CGM Glucose Variability Analyzer
Use increasing UTC timestamps and one glucose reading per row.
Your glucose time coverage inputs
The selected range is an explicit comparison, not a personalised treatment target.
Glucose time coverage report
146.67 mg/dL · time-weighted mean
Left-held intervals; final sample has no assumed future duration
Full result table
| Chosen comparison | Covered minutes | Covered time % |
|---|---|---|
| Below | 0 | 0 |
| Within | 10 | 66.666667 |
| Above | 5 | 33.333333 |
The selected range is an explicit comparison, not a personalised treatment target. No insulin adjustment, diagnosis or hypoglycaemia prediction is calculated. Coverage and recording span are reported separately.
Let’s work it out.
Ask about this tool, enter your values, or understand your result.
Ask in your own language.
How to use CGM Glucose Variability Analyzer
- Use increasing UTC timestamps and one glucose reading per row.
- A left-held interval uses its starting measurement until the next timestamp.
- The mean and population standard deviation are weighted by included minutes.
Worked example
The example covers fifteen minutes: three equal intervals at 100, 140 and 200 mg/dL; the mean is 146.6667 mg/dL.
Formula, accuracy & limitations
Treat each reading as constant until the next timestamp. Exclude an entire interval if it exceeds the chosen maximum gap. Compute duration-weighted mean, population SD, CV and below/within/above proportions over covered time only. Last sample has no assumed duration.
The selected range is an explicit comparison, not a personalised treatment target. No insulin adjustment, diagnosis or hypoglycaemia prediction is calculated. Coverage and recording span are reported separately.
The primary result is rounded to two decimal places. Calculation tables show up to four and supporting metrics up to six. Very small nonzero values use scientific notation. Display rounding is not fed back into the arithmetic.
Official source and method context ↗Enter a timestamped series
Use increasing UTC timestamps and one glucose reading per row. All rows must use the selected unit. The final reading closes the preceding interval but is not extended into an unknown future. Duplicate timestamps and reversed dates are rejected rather than sorted silently.
Choose the coverage rule before interpreting percentages
A left-held interval uses its starting measurement until the next timestamp. If that interval exceeds the chosen gap limit, none of its duration contributes to the summary. This is an explicit arithmetic model for irregular samples; it does not recreate the sensor manufacturer’s proprietary report.
Read variability with its denominator
The mean and population standard deviation are weighted by included minutes. CV is standard deviation divided by mean, multiplied by one hundred. Below-range, within-range and above-range minutes share the same covered-time denominator, with both chosen boundaries included in the middle band.
Keep sensor data separate from treatment decisions
Sensor lag, missing data and measurement quality remain outside the arithmetic. The chart is sampled for display when necessary, while every supported interval contributes to calculation. Results describe the supplied series and chosen range; they do not recommend insulin, diagnose diabetes or replace the original clinical sensor report.
Frequently asked questions
Why does the last sample have no time contribution?
Its following timestamp is unknown. Extending it would invent a duration, so it only closes the previous interval.
What happens to a twenty-minute gap with a fifteen-minute limit?
All twenty minutes are excluded from the covered-time denominator. The engine does not count the first fifteen or interpolate missing glucose.
Are the lower and upper boundaries included in the middle band?
A reading exactly at either chosen boundary is within range. Lower and higher readings enter their separate bands.
Will this match my sensor manufacturer’s report?
Different sampling, gap and smoothing rules can give different summaries. This page declares a left-held duration model and supplies no treatment recommendation.
