I‘ve read the topic So Long Sense. It’s been… meh by @roger1 . I have seen many other topics that contain the same basic subject as Roger1 complains about, and his started almost two years ago. I have had my sense only three month and I have all the same concerns and problems. I have read all the advice about “be patient”, “the discoveries take time”, Sense will eventually get it correct, and on and on. I believe I read that Sense started in 2015, well this is 2019, four years later.
I was real interested in the ML aspect and wanted to learn and experience the process. Unfortunately it does seem that Sense is somewhat of a beta experiment. I am not a programer or scientist nor do I know design process of a ML device, but I do know when something is basically wrong and a process is not working like it should. When something gets to that point, it is time to re-think and re-assess what you are doing. Something needs to change.
I know people get set in their way of thinking and believe they are right about everything, but when it isn’t quite right…
Things that Roger1 complained about are still there. I have two pool pumps, both 240 volt, two AC compressors at 240 volt. Sense identified the pumps one leg at a time and then I merged them. One I asked support advice about merging one of them and the answer I got was, that’s up to you. Not what I wanted to hear. As of now the pumps some times are recognized, sometimes part of the load is shown and part of it is in “Other”. Then some time while running half the load will just disappear. I also have an AC compressor, that registers 2830w, 1655w, and 1195w, makes it hard to get accurate data on usage.
Don’t get me wrong, I am not abandoning Sense, I still think it is a great concept, but before Sense spends a lot of time and programing effort improving the trends or how I compare to my neighbors, or what my goals are, put the effort into making the program do what it is supposed to do. Show me what, when and how much energy my devices are using. I will figure out how much it cost, when my high useage time is and any other facts I think I need to know. That’s what spreadsheets are for. With the CSV files you provided I can sort, calculate and extract a wealth of information.
If I had not installed an HS110 on each of my air handlers I would never know when and how long they are running. With the help of a VOM, I can calculate how much energy my HVAC is costing by the hour, day, month, etc. Now that is useful information. That is information I should be able to get from Sense alone and not by adding third party devices.
I hope that @RyanAtSense can pass this on to the Powers that be, and maybe it will help to get Sense back on the tract to do what it is designed to do.
I’ve been on both sides of this. One thing you mentioned that I also felt, perfect what your core features are first, before adding things.
At times, Sense looks like an unfinished project. It is like they started it and never quite got it finished. In the construction world, we had a saying, anyone can start a project, it’s the finishing that counts.
I’m with you on getting back to the basics.
I do really like Sense and have not thought about abandoning it, I ha e always felt they try to go in too many directions at once. Where is the captain?
It is really weird and inconsistent. I had what it thought was a coffee maker discovered only to never be seen again. My HVAC was discovered fully and no shows running with half the load under “other”. It’s regressing from it’s early learnings already and that’s disappointing.
There is an inherent impossibility built into any ML system: The Machine won’t be able to Learn unless the system scale is in line with the Subject.
In Werner Herzog’s “The Enigma of Kaspar Hauser”, Kasper emerges from total isolation in a Tower since birth. Standing outside the tower, in Herzog’s fantasy, Kasper tells the Doctor:
K: “Oh, how is it! A very big man must have built it. I would like to meet him.”
D: “A man doesn’t have to be as tall as the tower he builds. He can use a scaffold! I’ll take you to see a new building. You lived in this tower, where that little window is.”
K: “That cannot be! Because the room is only a few steps big.”
D: “I don’t understand.”
K: “Wherever I look in the room … to the right, to the left … frontwards and backwards - there’s only room. But when I look on the tower … and I turn around, the tower is gone! So, the room is bigger than the tower!”.
Sense has emerged from the Tower at this stage but is still learning what that means.
I truly understand the frustration around device detection. Mine isn’t perfect either. But Sense is doing what it was designed to do: provide deep energy insights in an actionable way. Yes, at the core of that is device detection, but Sense goes beyond that. In talking to our users — in reading their reviews, their survey responses, in my constant discussions with people here and in person, in our occasional customer webinars — we’ve learned that the bulk of our users care about saving money on their electric bill over everything else. If we can help with that goal by building out features like Goals, Custom Notifications, Comparison features, more integrations, and so on, we will. In a lot of these cases, these aspects also help with device detection, smart plugs and the recent Home Inventory being the most top-of-mind examples. For me, I’ve probably netted the most savings just by watching the Power Meter, but I’ve heard directly from many of our users how this variety of features, outside of device detection, have helped them save. If you haven’t yet, check out sensesaves.sense.com to see some of these stories.
Also note that this product work doesn’t necessarily come at the expensive of device detection. Indeed, there is some crossover in dev roles at a small company at Sense, but by-and-large, the Data Science team handles improving detection and the software team handles feature addition in the app. In some cases, Data Science labor is needed for new features, but it’s not like they’re entirely dropping device detection projects to assist on a feature like Home Compare.
Also, while you might be willing to put in extra labor with spreadsheets (and, well, probably most of us here in this forum are), we’re not necessarily representative of the bulk of the current or future Sense userbase. Honestly, a lot of us here were probably already digging in pretty deeply to our energy use before Sense. For the other folks, the “newbies,” features like Home Compare or cost tracking really are important and useful towards ends of saving money.
One final point that I’ll rehash: device detection is unique to every home. I happen to have pretty decent detection, albeit missing consumer electronics and some other things here and there (I swear no one on the DS team is pulling levers in the background). Many, many other people are quite satisfied with their level of detection. But, like any ML-based product, the output is pretty heavily dependent on the input. I still can’t get my bedside Alexa to recognize my voice, but my wife has no trouble. Experiences vary on this front for many, many reasons and we try and be as upfront about this as possible before customers purchase.
First Bernie Krause and now Herzog? I feel like we’d get along just fine outside of the Sense world. 
Bringing things back to base, so to speak: Knowing Sense founders have their origins in voice recognition tech I’m amused by your Alexa issue. “Alexa, tell my wife to tell Sense to turn off the lights!”. Hold on, that’s not gonna work. Tech always comes full circle.
My Voice story: New to America in 1994 I proudly purchased a Mac Centris 660AV, which I believe was one of the first “poor-man’s” machines that could do reasonable voice recognition. I set it up: “Computer, Open Window!”. Nothing. “Computer, Close Window!”. Nothing. Local friend walks in: “Computer, Open Window!”. Boom, immediate action. Even with a faux NY accent I couldn’t get it to work. After some fiddle I found that by naming my computer “Pickle” it would respond about 90% of the time. That lead to some pretty odd exchanges when people heard me experimenting but had no clue what I was doing. “What? There’s a pickle?? trying to escape through the window?” [We had a rat issue]
It gets better. Check out our VP of Tech’s bio: Meet George Zavaliagkos: Sense’s New VP of Technology - Sense Blog
While the Bio is a decent, how about the CEO and others higher up the food chain interacting with users in the community?
Are we too far beneath them? Am I the only one that feels ignored by them?
I guess I’m looking at it from my personal perspective that if my dream, my baby had so many people with negative comments, that these comments have come up for years, then I would feel a responsibility to address these users.
A lot of our higher ups and my fellow Sense team members have interacted with our community at various points. We’ve had multiple community webinars, question-answer content pieces (like AskSense), smaller webinars, local meetups, and others members of the team, including our CEO, have written blog articles. That said, we haven’t done a webinar in a bit so should probably do another one soon…
@samwooly1 have you seen AskSense - March 2019 - YouTube at around 6:10. There’s even a beep in there!
@RyanAtSense … something to note regarding AskSense: The CoFounders side-by-side makes me wonder if having a more conversational presentation with engineers would be more revealing (and fun). AT&T’s “More is better” commercials come to mind.
Your “Ask Sense” sessions are great, kind of missing lately.
To see the common complaints addressed would be nice. There are many issues that have been brought up many times Over The life of Sense that still remain. When there is an issue brought up in 2017 that remains seemingly unchanged in 2019, I wonder the overall direction.
Do you have a couple issues that are still outstanding from 2017? Just curious.
No, I sure don’t. I’ve only been a user since January 2019 but I read a lot here. When you go back and read about complaints or problems, that’s how far the same things being said now go back.
Resurrecting an old thread rather than creating a new, same ole same ole post.
So, the original post is from 2019 and now it’s late 2021. The bios are still great and the big picture initiatives are still great and the literary references are still great, but I’m a single user who wants to know more about what’s going on in my home, and now, 5 months after installation and aside from a handful of built-ins, Sense still doesn’t really know squat about my home. Sometimes Sense knows that my AC unit consists of an AC compressor motor, an AC evaporator fan motor, and an AC air handler motor, and other times it only knows that some mysterious, ~4kW, “Other” is running. It doesn’t do at all well with an electric stove or oven, but that’s not a surprise given that the heating elements are all resistive loads that most likely “look” the same. It does consistently see my coffee maker, my AC condensate pump, my clothes washer and dryer, and my microwave. Also, oddly enough, it always recognizes the appliance bulbs in my fridge and freezer. It’s so reliable that I’ve been able to set a 3 minute “on” alert to let me know when one of the the doors has been left open. Aside from those two bulbs, it recognizes no other lamp in my house. Oh, or my ceiling fans. I had imagined a ceiling fan would be a sort of a “gimmie,” given the AC motor signature, but no. A running joke with a friend who also has a Sense unit installed is, “Hey, it’s all ‘Other’ to me!” I’ve installed a couple a of Kasa plug strips and a number of single outlet controllers in order to get a more detailed picture of my home and that info has been useful.
I’ve read the “AI is hard,” “ML is hard,” and, “Hey, devices are not all we do,” explanations and I get it. I do. But I do still wish that Sense was better at being able to do what it’s marketed as being able to do. So, perhaps the issue is the marketing? Does there come a time when you just have to say, “This is the best we can do regarding individual devices,” and leave it at that? Perhaps Sense’s best life is as a big picture, literary reference generating, device/app/service. I’m okay with that. You just need to say it.
Oh, one more question regarding device detection and identification…
Does sense not use the device info that I manually entered at startup as a baseline for detection and identification purposes? If not, why did I waste time entering the info only to be asked to enter it again once Sense identifies one of the entered devices? Again, perhaps Sense is not at its best as a device/app/service?
Just trying to figure out what to expect from here on out.
Two inputs:
-
“Progressive Device Detection” should help with “back to basics”. Watch the Sense announcements.
2021 Data Science Update - YouTube -
The double entry of device data is indeed a pain, but each goes for a different purpose. The entire home list is used by Sense to figure out what devices they have in their user population, so they can target detection more intelligently based on customer need. You didn’t waste your time if you want Sense to think about what you have in your house.
At least it found your dryer.
I have a 240V Dryer that it has never found. Doesn’t even show up in “other” when it is on and in use.
As someone who has supported industrial automation UIs and user experiences, I can confidently state that there’s never a valid reason to ask a client to take on something that “is indeed a pain,” especially something as tedious and automateable (is that a word?) as redundant data entry. When Sense finds a new device, why not launch a dialogue to determine whether or not the device resides in My Home? If it does, then simply copy the details to the new device. If it doesn’t, then enter the new data. Boom. Done.
Thanks again.
Thanks. I’ll watch for it.
There are some valid reasons for the double entry, and the problem is more complex than you make it out to be.
- Historic reason - device name entry occurred earlier than the need for whole home inventory.
- Not 1:1 between home inventory and devices - some home devices consist of several detected Sense devices. The 2 air handler fans for my HVAC are detected by Sense, but are technically part of both my furnace and AC operation.
- Another not 1:1 - some devices that bop between legs of your house supply, like vacuum cleaners, show up as 2 detected devices, but are only one.
- Name management challenges - some devices enter Sense pre-named and categorized - smartplugs, Hue lighting groups, some TVs, anything coming via SenseLink. These may of not match up with the organization that you entered in the home inventory.
Just looking at the problem, I’m guessing Sense would need to invest considerable time in a device inventory management system with manual tuning and error detection if they wanted to handle all the cases. And since that’s all connected to the Power Meter database for each detected device, there are even more complications.
The historic reason seems to me to be nothing more than an excuse; you don’t simply live with limited functionality because a particular routine was there first.
I’ve never been a fan of the “This is hard,” answer. As stated in my first post, if the marketed functionality is too complex to actually implement, then perhaps the issue is the marketing, bringing us back to, “This is the best we can do regarding individual devices."
None of my customers would have accepted any of the offered explanations. They simply required that the particular system do what we said it would do, and if it didn’t, we made it right. Yes, even to the point of investing considerable time. As a Sense customer I think I’ll be happier with the system if I just keep “This is the best we can do regarding individual devices." in the back of my mind and not expect too much of Progressive Device Detection. Then at least I’ll not be frustrated.
Thanks.