Disappointed in device discovery

@andy, you’re not giving enough credit for the successes at many homes. You’re truly using your experience in your bubble as THE standard Sense experience. Many people are having very good success with sense, but some of these same successful users had set their expectations too high on how fast learning happens.
And for others listening, Sense has not over-hyped advertising, it’s just the way some users have interpreted their message. The J.D. Power Award winning car of the year has some customers who are disappointed. The early hybrid cars in early 2000’s had some disappointed customers.
ToH was a controlled environment, a demo lab. The same thing all manufacturers do to demo products in a short amount of time to keep audience attention, which ToH had strong control over the flow and usage of their limited video time and product messages.
Finally, I shop at lowes, they have disappointed me in many ways with service and how internet orders happen. I am happier to not go out of my way to beat them up… They have far more happier customers than my experience in my own bubble… I still shop there so I can use their services, because like the aggregated trends in Sense, it still has value while you wait for the product to continue advancement.
… … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … V heart V

1 Like

Not at all, just noting a large group of people that have not had good experience with Sense

1 Like

I have noticed a couple of things similar to your experience. For one thing, the less devices on a branch circuit the more quickly it will identify. The furthest device from my panel (a pump) was identified within weeks, but there is nothing else on that branch circuit.

Devices close to the panel that turn on and off frequently are found pretty quick. However my HVAC unit has so many fans and relays its sometimes hard to ID, when you do they need to be manually grouped together. If you have two (or more) nearly identical HVAC units this is challenging. It cannot differentiate between them. I ended up putting all of those in one giant group. :disappointed:
Another problem I have had is a drill motor in the garage being IDed as the garage door motor. I understand the horsepower rating and motor windings are similar, but one is 30 years old and one is 3 years old. I deleted them and it learned them the same way again. Bummer
Lastly, if it IDs a device like a toaster, or whatever, and then you plug it into another branch circuit… forget about it.
Sense is a cool idea, but I cannot imagine it will ever work 100% for device ID. One day electric devices will send an ID signal… but older devices will never fully work I’M HOME.

What I meant when I mentioned LEDs is that they can be hard to detect. Having LEDs in your home should not affect detection for other devices.

But in short, we hear you and are trying to come up with ways to be clearer about device detection capabilities on our website, blog, and in the press (on my front, I’m active here and on all of our social media platforms, which is giving us much better insight into the expectations and Sense experiences of our users, and how those might differ). It has never been our intention to mislead anyone, and if you feel you have been mislead and would prefer to return your monitor, please get in touch with me. We have honored refunds for customers past the 60 day guarantee period. If you are truly dissatisfied, please get in touch with me via PM. We are committed to making this right and correcting the path from here.

2 Likes

Hi Don,

Don’t be disappointed. A lot of the other answers to you question here are extremely helpful, but on the positive side, my sense detection was quite slow at first also. I am probably about nine months in now and sometimes my detection activity is completely overwhelming. It has figured out almost 50 devices now! I can be frustrating at times, like my sense just discovered my Samsung TV about 4 weeks ago. It even figured out the exact Model number! However, it is connected to a Dish Network Hopper 3 and now I am almost convince it is actually detecting the Hopper and thinking it is a TV? I say that because the Hopper doesn’t turn off for 5 minutes after I shut the power down and it takes about 5 minutes for sense to detect the TV shutdown. It’s fun being a detective!! Enjoy!

1 Like

Please return it, and let us work with people who want to use it, and to help.

1 Like

I agree that device discovery is pretty lame. The device is failing to reliably determine devices. I have studied machine learning for a long time, and I feel the device is over-relying on automation. A guided learning mode would improve device detection considerably, and save many users a large headache. Also – the ability to measure the output of individual circuits would make the given learning problem much easier. The devices learning model is constantly over and underfitting. It’s kind of ridiculous. Not worth the money imho.

2 Likes

If you don’t address either your product failures or your marketing, you won’t have solved anything.

2 Likes

When you say “over-relying on automation”, I think you mean over-relying on unsupervised learning ? Are you suggesting a human guided learning mode or a ground-truth based learning mode to enhance ?

1 Like

I agree Ryan, and I have suggested that Sense’s confidence in their approach to machine learning and device detection appears misguided at this point. I was onboard for fully automated device detection when this was a kickstarter project years ago - but today, now 2-3+(?) years later, I think there are better (accuracy, efficiency) approaches to training the Sense device detection algorithms.
My current configuration (from my previous reset) has been running since Dec 1, 2017 and 88-96% of my usage is ‘other’ or ‘always on.’
No complicated configurations or multiple panels here. Just an 1,800 sq ft single panel home.

3 Likes

Counterpoint

2100sf home, two adults and a child. All run on solar.

3 Likes

Awe, now you’re just showing off, @NJHaley! lol.
MachoDrone’s family doubled to a family of 8.
I noticed our water heater at 147kWh is only twice your usage, so it feels like we’re on par with you somewhere with a quick glance… MachoDrone is being proactive to bring down hot water to about 280kWh/billing period… Going solar after exhausting conservation efforts. I’ll be pinging you on my plans
… … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … V heart V

We typically use about 5kwh on the water heater each day. Two interesting things about that: first, it’s on a timer, an hour on at 5am and 5pm, but that only saves us 5-10kwh a month. Second, we live in a relatively warm environment and the water heater is in the garage, so comparatively we’d use less energy to heat it than anywhere else in the country.

It’s a six month old water heater, as well.

1 Like

Impressive @NJHaley !

I can’t match, given 4 EVs, only one of which gets identified by Sense (A white Tesla Model S), has a plus a much smaller solar array. I believe that another 15-20% of the Unknown usage is from the other EVs.

1 Like

Discovery and labelling sucks. I have motor 1, motor 2, motor 3, motor 4 and so on. Fan 1, fan 2, etc etc etc. Does sense think I have time to rush around my home trying to figure out what each motor or fan is as it come on?

Yup, my pool pump come on all the time but sense hasn’t figured it out yet. It knows when I’m ironing or the fridge light is on though! LOL

Yes @NJHaley, that is very nice detection… here’s how the other half lives:

1 Like

Sure seems like a lot of “always on.” I wonder if there’s a lot of background noise making things difficult. What’s making up that always on, do you suppose?

There are a lot of constants that on on in our house with regularity. Allergy clean air filters @ 300w each, our refrigerator might be in that bucket, though it’s new and shouldn’t be on all the time. Computer monitors, laptops, phone chargers in the walls, TV’s that are ‘off’ but are still ready for a remote to power them on.
I do have network device detection/identification enabled, but it hasn’t helped identify anything on our network except one LG TV. I just deleted that though, since we put an Apple TV on that LG and started streaming through the Apple TV, the LG TV that sense identified became completely unreliable.
We have a fridge, an AC, a furnace, 2 ‘whole house’ fans, the allergy filters, a double oven, a toaster oven, a couple computers, monitors, TVs, gaming systems, a bunch of lights, etc… those are all big consumers of power and are completely undetected.

Yup!