My Sense experience and observations

I have had it installed for about 2 weeks now, with the Solar option. It was relatively easy to install and configure.

It has identified 7 devices accurately and 3 incorrectly (which I deleted or renamed).

As I have had no technical issues I cannot comment on the Support aspect of the company. Below are my observations / comments:

  • I am doubtful it will get much better in resolving more devices in my home at this time, but their model is predicated on building a database from experiential data. So over time and with user cooperation it could conceivably get more accurate.

  • They need to develop a broader user inclusive means for device detection. I can turn a device on and off and see the change in the consumption at a very fine granularity. I should be able to tell the system it is device XYZ. Unfortunately you cant today.

  • The Solar sensing is very accurate. I was able to compare it to my inverter readings and was getting a ~5 Watt delta.

  • I am most concerned that if the company goes out of business or is acquired, that the device will essentially be rendered useless, or a new business model charging for the web based data base / monitoring will be enacted.

  • It is a useful device for a home energy audit. I have detected for example my water heater is probably the worst electricity consumer and due to its age would be the best appliance to upgrade at this time.
    After the audit, I cant say it is probably as useful. I think they should change their business model to some degree to go after electricians or power companies to provide customers with an audit service. (They need to grandfather those of us that bought it however)

1 Like

Thanks for writing in about your experience. Let me respond to a couple of your points in turn.

Why don’t you expect it to find more devices? 2 weeks is a pretty short period of time in Sense land (We advise reviewers to use Sense for 60 days before they evaluate). I’d definitely expect you to have more devices come online over the next few weeks.

We hear you. We’re working on other ways to get ground truth data into Sense (NDI helps with that, as do smart home integrations, and we’ll be announcing a big one soon). A straight up “learn this” mode just isn’t feasible at this time. You can read more about why here: Why can't you train Sense?

This is a fair point and something we’ve thought deeply about. We’re doing our best to create systems that would still give Sense functionality in the event that we have to close our doors. But, we’re very healthy and just got another big dose of funding, so we’re not too worried about that at the moment.

Hi Ryan

Thanks for your responses…I will try to respond to your questions

  1. Additional Device detection - I base this on typical learning algorithm behavior.

It finds devices based on signal signatures it knows and has reinforced with correlation from many users. So early on (ie 2 weeks) it has detected the ones it knows best from this accumulated knowledge. The rest go into the “Other” bucket or perhaps “Always on”. I havent seen any of the “always on” class get detected.

Once it goes through those I guess it keep analyzing until it thinks it has a possible hit. In my case the first 7 or so were both quickly detected and correctly identified. Since then it has detected some discernible signatures, but they were not correct or there was insufficient trend information to accurately identify the signal.

So basically it gets as many as it is sure of (clear signature) relatively quickly, then it takes time before it tries to identify the less clear ones. Here without a larger database of signatures it cant resolve the signal as accurately or quickly., Hence my comment regarding experiential feedback from users. Devices like electric dryers or washers I would have thought were easy targets, but it has yet to even identify these in my home. So I am presuming it is lack of user data from which it can learn and identify.

  1. User assisted detection

Sure any smart device helps identify its presence, but doesn’t necessarily mean you can correlate its existence with its consumption. I can tell Sense all of the devices that exist in my house (or at least most of them) without NDI, and as I mentioned above I could associate a change in the consumption with a specific device. But you are saying today it isnt feasible. I wonder then how you make the connection with NDI between existence and consumption. I dont see these 2 use cases as different problems.

  1. Company viability

Glad to hear. Mange the funds wisely :wink:

  1. Business Model

I note you didn’t respond to this…seems like it might help in finding stable long term markets.

On your points 3 and 4, you should take a look at Sense’s recent B round funding announcement and find some solace there… They have some strong new funding, plus one of the participants in the B round is a major power equipment manufacturer.

1 Like

@jdlow1959 My experience with device detection… See my graph. Don’t expect that many of your devices will be immediately found. It may take months. Also use the ‘suggested devices’ as a guideline. You really need to personally figure out what the device is. The suggestion list is just that - a list of possible devices. There will be times when SENSE will discover the same device twice. Or you may have a scenario where it will detect multiple signatures on the same device. Example: Washing machine - wash cycle, drain cycle, spin cycle. You can decide to leave them as separate signatures devices or merge them into a single device. Good Luck!



Sense’s learning does differ a bit in that it’s not entirely top down. Model building happens between the Sense monitor and your home, so it’s not a problem of just trying to match up snapshots from your home with snapshots in our database. While detection does usually slow down at some point, it’s usually after that two months or so, but this really does differ among homes.

As for the NDI question — good point! There is a difference between relying on users to manually ID and relying on, say, Philips Hue, to provide us with temporal links. But really, the best NDI data comes from devices that also accurately report wattage, like smart plugs. I highly encourage you to be watching the forum over the next week to learn more about that :wink:

As for the business model question, I did not purposely skip it! I think I just misread it. But we do have a Sense Pro program that targets electricians and homebuilders and we have worked with utilities ( More of that is definitely in the works.

Manage the funds wisely… Guess we’ll have to tear down the indoor basketball court we just built… Can we keep the Olive Garden franchise we bought?

Welcome to the forum btw! It’s awesome having such engagement right up front. :+1:

Hi Ryan

Thanks for the additional info…

As for the funds, keep the court, loose the garden :wink:

1 Like