Observed last night in the iOS app, the Dehumidifier was running but it’s load appeared under “unknown”, and not under the learned “Dehumidifier” device that has worked fine for some weeks. Confirmed by power cycle. Later on the same evening, the Dehumidifier load appeared under the learned “Dehumidifier” device again.
The introduces the possibility that even after Sense correctly detects a specific device load, it may be an “unreliable reporter” of that source of that load.
Thank you for sending this in. To explain this a bit further, we will only release a model for a device once we are convinced it meets the high bar for quality and accuracy that we maintain. Rarely, however, a device may not perform as well as expected once it is released. This can be caused by a variety of things, including conflicts with other devices and changes in the device’s signature. As our model-training algorithms see more data from the house, these problems should be corrected automatically. Hope this helps clarify a bit!
Yeah, not really. I appreciate that you have a quality threshold in your device detection, but that threshold had been met some time ago when the device was recognized by Sense and renamed by me. What changed? I can understand a problem that means you need to relearn it correctly before you show it to me again, but that isn’t what happened. The load for this hot water heater immediately appeared conflated under “Unnamed motor 6” which previously had a 100w load. You removed a device that appeared to be learned correctly without sharing a reason for doing so, and then put it’s load under an already learned device that’s clearly not correct.
“Change is bad (unless it’s awesome).” This change took something that looked fine and made it worse. I don’t get how this outcome is a win for me, or an engineering success for you.