The fleet interval, not the device interval
This is the arithmetic that catches everybody, and it catches them because each individual number is perfectly reassuring. A contact sensor lasts two years. A motion sensor lasts eighteen months. A lock lasts seven. None of those sounds like a chore.
Then you count. Fourteen contacts at 24 months is seven changes a year. Six motion sensors at 18 months is four. Five leak sensors at 30 months is two. Two locks at 7 months is 3.4. Four remotes at 14 months is 3.4. That is 19.9 device battery changes a year across 31 devices, which is one every 18 days. The weighted average device life is 18.7 months and you are still holding a screwdriver twice a month.
The general form is simple enough to do in your head: changes a year is the device count divided by the average life in years. Thirty devices at two years is fifteen a year, one every 24 days. Sixty devices at two years is one every twelve. The interval you experience scales with the size of the house, and no product decision changes that, because it is a property of the fleet rather than of any device in it.
Where the changes actually come from
The group generating the most work is rarely the one with the shortest life. On the default list the locks have the worst interval by a wide margin at seven months, but there are only two of them, so they account for 17 percent of the changes. The contacts last three and a half times as long and there are seven times as many, so they account for 35 percent. If you want fewer interruptions, the lever is the biggest group, not the worst device.
The locks do dominate one column, though: at four cells each they consume 13.7 cells a year against the contacts at 7, so they are the largest line in the cell budget while being a minority of the trips. Cells and trips are two different currencies and the same list ranks differently in each.
The cost that never gets counted
The cells on that list come to about $17 a year, which is nothing, and it is the number people quote when the subject comes up. The 19.9 changes at six minutes each are two hours a year, and at $30 an hour that is $59.57 — 3.6 times the cost of the batteries. Whether you value your own time in dollars is up to you, but the two hours exist either way, and they are not spread evenly. They arrive as twenty separate interruptions, each one announced by a device that has decided now is the moment.
That is the argument for the batch mode on this page, and it is worth being precise about what it does and does not buy. Doing a whole group at once does not save a single minute — every device still has to be opened, so the two hours are unchanged, and the page shows the same figure in both modes. What changes is the number of occasions: 19.9 interruptions become 4.1 scheduled sessions, one every 88 days instead of one every 18. If the cost of this job is mostly the interruption rather than the minutes, that is the whole saving. If it is mostly the minutes, batching does nothing for you.
It also has a cost the page does not model. Replacing a group together means throwing away cells that still had months in them, and the cell count here does not discount for that, so the cell line is optimistic for batching by however much life you discard. On this list that error is bounded by about $17 a year, which is why it is not worth modelling precisely.
One thing batching does not fix: the devices where a flat cell has a consequence beyond the inconvenience. Those want a calendar entry rather than a wait for the low-battery report, and which ones they are is a judgement about your house rather than a line in a table. Per-device life comes from the sensor battery life calculator.
Questions people ask
How many batteries does a smart home get through in a year?
Divide the number of battery devices by the average life in years, then multiply by cells per device. Thirty-one devices averaging 18.7 months, with one to four cells apiece, comes to 34.1 cells a year on the default list here. The count that surprises people is not the cells, it is the 20 separate occasions on which somebody has to go and deal with one.
Should I change them all at once or wait for each one to complain?
Both modes are on the page because both are defensible, and the page is careful about what separates them. Batching does not save time — every device still gets opened, so the two hours a year are identical either way. What it changes is the count of occasions: 19.9 interruptions become 4.1 sessions, one every 88 days instead of one every 18. Against that it wastes the life left in the cells you replace early, and it loses outright if a device failing between sessions matters.
Why is my change rate worse than this says?
Usually because the real battery lives are shorter than the box figures you typed in. Cold locations, devices at the edge of coverage retrying every transmission, and busier rooms than the rating assumed all cut the interval. Put in what you have actually observed rather than what was claimed, and the fleet figure will match what your calendar has been telling you.
Is it worth buying cells in bulk?
To a point. At 34 cells a year a pack of 20 is about seven months of supply, which is a sensible amount to hold. Cells lose capacity sitting in a drawer, and the self-discharge that costs a sleeping sensor a few percent of its budget is doing the same to your spares, so a five-year stock bought cheaply is partly a five-year stock of weaker cells. The page shows how long any stock lasts at your rate.
Does this cover rechargeable devices?
Only if you put the recharge interval in the life column and set the cell price to zero. The trip count and the time cost are then still correct, which is the main thing, because a device you have to take down and charge every three months costs you far more attention than a cell you swap every two years. What it does not model is the charger, the spare packs or the capacity fade over cycles.