CGM manufacturers publish accuracy data, but what do MARD, RMSE, and the other metrics actually mean for your daily use? This guide cuts through the technical language.
What Is MARD?
MARD stands for Mean Absolute Relative Difference. It measures the average percentage difference between CGM readings and reference blood glucose values (fingerstick or lab draws).
A MARD of 8% means the CGM reading is on average 8% different from the true blood glucose. At 100 mg/dL, that is ±8 mg/dL. At 200 mg/dL, it is ±16 mg/dL.
Current CGM MARD Values (2026)
- FreeStyle Libre 3: 7.8%
- Dexcom G7: 8.2%
- Dexcom Stelo: 8.7%
- Medtronic Guardian 4: 8.7%
Clinically, anything below 10% is considered accurate. Below 9% is excellent. The differences between 7.8% and 8.7% are rarely meaningful in daily practice.
Lag Time
CGMs measure interstitial fluid glucose, not blood glucose directly. Interstitial glucose lags behind blood glucose by 5-15 minutes. This means:
- During rapid drops (like after insulin), CGM may read higher than actual blood glucose
- During rapid rises (after eating), CGM may read lower than actual blood glucose
- Trend arrows help compensate — always consider the direction, not just the number
When to Trust Your CGM vs Fingerstick
Always confirm with a fingerstick before treating low blood sugar, if you feel symptoms that do not match your CGM reading, or before making a large insulin dose decision. Most FDA-approved CGMs are now approved for insulin dosing without confirmation — but clinical judgment always applies.
Calibration
Modern CGMs (G7, Libre 3) do not require calibration. Earlier models (older Dexcom, Libre 1) needed 2 fingersticks per day. Factory calibration has improved dramatically since 2020.
Bottom Line
MARD is a useful benchmark but not the whole story. Real-world accuracy depends on proper insertion site, avoiding pressure on the sensor, and stable temperature conditions. Any major manufacturer’s current-generation CGM will be accurate enough for clinical use.