute numbers across sensors, and do not treat a reading as a laboratory value.