The Well Pump That Ran Up the Electric Bill

Why this matters

A rural house with a private well has a load most high-bill checklists never mention: a pump that runs whenever the pressure tank calls for water, whether anyone is using water or not. When that pump runs more than it should, the bill climbs slowly enough that nobody connects it to anything.

This case is worth following because the first diagnosis was right, the repair was right, the customer was satisfied, and the bill did not move. The reason is a piece of physics that is easy to state and easy to miss on a service call: how often a well pump STARTS and how much energy it USES are set by two different things.

The signal

The customer called in late September. Nothing was broken. The electric bill had been creeping up all summer, and her own comparison against last year's account history showed it, month by month, in use per day:

A single jump points to something that switched on. A difference that grows every month, through summer weather that went up and then came back down, points to something that is getting worse. All reads were actual.

The office pulled the account's 15-minute interval data from the utility's online usage portal, with the customer on the line (see related: How to Get a Customer's Interval Usage Data Without Their Password). The overnight floor, the lowest the house drew while everyone slept, had been about 0.10 kWh per 15 minutes last September: a refrigerator, a freezer, standby loads. This September it was 0.13, and it was 0.13 in every single interval, all night, with no quiet ones.

That pattern says something specific. A load that cycles slower than the interval would show up as some intervals high and some normal. A load that never stops would show as a flat step. A load that cycles every few minutes averages out inside each 15-minute window and shows up as exactly this: an evenly raised floor. Fifteen-minute data cannot see the cycle; it can only say the cycle is faster than fifteen minutes.

The customer, asked to stand in the basement that night with the house quiet, heard it: the pressure switch on the well tank clicking on and off every few minutes.

Visit one: the right diagnosis

Before anything was opened. The pressure switch carries line voltage on open terminals under its cover, and the tank and piping hold stored water pressure. The tech left the switch cover on for everything except the final repair, and for that repair switched off the pump breaker, proved the switch terminals dead with a meter checked on a known live source before and after, per NFPA 70E Article 120 (120.5 in the 2021 edition, renumbered in 2024), with work practices at 29 CFR 1910.333(b)(2), and drained the system pressure through a faucet to zero on the gauge before disconnecting the tank. In some states, work on a well pump or its pressure system requires a licensed well or pump installer; the shop that took this call holds the license its state requires.

The measurements. With the house quiet and the switch cover on, the gauge rose from 40 to 60 psi (the switch's cut-in and cut-out settings) in about 22 seconds, then fell back to 40 in under three minutes. A full cycle about every 3 minutes. The tech measured the pump's draw with a power clamp reading true watts at the pump circuit in the panel: 1.0 kW while running.

The drawdown test. Drawdown is how much water the tank delivers between cut-out and cut-in. With the pump breaker off, the tech drained from 60 psi down to 40 into a bucket: 1.6 gallons. The air valve on top of the tank spat water when pressed. The tank had lost almost all of its air charge and was nearly full of water, which is what waterlogged means. Without an air cushion to push water out, the tank stores almost nothing, so the pump restarts after every small draw.

The diagnosis was correct. The tank was replaced, its precharge set with the tank empty of water to the value the tank manufacturer specifies (commonly 2 psi below cut-in, so 38 psi on this 40/60 switch), and the system restored: breaker closed with the switch cover on, the tech standing clear of the switch, the gauge watched through a full cycle to confirm the pump cut out at 60 and back in at 40. The new tank's drawdown, measured the same way: 9.5 gallons.

The clicking stopped. The customer was pleased.

The bill after the fix

The October bill, with the new tank installed on day two of the period:

The difference had gone from +2.9 to +2.8. Essentially nothing. The customer called back, reasonably: the shop had fixed the problem and the bill had not noticed.

The interval data made the puzzle sharper, because it HAD changed:

15-MIN INTERVALS, kWh, house asleep
           BEFORE TANK (Sep 28)   AFTER TANK (Oct 9)
  2:00 AM        0.130                 0.136
  2:15 AM        0.130                 0.100
  2:30 AM        0.130                 0.136
  2:45 AM        0.130                 0.136
  3:00 AM        0.130                 0.136
  3:15 AM        0.130                 0.136
  2:00-3:30      0.780                 0.780
  12-5 AM        2.60                  2.60
  12-5 AM, same night last Sep:       2.00

Before the tank, every interval carried the same small bump. After it, most intervals carried a bigger bump and one in six carried none. The shape changed completely. The totals did not move at all.

Why being right was not enough

A pump on a pressure tank does two things, and they are governed by different numbers.

How often it starts is set by the tank: how many gallons it hands out before the pressure falls to cut-in. Small drawdown, many starts.

How much energy it uses is set by how many gallons it has to pump, divided by how fast it pumps them, times its running watts. The tank does not change that at all. If the house is drawing water, the pump has to replace it, whether it does so in many short runs or a few long ones.

So the question visit one never asked was the one the overnight data had been asking all along: what was drawing water at 3 AM? A waterlogged tank explains fast cycling WHEN water is being used, because any small draw empties a tank that stores 1.6 gallons. It cannot explain cycling when nothing is open. Something had to be taking water out.

Visit two measured it. With every fixture off and the new tank's 9.5-gallon drawdown, the gauge took 15.8 minutes to fall from 60 to 40: water was leaving the system at 9.5 / 15.8 = 0.60 gallons a minute. Refilling from 40 to 60 took 2.16 minutes, during which the pump replaced the 9.5 gallons plus the 0.60 x 2.16 = 1.3 gallons that leaked meanwhile, so the pump delivers (9.5 + 1.3) / 2.16 = 5.0 gallons a minute at this pressure.

Now both visits reconcile with one calculation:

• The leak takes 0.60 x 1,440 minutes = 864 gallons a day.
• At 5.0 gallons a minute, replacing it takes 864 / 5.0 = 172.8 minutes, 2.88 hours, of pump running a day.
• At 1.0 kW, that is 2.88 kWh a day.

That is the +2.9 on the September bill and the +2.8 on October's, within what monthly bills can resolve. And it is the same before and after the tank:

• Old tank, 1.6 gallons drawdown: off for 1.6 / 0.60 = 2.67 minutes while the leak drains it, on for 1.6 / (5.0 minus 0.60) = 0.36 minutes, about 22 seconds, to refill against the leak. A cycle every 3.0 minutes, about 475 starts a day, 475 x 0.36 = 173 minutes of running.
• New tank, 9.5 gallons drawdown: off for 15.8 minutes, on for 9.5 / 4.4 = 2.16 minutes. A cycle every 18.0 minutes, about 80 starts a day, 80 x 2.16 = 173 minutes of running.

The tank replacement cut the pump's starts from about 475 a day to about 80. Every start heats the motor and arcs the pressure switch contacts, so that was a real repair that will lengthen the life of both. It did nothing to the kWh, because the kWh follows the 864 gallons, and the 864 gallons were still leaving.

The growing difference on the summer bills was the leak growing.

Finding where 864 gallons a day was going

Nothing in the house was wet, and the septic system showed no sign of extra load, so the water was leaving before it reached any fixture. The house had one branch off the well piping that ran outside: an 80-foot line to a frost-free yard hydrant by the garden.

A frost-free hydrant closes its valve at the bottom of its standpipe, below the frost line, and when it is shut, the water left in the standpipe drains out a small port at that buried valve into gravel, so nothing is left to freeze. When the plunger seal at the bottom wears, supply water keeps flowing past it and out that drain port, underground, continuously. Nothing shows at the spout. The only surface sign was a patch of soft ground and greener grass around the hydrant that the customer had put down to a wet summer.

The isolation test took twenty minutes. The hydrant branch had a valve in the well room. With every fixture in the house off, the tech closed it and watched the gauge: it sat at 60 psi for 20 minutes without moving, against the 15.8 minutes it had taken to fall the full 20 psi with the branch open. That test also clears the pump's check valve: a check valve leaking back down the well drops the pressure even with every branch shut, and here the pressure held.

The branch stayed closed until the repair, which stopped the loss and the cycling that night.

The repair. Replacing a hydrant means digging at it. The shop called 811 for a utility locate before any digging and waited for the marks, and then treated the marks as incomplete, because the lines most likely to be under a rural yard are private ones the locate service does not mark: the well's own buried pump cable and supply line, and the hydrant line itself. The crew hand-dug along the known line from the hydrant rather than trenching with a machine. The worn hydrant was replaced.

Putting it back. Branch valve reopened slowly with the hydrant shut, the line allowed to fill, then the house closed up and the gauge watched again: 60 psi held for 20 minutes. The hydrant was opened and shut once to confirm it delivered at the spout and drained only briefly at the buried port after closing, which is how it is meant to work.

The confirmation

The interval data three nights later: overnight intervals back to 0.10 kWh, with a single bump of about 0.036 kWh near 4 AM, when a toilet refill and the ice maker had drawn the tank down to cut-in. Midnight to 5 AM: 2.00 plus that 0.036, about 2.04 kWh, against 2.00 the same hours last September.

The November bill, the first full period after the repair:

That is inside the month-to-month noise for this house.

What the shop changed

The shop's high-bill walk-around sheet had been written for suburban houses on city water, and it had no line for a well. It gained one (the general list of forgotten loads is owned by The Loads a Customer Forgets They Own): on a well, with every fixture off, does the tank gauge hold steady for 15 minutes? If it falls, something is drawing water, and the pump's energy is following it.

The well service checklist gained the same question for tank replacements: a tank that is waterlogged AND cycles with nothing open has two problems, and the second one is the one on the bill.

References

• See related: Well Pump Service Call
• See related: The Loads a Customer Forgets They Own
• See related: Smart Meter Interval Data: What the Hourly Record Shows
• See related: Duty Cycle: Why Run Hours, Not Clock Hours, Set the Cost
• 811 and your state's one-call law for excavation notice; tank manufacturer instructions for precharge