I'm done with Sense

I’ve had a Sense installed in my 100% electric, new house for 1.5 years. In that time it’s found devices, lost them, found them again, and lost them again. And…repeat endlessly. It has proven itself to be remarkably unreliable regarding device detection. It finds things that are 100% correct and then loses them within a few weeks. It CANNOT find Mitsubishi Mini-split HVAC units at all, and I have two of them. They represent approximately 75-80% of my total energy consumption, but they can’t detect them. Really? They run nearly every day of the year, but Sense can’t see them? They run on completely different circuits from everything else, but Sense still can’t find them. And machine learning is supposed to be better than a clamp sensor because… what? I know commitment to an idea is a good thing, but there’s been no noticeable improvement in my Sense in 1.5 years. That is NOT an impressive accomplishment.

In Sense’s defense, I will add that it’s been very good at detecting “always on” load. And the solar measurement seems quite stable. But it should since there’s nothing to detect but current flow. The App is pretty nice, but no matter how good it is, the lack of data detection is crippling.

I’ve already chosen a replacement monitoring system. It won’t be using machine learning. Just simple circuit level power flow measurement. The day may come when machine learning trumps all, but as of now that day is still below the horizon.


I agree that Device Detection is very disappointing. In my case it has found small devices, like a kettle and microwave, but doesn’t find all of my biggest energy users, all of which are on dedicated 240V circuits - EV charging, Mitsubishi Mini-split Heat Pump system (1 outdoor unit, 2 indoor heads), hot tub, on-demand water heater. For our clothes dryer it detects certain cycles but not others… yes, device detection is a big let down.

However, since using Sense, and getting solar, I’ve found that the Device page is no longer very interesting to me. I frequently check the power meter and trends pages. It is fun to be able to see how much our solar system is producing when we’re out and about, and the trends help us to predict how many solar credits we are accumulating to then spend in the winter, when our system will produce much less energy.

So in my opinion, Sense is vastly useful and totally worth it if you have solar. It is much better than the SolarEdge monitoring system that I would otherwise rely on, since we can fine tune our energy usage to match how much our PV system is producing in real time. We can glace at the Sense app and know if it is a good time to charge the car, and know if we should charge it at Level 1 (1.5kW) or Level 2 (6.6kW). Without Sense, we wouldn’t be able to have a solar powered car, and that’s something that I’m quite proud of.

My suggestion is that Sense stop marketing towards the Device Detection feature, which is pretty flawed, and instead market towards home/business owners with solar or wind onsite. Device Detection is a minor bonus on top of what is otherwise a great product.

Also - look at other energy monitors on the market. Sense is very competitively priced, even cheap, compared to many other systems. So while I don’t recommend Sense to just anybody, I do recommend it to data nerds with solar.


Im sorry that the Sense monitor has lived up to your expectations.
I’m both very positive and a huge critic of certain things about Sense. I understand where device detection is important when it comes to the biggest energy users in the home and I, unlike you, have the big ones detected.
Those Mitsubishi splits are difficult for sense. The Mitsubishi is a cream of the crop product in mini splits, they just don’t get any better.
While I agree that marketing makes things sound and appear much better than the reality, does the fact that Sense provides your total usage and other information offset its inabilities?
I know it’s definitely more work for you to have to look and breakdown usage patterns where device detection has quite been the best. But does that give you enough insight to make the low $300 cost at least some value?
I can completely sympathize with you that after a year and a half, more hasn’t been detected in a meaningful way. But I also hope you realize that the data you have provided has helped others and will continue to. I came after you did and I’m sure that your data has helped contribute to what I think has been better detection. It is greatly appreciated.

Unfortunately, that’s not uncommon. Variable-speed motors don’t present a pattern for us to lock on to, which makes them incredibly challenging to detect.

One of our data scientists discusses that here: https://blog.sense.com/behind-the-scenes-tracking-air-conditioner/

That is certainly unfortunate to hear. I hope you’ve worked with Support at some point in this time. As I’ve noted elsewhere, the detection experience can be quite dependent on what’s going on in a home. Thus, while you may have not seen much improvement, others certainly have. DS has pushed out many major updates in that time that have resulted in significant improvements both in terms of detection quantity and quality.

Oh…I’ve spent may hours trying to track down Sense’s vague and data-thin snippets of information when it happily announces that “Sense found a new device!” On far too many occasions I’ve figured out precisely what the device was (laser printer, washer motor, septic pump, etc.) and properly identified it in the app, only to discover a few weeks later that Sense was no longer tracking it. A couple of days of data shows up, and then nothing. Gone like the wind. And then— surprise it shows up again…as a NEW device! And then it ghosts me again! This is a significant failure of this product. If machine language can identify the pattern to begin with, it should be able to continue identifying it going forward. The fact that it is so incapable of achieving an acceptable degree of reliability or reproducibility is really troubling. The part that is most outrageous to me is that I can see visually see certain devices appear on the timeline…over and over again. But Sense can’t identify them. As for the parts of Sense that do work, there are lots of energy monitors out there that can achieve the same results, at half the cost. Sense only real advantage over competitors is device detection and individual monitoring. And in my case at least, the device has continued to prove itself completely unreliable. And honestly I think the only real value that my data could provide to other end users would be to throw it out completely. There’s very little in there that anyone could describe as “reliable”. I truly harbor no ill feelings toward Sense. I bought it knowing that it was “speculative”. My feedback is obviously my experience and not necessarily everyone else’s. I have a friend with an older, low electrical use home who also owns a Sense. They have no problems but they also have very few devices, no mini-splits, old appliances and no solar. Perhaps that’s Sense’s sweet spot.

The point you make about reliability between yourself and the friend with Sense is a good one. Consistency between homes is something that is a problem that needs serious work. It is something I feel is lacking in marketing. There is a little mention of about different experiences between homes, just not nearly enough.
The variable speed pumps and compressors are very well known to be “noisy” and present detection difficulties. There should be a list of the these devices that easy easy to find and navigate to. Purchasing and installing the monitor only to find out after the fact that you have poor detection when you could have known upfront is a little unfair.


I think the biggest problem here is that your average sense user is going to be power conscious. We like power saving things. That means variable speed pumps/blowers/motors. They are the most economical. That also means sense can’t detect their power as a constant. If it’s a power conscious device in my house, sense has completely effed up on detection. It only is able to see constant value signatures which is pretty lame over all. We need a better way to mark what’s what. Let me choose what signature is what already. Then let your machine learning working around that.

For me, “device detection” is just a bonus to an already great system. While I do have a 10+ year old zigbee(?) based solar production monitoring system that came with the solar installation, Sense offers much more detailed information on both production and consumption. (Prior to the Sense, I had a TED5000 for many years, it always struggled with powerline communication issues and when the non-replaceable BIOS(?) battery died it was dead in the water.) From what I can tell after eight months of use, Sense is a solid hardware solution that also comes with great software (app&web), still improving with every update. Happy so far, exceeded my expectations. :smiley:

Yes, a circuit level monitoring system would be more accurate, but would also cost more and would be significantly more complicated to install and configure.

A better solution would be for Sense to support more self-monitoring circuits (similar to e.g. Kasa smart plugs) if you are really interested in a specific appliance. Maybe somebody will come out with a supported 240V high current smart adapter that you can place on your big devices.

Let’s not forget that the current technologies of the “smart home” are still immature. Mini-splits of the future are hopefully self-monitoring and WIFI enabled, so sense could add support and just query them. :wink:


I too have the issue of Sense finding devices and then suddenly loses the device. I cycle every device when it is found to ensure that it is the correct device. The same thing has happened with my freezer AND vanity light. They WERE EACH FOUND AND THEN MYSTERIOUSLY LOST BY SENSE. Not what I would expect to happen. There is something bigger going on here. Support is trying to help, but some of there responses don’t make SENSE(pun intended) Here is an example. “Thanks for letting me know. I understand, however, again looking at the signatures and wattage usage difference between what you are saying is the freezer and what this detector was picking up I would be inclined to say that the detector was tracking another device. Due to this it may be best to delete the device to allow Sense to relearn the signature and device behavior to see if it can regain track of the device in the future. Unfortunately, since the device is not being tracked I am not able to force the detection back upon it.” Based on what I mentioned earlier about Cycling every device to ensure that the correct device was found, this makes no SENSE. The Freezer was there for almost 3 months and then GONE. And I validated that it was the Freezer every week while it was found. Please HELP. I want this device to work.

I’ve had my sense for 2 years. During this time I’ve reset the data twice because of the issues you describe above (same devices detected under different names). It just becomes too confusing to spend my time trying to figure out if the device is new or old. The greatest improvement to detection was to use the HST110 sensors. I guess anybody can do that…don’t need a $300 sense. I have a Solar City + Tesla solar/battery setup that tracks the totals…and the electrical utility does that monthly; for me individual detection is THE reason I purchased it. For example I want to know if my well pump is running continuously, if my wine cellar split AC has stopped, if my pool pump is running, when my EV charger is charging my car and for each how much current it’s using. Simply stated Sense can’t do these things by itself. All in, with installs, HST110 purchase (have 10) and electrician costs I’ve invested $750 on this…I would be lying to say I’m satisfied. It’s a learning experience. But honestly, the solutions is a smart circuit breaker…and if I go anywhere…it will be that direction.


Are you aware of the partnership between Schneider Electric and Sense?
Before moving on two another product, maybe there are things coming that will be what your looking for and you’ll already have a piece of the system.
I do t know what is to come but Schneider is a major supplier for electric components and I do t think they would invest large sums of money into something they felt wasn’t top notch.
@RyanAtSense, I couldn’t locate the Schneider link, can you help out please?

I saw that announcement regarding cooperation & investment. Good for Sense. That said, I will go with good old school current measurement as opposed to AI for anything in the future. The best thing Sense ever did was realize ground truth is valuable. I’m learning it is that and more…the most reliable/best way to get the job done.

Sweet, another “let’s bash Sense” thread.

It’s fair to point out shortcomings, with the hope of improvements. I am very patient with Sense on slow device detection. My Chevy Volt is only now after 3 years showing some signs of being detected, though not the full amount of energy. And I do get real value out of tracking changes in Always On, though I wish they were just presented as Watts over time, since it’s already an average value.

But the trust really erodes when Sense 1) loses devices, sometimes rediscovering them later, and 2) confuses devices with each other. The reason is that I have learned from repeated experience that I can’t look to the device page with any confidence. I’ve also learned that the hard work I put into identifying a device can easily vanish into thin air the next day, severely dampening my enthusiasm for this.

How much is my dehumidifier running in the basement during the wet season? Wow it stopped running for a week? Maybe it did, maybe it didn’t. Is it worth getting a toaster to replace the toaster oven? Let’s just check. Hmm, it hasn’t run in two months, even though it gets used daily. A new Refrigerator 6 is running? But I didn’t get a new fridge.

In the end I just spend less and less time engaging with Sense, sort of like that nice and enjoyable person you know who sadly just isn’t that trustworthy.

I only wish there was a way in the interface to let Sense know: “I am really very confident in this device. Please don’t re-weight it out of existence”. Just to let us put a finger on the weighting scale…

1 Like

A post was split to a new topic: Has the underlying Sense hardware changed over time?

I too have demoted Sense from tool to toy. I’m 2 years in and it still hasn’t found my pool filter pump, hot tub, and other large loads I’d like to track. A 2kw radiant floor? it found 1 phase only so it’s reading half. Yet it keeps finding devices that use 0kw at zero cost! Support says to delete them but I don’t bother any more,

I’ve had the same negative experiences with Sense…

2+ years its simply gotten worse and worse in every way. I posted details several months ago.

I have no faith the system will get any better, anytime soon. Sense refuses to find a way to allow users to input into the training models data so it can learn true patterns on devices / combination of devices. It relies on guessing based on “maybe” accurate guessing. And the results show this is true since it misses many devices that are ‘always on’ for years… and when it stops tracking, and recreates devices that already exist (but are no longer tracked).

Sense technology is very early Alpha, not anywhere near ready for commercial use. It is maybe 50% reliable on detection devices - at all, and 25% reliable that it continue to see / track those detected devices longer than a few months at a time. My house devices and usage patterns do not change, at all, so its not that introducing interference would be an issue.

Its a toy, nothing more at this point. I’m sure this type of technology will get better over time but not sure Sense will be the one to bring reliable tech to market. I hope they prove me wrong.

1 Like


That’s the one, thanks for the link @ixu.

I had forgotten the other names involved with the funding because Schneider overshadows them for me. I’ve long been a user of Schneider products.
This was about a year ago, I ha ent heard of anything more recent yet.

It’s a fun exercise, as a thought experiment, to parse the community postings as though they are part of Sense’s ML … and to a greater extent we are responsible for improving (or “disproving”/“disapproving”) the concept. Throwing out a recent post as an example of something “proving”/“approving” to ponder …

If anybody wants a fascinating read regarding, among other things, “empirical temperature” and the history of caloric, I recommend “Thermodynamic Weirdness” by Don S. Lemons. It puts things into perspective in a way that only the history of science seems to be able to achieve.

1 Like