With Smart Plugs, what becomes of Always On


I am experiencing the same issues. I added three smart plugs to my heavy always on areas and have seen no change in my always on amounts. It has been 3 days as of right now. I was hoping by adding these smart plugs would have lowered my always on to the correct amount. I sort of like the two bubble idea, having two always on bubbles, one for detected always on and one for undetected always on. The more breakdown of where my power is going the better. I realize this is a new issue with the addition of the plugs so I’m confident they will get it ironed out.


I think simply having a contrasting color bubble for indicating identified always on devices should be simple to implement instead of having 2 always on bubbles. In set up for HS110 identified devices put a selector to indicate that this device is always on.


I don’t want 2 always on bubbles. Why make this so complicated?


@ben, @RyanAtSense,

Here’s one move view on “idle” vs “active” on smartplugs. I’ve been plotting distribution of power over time on all my smartplugs to see if there is a good way to “spot” devices that have a visible non-zero “idle” and an “active” state above that. Since I posted my waveforms above, I’ll share associates distributions for hourly energy on the smarplugs. The width of the area at any given energy is proportional to the number of hourly readings at that energy.

  • The green circles show devices where Sense has found an idle.
  • The red arrows indicate devices and points where an “idle” level seems to be missed.

I’m thinking that Sense should do some kind of statistical test like this to identify one or two modes in the power distribution to compute idles for specific smartplugs loads.


Interesting comments and food for thought.

Smart plus make this more complicated than what has been pointed out already. In my case, before I installed my two smart plugs, I knew most of my Always On causes: my power strip in front of my TV and other entertainment devices, and the power plug strip in front of my computers. There are other always on that I know exists like my DSL modem, night lights, etc, but for this discussion, let’s ignore them.

I was on a mission to cut my Always On, so before smart plugs, I would shut off my two power strips overnight and saw the reduction of the always on number the next day. Wherein lies the problem, but I didn’t catch it until I installed my smart plugs. The Always On number is unlike any other device number. It is an assumption of the power used by devices during the quietest electrical usage in 24 hours. In my case it is in the wee hours of the night. In my home, before I did anything it would run around 180W. I could see one power strip contributed to 68W, the other about 60W. None of my power strips control devices that Sense has identified, so I didn’t encounter the issues discussed above.

Then I installed two smart plugs and the first thing I did was to program them to shut off overnight (after I powered down my PC’s - that’s another issue - having your smart plug shut down power to your PC while it is doing shit you don’t know it does. I did that fat fingering it off by mistake.) The firsts few days I was happy. My always on dropped from 180W to 60W and I didn’t have to do anything to make it happen.

I could actually calculate the ROI of buying the plugs - about 3 months. It was during this calculation that I realized a mistake I was making.

My Always On only drops for 8 of the 24 hours. I didn’t really save 120W over 24 hours, but rather only for 8 hours. By shutting off my power strips, either manually or using a smart plug, I faked out how Sense calculates always on. I was fudging the numbers. I recalculated the ROI and it is still good - around 9 months.

So, now I’m scratching my head wondering just how do I want Sense to handle this? Smart plugs only make it easier for the power strip to be shut off, it didn’t cause the issue of Sense reporting a ‘fake Always On’, I did. (FWIW, my night lights also fake this number out, for they are on only during the quite times at night; they are not really always on.)

This does not mean there are no issues with Always On with the advent of smart plugs? Nope, they are covered in this above posts. But I just wanted to point out, however you great Sense engineers figure out how to fix these issues, don’t forget, if you only sample the quite times, which happens only at night and I up plug phantom devices during night time, I cause Sense to report a number that is not totally accurate. To make it more complicated, today, I programmed the smart plug that runs my entertainment devices to shut off for another 8 hours during the day.

As an ex-engineer, I wouldn’t want to have to figure out the resolution to these issues. Good luck guys!

This is all assuming I have correctly reverse engineered how Sense determines Always On. Wonder if @BradAtSense, would comment?

Thanks folks for the create conversation and analysis. Let me know if I got my head in a place it shouldn’t be.


For some reason I now no longer have an “Other” category. Which seems odd, but ok. I can go with it I guess. Anyone else being this lucky too?


The odd things is though that my “total wattage” used is moving around between 170W and 190W, so either, the devices it detects as “Always On” aren’t actually really always on, or using varying levels of energy I guess, but there shouldn’t be any motors, etc be running right now at home as nobody is there. The most that should be running is all my Google Homes, internet modem, couple of smart hubs (Hue, SmartThings, Harmony) and 2 cable boxes (in standby)


The same thing is happening to me. I’m pretty sure this has to do with the Always-On-being-double-counted issue (see the two other recent threads related to Always On). Because of the Always On double-counting, my bubbles often add up to more (sometimes significantly more) than the total power draw. The Other bubble just shows the difference between the total draw, and the “identified” draw shown by the various bubbles. In our case (because of the double counting) the remaining “Other” amount is actually negative (since the total draw is less than what is shown by the various bubbles), so it shows no Other bubble.


Yeah that would totally make sense (no pun intended :P)


I have seen this, even without double counting. Why ? Always On is a 24 hr kind of measurement. There can be cases where your low watermark from yesterday, is higher than your current “Always On”. In that situation, the 24 hour Always On measurement steals from Other, possibly taking it down below zero. I’ve seen it happen, even before the smart plug integration.


Makes sense. So basically having real time numbers mixed w/ average over 24 hour numbers in the same view is not a great idea :slight_smile: Maybe the “Always On” number should be in a different “shape” (triangle/square?) on the view and it should not be subtracted from anything, but just be there as a “this is what your average over the last 24 hour always on usage was”. That way you can more easily make the case that all the bubbles should add up to your "current’ actual usage and the “Always On” number is just there for reference.


I have periods of time in the mornings where I have no Other use detected.


I like having an Always On calculation because it is based on real measurements, and it’s a worthwhile attempt to pull out the uncategorized Always On usage in your household each day. Maybe it needs to colored a little differently, like Other, since they are really the two different components of the uncategorized power usage.

I would also like Sense to be a little more transparent in how they calculate Always On, since that might help us conceptually with how to deal with it. But it sounds like they are still figuring which flavor of calculation is best. And there are many more possible computations to arrive at Always On than you might imagine.


Add me to the list of having this same issue. Ever since I added the HS110’s to many of my always-on devices, it seems common to no longer see Other


I agree. Always On is misleading. Having it shown to sum up averages is great, but it should not count as part of the wattage calculations. I’d rather see an “other” bubble awaiting detection than an Always on average. I too m running into the double counting issue with Wemo smart plugs.


This right here. My double counting issue has dramatically lowered the “other” category. That being said I know that support is working on it and my double counting has shrunken down dramatically over the last week.


I want to re-iterate here that we’re working on figuring this out, thus we’re refraining from sharing too many details at this time as they’re subject to change.

One of the particular challenges that we’re facing (and working hard to solve) is correlating the usage of the smart plug with the correct leg. Self-reports of this info have proven to be less than accurate, so we’re relying on ML to make those correlations at the moment. Once those correlations are made, your Always On (which calculates per leg) should accurately reflect your smart plug devices. This is not ideal however and can take longer than expected.


You got me thinking again, which is dangerous… I hadn’t thought about the added challenges of managing Always On and smartplugs offsets across both legs of a home’s split-phase delivery, but that is necessary. @oshawapilot and others are bravely experimenting with using HS110s with 240V chargers and other devices that cross both legs. Have your guys thought about that situation as well, where a smart plug splits its load across both legs ??


I can’t resist sharing one more analysis on my smartplugs usage to see if there are visible states, like off, idle, and active in the hourly energy distributions. Sense has far more granularity than hourly readings, but hourly readings still give a good idea of whether there is some clustering of power usage that could be used to define states.

Sense found “idle” in these three distributions:

Washing Machine - pretty clear that there is an “idle” state above 0kWh (Washing Machine uses about 2W when plugged in).

Playroom outlet strip - League of Legends Gaming computer on and off. You can see two clusters here, an “idle” when my kid’s PC is switched off, and another cluster when the PC is on and the GPU is engaged playing LoL.

Family Room outlet strip - AV system on and off - You can see two clusters here, an “idle” with the TV, Yamaha Amp, Apple TV, local switch, and Tivo Mini all in standby (or on in the case of the switch), and another cluster when the AV system and TV are on.

But NOT in these four distributions:

Office outlet strip - removing and adding my laptop

Sonos (2 Connect:AMPs) - blaring the speakers at 50W then back to off. The highest hourly value is 0.024kWh, with the peak around 0.012kWh, but I did take the amps up to 50W to try to force Sense to see an “idle”. Didn’t work.

Toto bidet - the spikes are the spray washer… They go as high as 0.039kWh. The heating element and controller “idle” around 0.013kWh.

Hot water recirc pump with off timer… 0.001kWh when the timer is off (powering just the timer), uptimes to 0.045kWh when the pump is running. Very clear “idle” and active behaviors.


One of the particular challenges that we’re facing (and working hard to solve) is correlating the usage of the smart plug with the correct leg. Self-reports of this info have proven to be less than accurate, so we’re relying on ML to make those correlations at the moment. Once those correlations are made, your Always On (which calculates per leg) should accurately reflect your smart plug devices. This is not ideal however and can take longer than expected.

I have an odd scenario in my house that will likely mess this process up. My energy recovery ventilator is powered using a 240v to 120v transformer, and has a HS110 smartplug between the transformer and the ERV input. So half of that HS110 usage will be put on each phase. I wonder what will happen in this scenario?

I would imagine that the people who are getting creative and using the HS110 to monitor 240v loads will encounter similar issues.


My 2 cents, if I buy a H110S to sense and track a load it should drop off the always on bubble & become a load with it’s own bubble. If I’m savvy enough to buy the individual monitors then I can also look at the trend plot to see the low points and find my always on.