When the sensor “lies”: why the CGM reads too high, then too low

The core: blood first, tissue after

Evidence level: established

A CGM measures glucose not in the blood but in the fluid between the cells. Blood sugar changes first, the sensor glucose follows – the lag is, depending on the situation, roughly 2 to 20 minutes.¹ As long as the value barely moves (rate of change under about 1 mg/dl per minute), blood and tissue are practically the same.² Only when it goes fast does the gap become visible.³

You can see exactly this on my beer-garden evening

CGM overshoot and undershoot around a fast change (n=1)
Schematic single-case CGM curve. On a fast change the sensor peaks around 190 (overshoot), then dips to about 78 below the target band (undershoot), before settling around 100. Because the sensor measures in tissue fluid, it lags 2 to 20 minutes behind and overshoots then undershoots on the turnaround.
The swing – and what happens next
Single-case observation (n=1) – schematic from my own CGM curve

Target band 70–160 mg/dl

20015010050

~190
Overshoot

~78
Undershoot

settles (~100)

Why: the sensor measures not in blood but in interstitial fluid. During rapid changes
the sensor lags 2–20 minutes behind – and on the turnaround overshoots first, then undershoots.

Fig. 1: My curve from 16 July, schematic – overshoot to ~190, undershoot to ~78 below the target band, then settling around ~100.

After the unexpected sugar surge (the detective work is here) my sensor showed 205 mg/dl at 19:42 with a steeply rising arrow. Later in the day – in hindsight, with a smoothed curve – the peak reaches about 190. Then the curve falls steeply, shoots below the target band to about 78 mg/dl, and afterwards settles around 100, where it stays calm for hours.

This is exactly the pattern that unsettles many: the sensor overshoots on a fast rise and undershoots on a fast fall, before it catches itself. That’s not a malfunction – it’s the physics of the tissue lag plus the smoothing algorithms in the device.

When the deviation is largest

Typical situations with fast changes – and therefore the largest deviation:¹

After eating, especially with fast carbs (my beer-garden evening)
During and right after exercise
After treating a hypo – the sensor still reads low although the blood is already rising
After medications that move blood sugar fast

The other gap: the sensor change

On 13 July my notes simply say: “gap due to sensor change”. That’s the mundane but important kind of gap: between the old and new sensor there’s simply no data. On top of that, a fresh sensor is often not yet up to operating temperature in the first hours – the values can deviate more at the start.

How I handle it: mark the gap as a note. Otherwise weeks later I puzzle over why there’s nothing there – or worse: I interpret a gap as an event.

Key takeaways

Trend beats single value. With a steep arrow the direction matters more than the exact number.
Don’t react to the swing, but to the pattern. The undershoot to 78 was no cause for panic – the curve caught itself.
If feeling and display conflict: finger-prick. If symptoms don’t match the sensor value or it’s implausible, use a fingertip measurement per the manufacturer.
Note gaps so the later analysis stays honest.
In calm phases the sensor is most accurate – that’s where absolute numbers are most worth looking at.
CGM measures in tissue fluid, not blood – lag typically 2–20 minutes.
With calm values, blood and tissue practically agree; with speed the gap opens.

Sources

Sources as of: 21 Jul 2026.

CGM tissue lag (~2–20 min) and largest deviation during fast changes. Reference. pmc.ncbi.nlm.nih.gov
Blood and interstitial glucose agree at low rates of change (<~1 mg/dl/min). Reference. pmc.ncbi.nlm.nih.gov
At rates of change of −1 to 3 mg/dl/min, MARD is about 9.4–11.4%; above 3 mg/dl/min it rises to 19.0% (flash glucose monitoring, meal test in type 2 diabetes). PMC. ncbi.nlm.nih.gov/PMC7199533