Reading Energy Data Without Misleading the Customer

Why this matters

Customers bring energy graphs to almost every trade now: from the utility's app, a panel monitor, a thermostat's runtime screen. Whatever a tech says while looking at that graph will be repeated, to a spouse, a review site or a lawyer, as "the technician told me". Most bad statements are not lies. They are honest readings of a number that meant something other than what it seemed: a power figure read as energy, a 15-minute average read as a peak, a cold month read as a failing heat pump. This card covers the three readings that go wrong most often and the claims to keep out of your mouth however the graph looks.

Power and energy are different quantities

Power is a rate: how fast energy is being used at a moment, measured in watts or kilowatts (kW). A 4.5 kW water heater element draws 4.5 kW while it is on.

Energy is an amount: power multiplied by time, measured in kilowatt-hours (kWh). That element running for 2 hours uses 4.5 times 2, which is 9 kWh. The utility bills energy.

Graphs mix the two constantly, and the axis label is the only thing that tells you which you are looking at. A live view or a "now" reading is usually power. A daily bar chart is usually energy. A line chart over a day can be either, and frequently it is power averaged over an interval, which is the next trap.

Two rate structures make power matter on a residential bill as well as energy. Time-of-use rates charge more per kWh at certain hours, so the same energy costs different amounts depending on when it is used. Demand charges, still uncommon on residential rates but spreading, bill the highest power drawn in an interval during the month. Before you say anything about what a load costs, find out which rate the customer is on; it is on the bill.

Interval data: averages hide peaks

A smart meter or a monitor does not record every instant. It records the energy used in each interval, commonly 15 minutes, an hour or a day, and a power graph built from it shows the average power across each interval.

That average preserves energy and flattens peaks. Say the 4.5 kW element runs for 6 minutes inside a 15-minute interval. Energy in that interval is 4.5 times 6 divided by 60, which is 0.45 kWh. Spread across the 15 minutes, which is a quarter of an hour, the average power is 0.45 divided by 0.25, which is 1.8 kW. The graph shows 1.8 kW for that interval. On hourly data the same 6 minutes shows as 0.45 kW for the hour.

So the same event appears as a 4.5 kW spike, a 1.8 kW bump or a 0.45 kW ripple depending only on the interval. Two practical consequences:

• Do not read a peak off interval data. The true peak was at least as high as the interval average, and often much higher. To see a real peak you need a monitor that samples fast, or a clamp meter at the load.
• Do not compare graphs with different intervals by eye. An hourly graph from the utility and a one-minute graph from a monitor will look like different houses. Compare energy totals over the same period instead; those agree, give or take each device's accuracy.

Two smaller traps live in the timestamps. An interval may be labeled by its start or its end, so a load that began at 2:00 can appear in the 2:00 bar or the 2:15 bar. And around a daylight saving change, one day has 23 hours and another 25, which shows up as a dip or a bump in a daily chart with no physical cause.

If a customer's monitor and their utility meter disagree, that comparison has its own method; see related: Reading a Connected System's Data Against Your Own Meter. This card does not re-derive it.

Weather normalization: comparing a cold month with a mild one

Heating and cooling use follows the weather, so comparing this January with last January says as much about the weather as about the house. The standard correction uses degree days. For each day, take the average outdoor temperature and subtract it from a base, conventionally 65 F; a day averaging 40 F contributes 65 minus 40, which is 25 heating degree days (HDD). Days at or above the base contribute zero. Cooling degree days work the same way above the base. Monthly totals for most locations are published by NOAA and many utilities print them on the bill.

The 65 F base is a convention about when a typical house starts needing heat, not a law of physics; a tight, well-insulated house starts needing heat at a lower outdoor temperature, so its usage tracks degree days on a lower base. For a rough before-and-after comparison on the same house, the conventional base is good enough, because both periods carry the same assumption.

Normalization also needs the baseload, the energy the house uses regardless of weather: refrigeration, water heating, lighting, electronics. Estimate it from the mild months, when there is little heating or cooling, and subtract it before dividing by degree days, because dividing the whole bill by degree days charges the refrigerator to the weather.

Worked example: "the new heat pump is costing me more"

A shop installed a heat pump in the fall. In February the customer calls with two January bills side by side. Both cover 31 days.

• This January: 1,800 kWh. The bill lists 1,100 HDD.
• Last January: 1,500 kWh. The bill lists 900 HDD.

The raw comparison. 1,800 divided by 1,500 is 1.20, so use is 20 percent higher. That is what the customer sees.

Take out the baseload. The customer's mild-month bills from spring and fall, before and after the install, sit around 600 kWh a month with little heating or cooling, so use 600 kWh as the baseload for both Januaries. That is an estimate, and it is the one input in this example a reader should question first. Heating energy this January is 1,800 minus 600, which is 1,200 kWh. Last January it was 1,500 minus 600, which is 900 kWh.

Divide by the weather. This January: 1,200 kWh over 1,100 HDD is about 1.09 kWh per HDD. Last January: 900 kWh over 900 HDD is 1.00 kWh per HDD. Per degree day of heating, the house used about 9 percent more this January than last.

How much of the increase was weather. At last year's 1.00 kWh per HDD, this January's 1,100 HDD would have needed 1,100 kWh of heating, for a total of 1,100 plus 600, which is 1,700 kWh. The actual bill was 1,800. Of the 300 kWh increase, 1,800 minus 1,500, the weather accounts for 1,700 minus 1,500, which is 200 kWh, about two thirds. The remaining 100 kWh, about a third of the increase, is not explained by weather.

What that means, and what it does not. The comparison is between a new heat pump this year and whatever heated the house last year, so a 9 percent rise per degree day is a real signal worth checking. It is not proof of a fault. Candidates include auxiliary electric heat running more than it should, a lockout or staging setting, defrost behavior in an unusually wet month, a change in how the household sets the thermostat, or the baseload estimate being off by 100 kWh. Each is checkable: the thermostat's runtime history shows auxiliary heat minutes directly. See related: It Worked Until Last Tuesday and Nobody Touched It, which works a case where exactly that turned up.

What changes the answer. If last year's heat was gas, last January's electric bill contained almost no heating at all, and the comparison above is meaningless: the heating moved from the gas bill to the electric bill, and the only fair comparison adds both, converted to the same energy unit. A billing period of 29 days against 33 days needs a per-day correction before anything else. And one month is one sample; a cold snap concentrated in a few days, or a week of guests, moves a single month by more than 9 percent on its own.

What you can say, and what you should not

The negative space matters more than the method here. These are the statements to keep out of your mouth, whatever the graph shows:

• "This appliance is failing." A graph shows a pattern changed. It cannot show why. Say what the pattern is and what you will check.
• "You will save a set percent." A projection from one bill, or a maker's brochure figure, becomes a promise the moment you say it. Give a range, say what it depends on, and say it is an estimate.
• "The new equipment saved you a set amount." Not without weather normalization over at least a full season, and not when the fuel changed.
• "Your bill is wrong." Meter and billing disputes belong to the utility, which has its own process and its own meter test. You can say the readings do not line up and help the customer ask the utility for a meter check.
• "That costs you this much a month." Not until you know whether the rate is flat, tiered or time-of-use. Speak in kWh, and let the customer apply their own rate.
• "That is too much." Too much against what? Name the comparison: last year's same month normalized, the nameplate, a similar house. A number with no comparison is an opinion.

What you can say is the pattern, the comparison you made, its main assumption, and the next check. In practice it sounds like this:

CUSTOMER: So the heat pump is costing me more.
TECH:     Your use is up 20 percent, but this January
          was a lot colder. Adjusted for the weather,
          heating used about 9 percent more per cold
          day than last year. That's worth checking,
          not proof something's wrong.
CUSTOMER: Checking what?
TECH:     How much the backup electric heat ran. The
          thermostat logs it. If that's high, it's
          usually a setting, and I can see it today.

Every statement in that exchange is something the numbers support, and the one assumption that matters, the weather, is said out loud.

Before you say it, three checks

Before any statement about a customer's energy data: confirm whether the graph shows power or energy from its axis; confirm the interval and both periods' lengths; and confirm whether anything other than your equipment changed between the periods compared, including the weather, the fuel, the occupants and the rate. If any of the three is unknown, say so in the same sentence as the claim.

References

• NOAA National Centers for Environmental Information, heating and cooling degree day data
• The customer's utility rate schedule, for flat, tiered, time-of-use and demand provisions
• Manufacturer documentation for the monitor or thermostat, for sampling interval and timestamp convention
• See related: Using a Home Energy Monitor to Find a Problem Load
• See related: Circuit-Level Monitoring and Smart Panels; Reading a Connected System's Data Against Your Own Meter