Defective Monitor


After reporting some data missing from my Unknown Power (a reboot fixed it), I’ve been told my monitor is defective and I essentially have to restart the entire learning process after they send me a new one, all of the data is lost. I am appalled that 4 months in and there is essentially no data backup/restore process. Has anyone else experienced this?


@marcbh, I’m really sorry to hear that. If there is a defective sensor, that means the data from your monitor likely has issues with its data collection. Unfortunately that means that carrying over data in that manner would negatively impact device detection on your new monitor. Completely understand the frustration around this and I certainly wish this was caught sooner!


Ben. Who owns the data collected from the Sense module?
Are there still future plans to make individual"s data available?


@1agkirk2, the short answer is that the user owns the data collected from the Sense monitor, but Sense (the company) has the right to use that data in the ways outlined in the privacy policy (anonymized etc.):

There are still plans to make an individual’s data available to them but I don’t have a specific timeline for that at the moment. It is something that will be tied to the web app which has been stalled, but which work will start up on again very soon.


@BenAtSense - Is there any chance that in the short run, some CSV’s or Spreadsheet files could be emailed to us on a regular basis?


@miracj, unfortunately that would also require some development work on our end and when we do tackle data export, we want to use the time to build the export functionality into the web app. Sorry that development on that front has been slower than expected.


I feel your pain man, I did a reset (Uninformed Engineer at Sense had me do it) and to date the unit has never recovered. Sure its detected a few things, initially 8 then it just stopped. I lost 2 months and 15+ devices… never to get them back apparently. My guess is they host on a 3rd party provider like Amazon and don’t have access to this data in a format they can backup and restore.


Whatever the reason is, it’s poor architecture that a hardware failure can lead to complete data loss. My friend asked me the other day if he should guy Sense, what do you think my response was?

While a detective monitor might not detect devices well, the signatures it learned about from my home and the identication I provided should allow the new monitor to at least be trained much more easily.


@marcbh I think your kinda being unfair to Sense about the data loss. While the situation sucks at the end of the day if you have machine learning data of your house usage, and if that data is defective, you yourself wouldn’t want to keep that dataset. It could easily cause issues going forward with it’s ability to detect correctly. (and potentially other users?)

I do not know the details of what exactly was wrong with your sense and how corrupt your data was, but on a high level with how machine learning works, it sounds like they did the right thing.


I am familiar with the way Machine Learning works. It was identifying many of my devices correctly, but apparently stopped detecting new devices and I also had some missing data. I never said anything about corrupt data. At the very least, the user should be given the choice “Restore Data” or “Start Fresh”.

All of the data my old monitor identified was correct. These signatures should be stored/backup up to the Sense Cloud and re-downloaded to my new monitor and when the new monitor sees IDENTICAL patterns, it should identify these devices correctly immediately. I shouldn’t now have to wait 3 months for my Tesla to be learned again. I spent $300 on a piece of hardware that failed after 4 months, and now have to start all over again. I don’t think it’s unreasonable to expect some sort of data restoration process.


I’m not certain you are entirely correct in your suppositions. If you know a bit about electronics and machine learning, you would know that the new monitor would NOT see IDENTICAL patterns. Maybe close, but the whole idea behind machine learning/CNNs is that no two signatures, even for the same device, are IDENTICAL, especially given the context of the wiring grid of the house plus associated “noise” from all the other devices. And to be clear, two Sense devices probing exactly the same grid using the same clamps will show slight, but discernibly different digital current and voltage results, even if they are only sampling at 8-bit resolution, due to device variation between the two Sense modules.


I’ve told everyone I know, “RUN, do not walk, away from this boondoggle”. It’s to worth the shipping cost.



Think of all of us who don’t have the money for a Tesla (or similar) toy, and also spent the $300. And, after 7 months, my Sense can’t even detect my Sears Dryer.




Interesting… My Sense found our Maytag dryer within 7 days of a reset.


Perhaps I used the wrong wording with “Exact” signatures, but for Sense to be successful, it has to get smart enough to distinguish small variations of devices from home to home (or monitor to monitor), otherwise how will it build a large enough database of detected devices that others can take advantage of? If it’s reliant on EXACT signatures per monitor, I don’t see how this will ever scale and be reliable. If you use Sense for a couple of years and then have a hardware failure and lose EVERYTHING and have to start over, that is just a flawed business model.

I am disappointed in this experience, but I can’t get too upset as this is a brand new technology and this is one of the drawbacks of being on the cutting edge of technology.


I value $300 just as much as the next person. Just because someone has an expensive car, house, etc, it doesn’t mean they don’t have a sense of value. It’s also not a toy, it’s my main mode of transportation that has brought a lot of joy to my family (the kids LOVE it) and it’s not as harmful to the environment like an ICE automobile. Once the Model 3 becomes mainstream, you will see a lot more of them on the road than now.


@marcbh, I feel your pain of having to replace and relearn. And long term, Sense will need to be able to replace a device without losing any detected devices. But given what I imagine their architecture to look like, I can see how it might be extremely difficult to segregate and save a specific users environment.

Good luck and may this round of detections be much faster and more accurate.

closed #18