Real time model vs back dated detection

I’ve noticed in the last when a device is detected the past trends are no where near consistent with the real time detection. This is really evident with my water heater’s device. Real time it gets the on detection correct, but usually turns off prematurely. Now when the device was originally detected the trends were spot on with what I was seeing in the reheem app.

I wanted to see if this was a fluke, so I deleted the device. Interestingly it was redetected a few days later. Again the backdated trends were spot on. Even over the period of time the previous real time device was failing. Again, the new real time device is having the same problems.

Real time device detection is cool, but accurate trends are more important to me. It would be nice if whatever function runs back over our data when a new device is found could be done on a more frequent basis to clean up missed or prematurely terminated real time detections.

I imagine this could be a notification in the timeline every morning or a weekly:
Your water heater trend was corrected by 2.2kWh for yesterday.
Your last EV charge was 4.5kWh higher than originally detected for the previous charge.

2 Likes