Way to extend the always-on look back?


Would it be possible to make the always-on evaluation period configurable? More specifically could there be an option to extend it from 2 days to 3 days? As summer approaches I find that my always-on goes up dramatically, specifically from the weekend when my kids have all types of electronics on. Our lowest power usage comes in the daytime during the week when my kids are in preschool and my wife and I work are working. Overnight our undetected pool pump runs which spikes the power usage for that whole period of time, (10 hours). It would be great to extend the evaluation so that always-on doesn’t keep spiking from the weekend.


A foolish question on my part - why do you want to change the definition of Always On to reduce your weekend Always On ? Changing the period and reporting isn’t going save you any power and it seems like you are actually getting a real reflection of your Always On usage that does have a legitimate uptick on weekends.

My weekend always on is showing a false reading. It balloons to 400-500 on the weekend and sits around 200 during the week. This balloons up largely due to undetected devices that run more frequently on the weekend. Always on should be just that always, not a reading that goes up over the weekend and drops the next Monday week after week.

1 Like

I guess I start from the viewpoint that Always On is an indicator, not an absolutely factual summation of all the Always On devices in the house. It can’t be, especially if you have many, many devices in your house and especially if you have things running through every night when most homes are typically at their quietest. I see it as an imperfect but useful measure since the 48 hour window helps get past the cyclic stuff every day.

I don’t see your weekend reading as a false reading - it’s showing you power that has been left on for long periods, though not necessarily truly Always On. But if you have significant steady power use (pool pump) all night, you are going to be susceptible to “false” Always On readings.

One more point - the whole question of detected vs. undetected doesn’t really matter for purposes of Always On. Per the current definition today, Sense uses the 1% bin for each main to derive Always On, not the 1% bin of the mains minus any detected device.

From your last comment, I think I understand your goal - get close to the real, underlying, pure Always On value. And for that, I can see why you might want to change the window.

@senseinaz: Agreed, it would be nice to customize the behavior of “Always On” in several different ways (changing the “look back” period does seem like a useful possibility, and this has definitely been discussed before in other threads, but it’s still helpful to itemize it here).

Sense has clearly worked to improve the “always on” functionality over time, and I think it is much improved overall. But even with these improvements, many people continue to be confused by the behaviors, and the behaviors are still not “one size fits all”. (Note that I include myself as being confused/frustrated at the “always on” quirks at times: and I’m an “advanced” Sense user with several years of experience.)

Unfortunately, I don’t have a specific solution to propose here (that hasn’t already been discussed extensively elsewhere), I just wanted to validate your wish/challenge.

@kevin1, For a lot of us Sense “old-timers”, it can be frustrating to have these same questions asked over and over, but while there are some “routine” explanations… we should keep in mind that having dozens of people struggling with the same thing means that the core features do need some improvement.


I just updated my first reply - I realized that I might have come across as characterizing @senseinaz‘s request as foolish, whereas I meant that for my my follow-on question. Sorry. Agree that core functionality should continue to be enhanced to improve accuracy and usability. Many users, including me, are sometimes surprised by the behavior of Always On, especially because we can’t see the underlying statistics that are used calculate it. But I’m not sure if letting users customize the parameters of a calculation that most don’t really understand, would make the results more predictable.

ps: My goal, typically when asking people about their new wish list requests, is to ask the same question that the guys who would have to specify and implement this new functionality would ask.

1 Like

@kevin1, thanks for the clarifications and additional insight. :smile:

And I certainly sympathize with the challenges Sense faces: it is a very legitimate challenge to balance allowing users to customize or influence settings (usually trying to “compensate” for something that Sense isn’t figuring out or explaining quite right) with trying to have “the system” just figure everything out intuitively (when we all see areas where this is lacking).

LOL: as an engineer myself, sometimes I wish I worked at Sense and could directly contribute to these enhancements… but then I am also highly aware of the complexity and immensity of the data they are working with, and I’m happy to leave it to someone else! Hehe! :wink:

1 Like