Total kwh accuracy?


I am curious how accurate senses kilowatt hour detection for the whole house is. One of the several reasons I bought Sense was to figure out why my electric bill was so incredibly high. Now I’ve barely started using Sense three days, but at this moment I’m averaging 80 to 85 kWh a day. If this holds, That’s probably the low-end once it starts to get hot since our air conditioning is electric based.

That would be say 2700 kwh/month at less then $.10 kwh is still $270 a month. I am on an payment plan that estimates my monthly payments. I pay $510/mo all year. Gas heat and hot water. I no the cost will go up come summer but if in general 6 months a year stay at $270/ month, seems like there is a math problem unless my 72degree ac is costing $500/mo on its own. So. I just want to know the kwh accuracy. Thx


I have both a Sense and a Rainforest Eagle, which shows the net meter reading on the electric meter on my computer. I have found both to be very close, though it’s complicated to figure out exact correlation numbers since they both have differing sample rates and latencies - the Sense samples and updates “instantaneous” values more often and with smaller latency than the Rainforest Eagle. I’m not sure if the Sense data is simply sampled or averaged across it’s quarter second updates. I believe the Eagle averages across the what appears to be the meter’s 4 second or so sample period.

Some comparisons this evening…

1,593W (Eagle/Meter) vs. 1,580W (Sense), 792W (Eagle/Meter) vs. 799W (Sense)


Just for the fun of it, here are two different view of my power usage today:

From Sense

From my power meter via Eagle

The Eagle only shows net power, while the Sense shows separate solar input and usage data. But you have to do the integrals if your really want to see if there is a systemic difference between the two, because of the sampling and other measurement differences.


I have a custom emontx based device I built a while back to monitor usage. It matches up nearly identical to Sense. You can check your daily Sense numbers against your midnight to midnight meter readings. I have 2 panels and only one sense monitor at the moment so I can’t compare totals against my meter but I’m sure the numbers would be nearly identical.

Depending on where you are located, this time of the year is pretty mild most places so A/C and heat production should be minimal. Before I got Sense, I went around my house and found all devices that were using lots of energy that I didn’t need running all the time. Typically, this time of the year, I was using 2200 KWh per month. I’m now down to 1400 KWh per month. With Sense, I’m hoping to get a bit lower. I’m now averaging 35 KWh per weekday. Weekends run a bit higher at 45 KWh because my wife and kids are on the plasma TVs all day.


Not sure exactly what you are talking about regarding doing integrals and the sampling time. There is a page on the Sense app that shows total usage for the day, the week, the month and the year. That day goes from mid-night to mid-night. The week, month and year is really only good for the previous week, month and year as it does not include the current day, week and month respectively in the totals.


Howard, good point. Still requires a little wrangling and a spreadsheet since the Eagle is net power and Sense separates solar and full usage. But here’s a view on the last week in kWh. Not sure if there is a way to squeeze more decimal places out of the Sense for daily usage.


More data - The first full day of Rainforest Eagle Usage was March 2nd. Sense appears quite accurate, typically within +/-1% except when the solar contribution becomes a significant % of supplied power and the total usage is relatively small. Looking at the graph for March 12th, the 5% error day, it looks like the infamous “solar dip” plays a role.


Anytime you get agreement within 5% for two sets of non-revenue meters I think you are doing fairly well.


I’m typically within 2 kWh on a given day between what the POCO meter reads vs. what Sense tells me. That works out to a variance of 3% or less. The main reason for my variance is that I have loads in the same sub-panel as my solar panels. So when I use those loads (in my shop) they will drive up the variance as everything in that sub-panel goes through the Sense Solar CTs.

Here are my totals for the last 30 days:


I have tested Sense’s accuracy with a Kill-A-Watt plug and Rainforest Eagle monitor and it’s nearly spot on. Last month my usage was 2kW less than what my utility reported I used so over the course of 30 days that’s pretty dang accurate.


Solar production. Just took reading for June (yea I know it is 7/2) and Sense showed my solar production at 1235 kwh, my inverters claim 1299 and my meter says 1280 if I add in the 37 kwh that Sense says was made yesterday and to right now today the numbers look like this:
Sense - 1272 99.375%
Inverter- 1299 101.484%
Meter - 1280
I would say they are all in perfect agreement within the slight error one would expect.


Bumping this thread, as my observed results seem to deviate a bit from those shared above.

I’ve recently installed the Sense monitor on my mains - and also have two HAN gateways paired to my electric meter providing realtime usage information.

Both the Rainforest Eagle and Universal Devices ISY994z agree with the realtime display on the meter face (not surprising, since this is their data source).

However, the Sense consistently reads 50-60W less - regardless of relative load. Meaning - whether I’m way down at baseline usage of < 500W, or peaking while charging the EV and running the microwave and electric clothes dryer - I still seem to see this ~50W variance - so it’s not ‘proportionate’ to load.

This would lead me to believe there’s some circuit or draw not accounted for - but I’m 100% confident I’m clamped to the mains - the only connection into the house from the meter. There is a sub-panel downstream of the main panel where the Sense is installed - but, as expected, all the load generated there (from a remodeled kitchen) are all reflected on the Sense.

At higher loads, this ‘error’ definitely comes down to a reasonable deviation - but at baseline (~500W) - it represents at least 10% under-reporting of usage.

Is there any way I can further isolate or diagnose the discrepency?

Thanks in advance,


Hmm. That’s pretty curious. I’d reach out to our technical support team: They can take a closer look at the data and may have some ideas on why this is happening.


Thanks, Ben! Request submitted. I failed to mention in my submission that I’ve got photos of the installation, if there’s any question regarding how the CT clamps are oriented / connected, etc. Another thread in these forums seemed to suggest that clamp position / orientation could impact readings - but even that thread only mentioned a variation of 4% based on clamp placement - and I’m seeing 10% variance from what the upstream meter is reporting.

Thanks again!



Quick question - are you looking at instantaneous power (Current Usage) from the Eagle vs. the Power Meter/Bubble power readings ? Or are you looking at energy ? I have both the the Eagle and a Sense, though my setup is a little different since I have solar as well. But the thing that I have discovered is that the Eagle/meter sample rate is much slower, even in the fast poll mode, than the Sense. I haven’t been able to to figure out via Eagle documentation whether the Current Usage number is the highest power point during the entire poll period, or merely a sample at a fixed interval. If it is the highest value during the poll period, then that could explain the differential. My experience is that the energy numbers (integral of power over time) correlate quite closely between the Sense and the Eagle, but looking over time weighs against discrepancies when the power usage is small - relative accuracy at higher power levels swamps poorer accuracy at low power usage.


What is the claimed accuracy of the Sense CT clamps??? A good LEM CT will be accurate to 0.5 percent or better. Can we get the data sheet on the Sense CT?? What company is making the CT clamps for Sense??


Hey Rex,

Unfortunately I don’t have a data sheet on our CTs I can share at the moment (although we are certainly open to it). Our CTs are accurate to 0.5% and likely more accurate than that, as we custom make and calibrate our CTs to reduce inaccuracy. Since you’re interested, and it’s an interesting question, we’re going to write a blog post on the topic with a little more information.


will you give us a nudge when the blog goes up? I’m curious as well


yes I’m interested because my career in the USAF was in Metrology.


Just to pinch in my two cents. My local utility, PG&E, makes total kwh usage available online after a day of two.

I’ve compared sense and the utility’s for a week and it was 99% accurate.

39kwh, vs 39.1 sense reported
85 vs 84.9
And so on