Better Insights into "Unknown" Usage (particularly to Guide DCM and Smart Outlet Placement)

The “Other” category seems to be the bane of any Sense user’s experience – it’s offers no insight into what devices are consuming power and therefore no value towards more intelligent energy usage. In a typical month, my “Other” hangs around 30% of my usage – usually more than any single device. As I assume is the case for most enthusiastic Sense users, my ideal case would be to see my “Other” eliminated completely and my overall consumption accurately allocated to each device that contributed. Despite their impressive work thus far, I understand that Sense’s native detection will likely never be able to 100% separate and allocate total usage across all devices on it’s own – which is where smart outlets and DCM become necessary. But as of now, Sense users may not necessarily be getting, or know how to get, the maximum benefit from add-ons like DCM and Smart Outlets / Power Strips. I’m currently experiencing this with a KP-115 plug and HS-300 strip I have on hand, as well as the DCM I would like to add in the near future.

My proposal is for better insights into “Other” usage to help a user to better allocate DCM and smart outlets to where they would offer the most benefit. While what is considered “most beneficial” may vary between users, I feel most would want to reduce the amount of usage being attributed to “Other” and get better insight on what devices are contributing to “Other” (even if they are never recognized in the device list).

One of the easiest ways to provide such insight would be to enable the Power Meter for “Other” as originally suggested here, which gives a basic idea of what devices are contributing to “Other”. This way we could estimate what the biggest contributors to “Other” are without having to sort through the noise from known devices.
I think the enhanced proposal I posted (post #15) under this previous thread may offer even more value. A way to “Filter” devices from the main Power Meter (I’m picturing a dropdown checkbox list of known devices) would offer value in many ways by making it easier to see how each device contributes to overall usage. Similar to this suggestion. In the context of this enhancement suggestion, it would allow the user to visualize “Other” usage against known device(s) usage on the same timescale. We could then see if changes in “Other” usage coincide with events from known devices to better assess where add-ons would provide the most benefit. Example: a spike in “Other” usage at the same time as A/C compressor start-up may suggest HVAC fan is a main contributor to “Other” (assuming HVAC fan isn’t already a known device). This may also suggest that Sense’s learned power usage waveform for the A/C compressor isn’t very accurate, and that the A/C compressor may be a great candidate for DCM.

The overall point or “problem to be solved” is this… I’m a Sense user holding a pair of flex sensors or a smart outlet/strip (that I plan on using exclusively for measuring power consumption, not remote on/off control). How do I know where these add-ons would be best employed? Currently I find this very difficult to determine.

Given how easy it would have been to have include when the individual power meter feature was added for the know devices, I imagine there is a deliberate reason the Sense team didn’t enable this for “Other” as well (@JustinAtSense, any insight here?). It may have simply been that it felt untidy to include “Other” given that it’s a random recording of potentially many devices + the error from any known devices. In other words, it’s the remainder after “Always On” and all known devices have been subtracted out. There’s really know way of controlling how useful this is or limiting the difficulty in interpreting it for each user. For the reasons argued above, I think there is a lot of value in providing this to the user – even if some dialog is required along side it to ensure users don’t lose sight of what it is their looking at.

The reason may also be out of fear of revealing the inherent inaccuracies or “how much is missed” from the waveforms of known devices learned through native detection. This may be a valid concern, especially given that it’s easy for most users to expect perfection in terms of device detection without having any first-hand understanding of the underlying challenges associated with it. If this happened to play any bearing on the decision, I’d add that I personally don’t believe it’s fair for any user to expect the detection algorithms to be perfect. As a consumer I chose Sense because I believed they had the best ability to detect individual devices from the main power lines. I don’t expect them to be perfect, but only to remain superior to my alternate offerings from their competition.

One final suggestion for this feature enhancement… Maybe Sense could even use some of the insight gathered across the user base to make further suggestions on where add-on detection methods would be best employed. Rough example: provide each user a list of how the calculated usage of some of their more major/common devices compares to other users… if the ‘per cycle’ or ‘per month’ power usage of my clothes dryer uses 1/10th the average of other users, and clothes dryers are typically one of the highest usage devices for Sense users, then I may want to consider DCM for my dryer. Sounds crude but would need the Sense team to add their flare to it.

I agree it would be great having the ability to compare usage to other users at a device level. However this would be quite problematic.
Quite often only parts of devices are discovered. For example one users “AC” might be the compressor, another the compressor and blower. One users furnace might be the induction motor, another’s the induction motor and blower motor. A washing machine could be the entire device or only portions of cycles. Some users report only cycles of specific wash settings are detected.

Sense detected the small burners of my electric range independently from the large burners. Sense thinks I have two completely different stove tops. I like it that way, many others merge the two into one device.

These detection and labeling differences of similar devices of the same type would make comparison highly inaccurate and certainly frustrate a lot of users. Everyone would think they are comparing apples to apples when in reality they would be comparing ones apple pie to another’s apple sauce.
If Sense ever got to the point of being able to compare apples to apples that were picked from the same tree things could get very exciting.

I too would like a power meter for “other”. I have not attempted to search if Sense ever explained the reasoning for not including it. I agree with most everything in that thread, especially fond of the filter suggestion you mentioned in the first thread you linked.
Theoretically a “power meter” rather the signature an individual device could simply be subtracted from the overall “power meter” leaving the potential to display only “other” and “always on”.
Going full circle this might explain why there is no distinct “other” power meter as it would have to be a combination of “other” and “always on” to be accurate. “Always on” is only an estimate. Basically an average of what’s left over when you take out all the peaks and valleys long term. Average changes over time which would make “other” confusing to look at on a detailed graph like the power meter if it was separated from “average”.
In a way this gives more weight to your idea of being able to subtract specific devices.
Circling around again, if you were to look back an “other” it would also have to include any device that had not been discovered at that point in time, thus adding even more confusion.

I could very well be thinking this all wrong as It’s been long overnight with way too much bad coffee.

I think you have hit upon why Sense doesn’t compute a device Power Meter for Always On or Other - neither of those are based on direct measurements. Always On is a rolling statistical calculation (1% bin of data for the last 48 hours), and Other is a rolling subtraction of all identified devices plus Always On from Total Usage.

@matthew_lasorsa, I appreciate your thoughts on how to squeeze out more insight into the components of Other. One route I have found useful is simply trying a Traveller HS110/KP115 against some of the suspected bigger consumer that haven’t yet been detected. If you leave the Traveller on each device for 48 hours or so, you should acquire enough of a waveform history for Sense to provide a good view of average and peak usage, as well as a Always On number for just that device. For undetected wired devices and 240V devices, you could do the same with DCM, though that would require regularly moving around CTs in your breaker box(es).

One of the the reasons I posted some of my smartplug waveforms in the Sense Community Device Library was to archive the usage waveforms of some of the non-detected devices, so I would have a history of what I tried.

It is broken down as an estimate, but the main “meter” “power meter” is accurate. Theoretically it would be possible to subtract a single devices “power meter” from the overall “power meter” allowing for a “power meter” or only other and always on checkboxes on the main “power meter”. But I imagine something like that would eat up a lot of extra computing resources, cause additional confusion as the total on the graph would not match up to the bar graph.

I suppose another and likely better option would be a a colored overlay showing individual devices on the main graph.
Main power meter graph would remain orange but when a known device is detected it’s signature represented on the meter would be it’s own color.

I’m sure someone could easily overlay something like this as a 3rd party addon type of thing with less accuracy but that would require having the ability to export the graphs themselves.

Either way I imagine it being quite buggy and require a lot of dedication. Maybe something to look into (long) down the road?

A big problem with DCM is remote sub panels. The breaker that power my HVAC system is in a sub panel in my basement. 75% of my average usage is likely feed off that panel. My main panel is on the far wall of my garage, which also feeds another sub panel located centrally on my main floor. I have no real practical way to monitor my HVAC until all components are detected naturally which is somewhat of a problem itself.
My furnace induction motor has been detected. The blower motor is lumped into “other” at initial start up and for a few seconds after then falls into the bubble with the induction motor. When I use ventilation/fan mode without heat the blower motor stays in “other”. This in itself is another reason why an “other” power meter could be really confusing and somewhat misleading.
If the main meter was color coded it would make it a lot easier to pinpoint exactly what is going on. As I would see a large spike on the graph every single time right before the furnace color would jump. This could be a really useful feature that could be included if Sense were to ever consider implementing some type of advanced mode.