I have two devices that were “discovered” that I can’t match/validate to anything… One is a coffee maker (Sense has already discovered my “real” coffee maker which it accurately detects when it turns on and off)… And one is a “heating device” which seems to consume around 90W when it detects it… Neither has been detected for a while now (a week or more)… I guess this feels like a long time because I have only had Sense installed for a little over a week…
So what exactly happens when I delete a device? Does it send any intelligent information back the Sense servers that the device doesn’t exist? Or is it like throwing a fish that you caught back into the ocean…and it’s possible that you’ll simply re-catch the fish if it detects it again?
Any suggestions or guidance on devices that fall into this category? I know I want to keep these devices and give myself a chance to discover what it actually is…but at some point, do you give up and “toss the fish back into the ocean”?
Hi @MikeekiM - great question.
We actually talk about this a bit in a blog about this very subject (which you can find here). I included a relevant snippet from that blog below that I believe answers your question.
When you ‘delete’ a device, it is the device model that you delete, both from the cloud-based learning algorithms and from the monitor itself. Going forward, Sense will no longer recognize that specific device pattern in your home (though, it may re-detect it.)
Thanks @JustinAtSense… This is exactly what I was looking for…
It is a little confusing what it means that the device is deleted from cloud-based learning and the monitor… But I think I get it… Let me again, use the fish/ocean analogy…
What I am hearing is that this is EXACTLY what is happening… By catching the fish, I have in my possession, a known quantity… Once I throw the fish back in the ocean, I have “deleted” it from my boat…but the fish is still out there in the ocean ready to be caught again if I “detect” it and “catch” it again… But the fact that I have thrown it back in the ocean makes it impossible for me to get it back unless I re-catch it again… (and there’s nothing stopping me from catching it again)…
Sorry… As you can see, I am into analogies…it helps me make sense of the world (did you see what I did there?)
I would be careful with that qualifier. In that blog post, it talks about the system identifying a deleted device as noise and not detecting it again. I am always reluctant to delete a device because of the concern (perhaps unjustified) that a device will never be detected again since the system will put it in the “noise” bucket after deletion. In my experience, once you delete a device twice, it either won’t come back or it won’t come back for a very long time; multiple months potentially.
I haven’t deleted anything for the same reason… I first want to completely understand what’s happening on the backend before doing so… I have the same fear as you… That’s actually why I put the qualifier there…If this is NOT true, I want someone to tell me that I have it wrong…
I want to know whether the back end is merely removing this recognized pattern and device, and putting it “back in the ocean”… Or if they have tuned the ocean to treat this fish as noise and if it is valid, won’t be caught again because I did something that created an automatic filtration within that ocean… I don’t mind throwing the fish back…I do mind committing to treating it as noise and making the ocean intelligent enough to prevent me from catching it again…
Thanks… I think the big subtlety is whether the ocean has changed as a result of me throwing the fish back… Right?
I think both me and @ramon understand that throwing the fish back removes it from our boat…and that getting it back in the boat is impossible once you’ve thrown it back into the ocean unless you actually catch it again… But has the ocean changed in anyway to make it more difficult to catch it again (if it turns out the fish was actually a valid device and was removed by accident)?
I have a twist on the deletion issue. I’ve had my Sense 5 months now. It’s done quite well, even picking up our Prius Prime reliably. We also have mini-split heat pumps, a single condenser driving three air handlers. I know that inverter loads are difficult to identify, so just a week or so ago I added a dedicated sensor to the heat pump circuit. It seems to be working fine.
Now the problem…
We had a regular resistance electric hot water heater which Sense reliably identified. About a week ago I replaced it with a hybrid electric/heat pump water heater. Since then it seems that other previously reliably identified objects have been wacked out, reported unreliably, and the phantom electric hot water heater is sometimes still identified. I assume that the varying inverter load from the new water heater is corrupting a number of other object patterns. I know I can’t teach Sense. I suspect it will help if I delete the old resistance hot water heater. Are there other things I can do to identify the fact that Sense should be trying to fine a “new” object rather than incorrectly apportioning a new pattern among previously existing devices?
Can’t wait to read Data Science team answer. I deleted several times devices in the first weeks after installation. I hope I did not mess with the system by doing so.
@JustinAtSense - Following up with you on this one… Did you ever hear back from the Data Science team?
Also, I had another question for you and the community… At what point does one “give up” on an identified device as “noise” and inaccurate? I have a “Coffee Maker” that was identified as my first detected device (if I recall correctly)… It has NEVER showed up again… At what point do you decide to throw the fish back into the ocean?
I’ve been wondering about this too. I bought a new stand up freezer and my Sense found it a couple of days afterwards. But it hasn’t ever been “on”, it can reliably find the defroster though! Granted my sense has to sift through a LOT of refrigeration in my house.
I get that sense errs on the side of caution and avoids using too much user input to avoid errors. I wish I could see the live data to see what it’s up to.
So on our end, we won’t re-issue the exact same model if a user deletes it. However, we will start training again from scratch the next round, and may re-find that signature and issue a different model for it.
There are quite a few reasons this could be happening, so it’s hard to point to one singular answer as the cause. Considering that it’s been a few weeks, my initial recommendation would be to delete the device and continue with regular usage of your coffee machine so Sense can “try again” with a new model for the device that is hopefully more accurate.
Could be different…but could also be completely identical right? Refinding it exactly as it was discovered the first time is completely possible, right?
I am specifically referring to the possibility that a Sense owner thinks that a device has been incorrectly identified, and deletes it…not realizing that it is actually a valid device…
So he “throws that fish back into the ocean”… But Sense can re-catch it, because it is, in fact, a valid fish/device…
I think the consensus was that we didn’t want to throw fish away that would be no longer catchable. Knowing that Sense isn’t going to give up makes it easier.
Precisely!!! You and I are on the same page… It’s easier to throw the fish back (and I would only do so after thinking long and hard about it, and give myself some time to try and validate it as a valid device)…
But if for some reason, I am 100% confident (but mistakingly so), and decide to delete it…I don’t want to be faced with a situation that is unrecoverable because Sense is interpreting that it should never catch that fish again because, after all, I threw it overboard!
So yes…It is good to know that Sense is able to re-detect it if relevant…
.