Whenever Sense discovers a device, and assuming I’m able to pin it down to a specific… thing, how important is it to enter the make and model number?
I’m only ten days into my Sense experience, but I have isolated and entered make and model numbers for two water heaters, a dehumidifier, refrigerator, and a high efficiency washing machine I just connected to an HS110 smart plug. <<-- Mainly because I’m OCD about this stuff. Incidentally, Sense thinks the washer is on all the time just because the plug is “on” and the washer is drawing 1 watt. I suspect that will change with time.
I initially put in the info for my central air but since it seems to have trouble telling the difference between the upstairs split-system heat pump and the downstairs gas-pack, I removed the details other than “Trane” (which they both are).
Regarding the central air, based on power meter analysis, it no longer seems to recognize the downstairs unit but I’m not sure… The upstairs unit has a soft-start fan and the current signature is definitely different than the downstairs unit. I think maybe Sense has figured it out but since it’s been fairly cool all week and the air hasn’t run I can’t tell.
Anyway, my question is really: Is the make and model really important? Does this info get sent to Sense servers for analysis, which I suspect is true, or is it just for local reference?
My assumption is that the make and model data does go back to Sense and is useful from a “research new features” perspective. We know that data goes back to Sense servers since that data gets populated on other devices running the phone app and web app. I don’t think that specific info is used for training or detection today for a few reasons
there are so many makes and models and still relatively few Sense users, that a large % of specified devices are likely nearly unique (1-3 users). Not so good for training.
Sense hasn’t gotten “militant” about asking for detailed make and model info, nor have they gotten very directive in their guidance on entering makes and model numbers.
crowd sourced classification is only done at the gross device type level today.
But I bet that they do use the data to do research - how does the same exact device look in two different homes ? And I can foresee the day when you might see crowdsourced classification data for the most common devices (is that a Keurig or Nespresso ?).
In that case, I’m glad I removed specific data from my “AC” device other than the brand until it tells the difference between the two units. I’d like to think that I’m at least contributing a little bit to the overall accuracy of device identification.
Inputting make/model isn’t going to have any sort of immediate effect on device detection in your home or others; we’re not doing any sort of “matching” your make/model data to some big database of make/models, for reasons that @kevin1 mentions. In short, it is for research purposes, so we can dig in deeper down the line on specific make/models and improve detection broadly for specific device types. I would encourage users to put it when possible, but if that involves ripping your fridge out of the wall, it’s not necessary
