Subject: Feedback on Zone Sense accuracy: Short intervals and Zone 2 threshold shifts

Dieter1960

I’ve been testing Zone Sense recently and wanted to share some detailed feedback regarding its behavior during specific sessions (treadmill and outdoor runs). I compared the Suunto data with Garmin’s LTHR-based zones and performed a deep dive into the raw HRV data.
Testing Conditions:
For these tests, I used a Garmin HRM strap paired simultaneously to both my Suunto Race 2 and my Garmin Epix Pro (Gen 2). This ensures that the heart rate and HRV (R-R intervals) data source is identical and highly accurate for both devices.

Short Intervals Visibility (Treadmill)
During 1-minute intervals at 13 km/h on the treadmill, Zone Sense remains almost “blind” to the intensity shift. Even with the accuracy of the chest strap, the rapid change in effort doesn’t seem to register within the algorithm’s detection window, likely due to the short duration of the burst.
Underestimation of the First Threshold (Aerobic)
During steady Zone 2 sessions (based on Garmin LTHR outdoor baselines), Zone Sense consistently marks me as being above the first threshold (moving into “Yellow/Red” zones in the app).
Data Verification
To verify this, I performed the following analysis:
• Exported the .fit files from both Suunto or Garmin.
• Converted them to .csv via fitfileviewer.com to access raw R-R intervals.
• Analyzed the data through a specialized AI model to identify physiological breakpoints.
The result: The raw R-R intervals confirm that for the Zone 2 session, my heart rate variability markers stayed well within the aerobic zone. I never actually crossed the first threshold, which contradicts the current Zone Sense interpretation. It seems the algorithm is currently too “conservative” for my profile.
Has anyone else noticed this bias regarding the first threshold or a lag in detecting short, high-intensity intervals? I’m happy to share further data if it helps the dev team!

borgelkranz

@Dieter1960 doesn’t the documentation state that ZonseSense cannot detect short intervals at all?

Brad_Olwin

@Dieter1960 If you read the documents and watch the videos, you will gain answers to your questions. ZS is ineffective on short intervals, I personally would not use on any interval less than 20 min or so. There is a lag of 1 to 2 min for ZS to implement changes.
For AT I have tested ZS prior to the public release and since then (for over 1 year) and for me I get both AT and LT accurately but only if I spend significant time in HR zones at or above AT and LT. ZS is primarily intended for live values and its ability to detect zones will depend on how recovered you are. On days I am well recovered for example, after a recovery week, I get AT values near my lab measured values. On days where I am in a training block my AT from ZS can be 20bpm lower than my measured values.
So:

Don’t use 1 or 2 exercises to calculate either AT or LT, a lot are required for me.
ZS is an excellent measure of RPE for me and for some others on the forum.

In my opinion you are asking for ZS to do something it is not intended to do. I doubt the AI analysis is as good as the algorithms (patented) that ZS uses from Monte Cardio (not sure that is correct.)

You can search the forum for a lot of discussion on ZS.

Dieter1960

Thanks for sharing such a detailed feedback. Your experience as a long-term tester brings a lot of clarity to how Zone Sense (ZS) actually behaves in the field.
A few points you mentioned really resonate:
• The Lag & Stability: You’re spot on about the 1-2 min delay. It confirms that ZS is a physiological tracker rather than a real-time reactive tool like pace or power. It’s clearly not built for short bursts.
• Fatigue vs. Lab Values: Your observation about the AT dropping by 20bpm during heavy training blocks is a perfect example of what ZS is meant for. It tracks daily physiological capacity rather than theoretical fitness. It’s more of a ‘form’ gauge than a fixed calibration.
• Consistency: I agree that judging ZS on a single session is a mistake. It takes time and a significant volume of data in the higher zones to see the patterns emerge.
Using it as a dynamic RPE guide instead of a rigid zone system seems to be the most effective way to use the technology. Thanks for the heads-up on the algorithm origins as well !

Dieter1960

Monitara