The number, and where it comes from
This is one of the few maintenance statistics with a federal agency behind it. The U.S. Department of Energy's guidance on air conditioner maintenance states that replacing a dirty, clogged filter can lower an air conditioner's energy consumption by 5 to 15%(energy.gov). Not the building's energy, the unit's own energy. It is also the cheapest fault in HVAC to fix, which makes it the clearest example of how deferred maintenance turns into a utility bill.
The figure in dollars
A 5-ton commercial cooling system in NYC costs about $1,432 a year to run on a deliberately conservative $0.18 per kWh marginal rate (the derivation is shown in What HVAC really costs a NYC building per year). Apply the DOE range to that line:
| Filter penalty (DOE 5–15%) | 1× 5-ton unit | 50-ton building (10× 5-ton) |
|---|---|---|
| Baseline cooling OpEx | $1,432/yr | $14,317/yr |
| At 5% wasted | ~$72/yr | ~$716/yr |
| At 15% wasted | ~$215/yr | ~$2,148/yr |
Those figures sit on the conservative marginal rate. On an all-in NYC commercial rate of $0.24 to $0.30 per kWh, multiply them by 1.33 to 1.67. And this is one fault. A low refrigerant charge, the most common service fault, costs another 10 to 20% of unit efficiency on its own (Downey & Proctor 2002, a 13,000-unit ACEEE field study). A neglected unit usually carries several faults at once.
Why nobody notices
A dirty filter almost never generates a complaint, because the system compensates. The unit runs longer to move the same heat, and the space still reaches setpoint. The occupant experiences nothing. The only place the fault is visible is in behavior no human tracks: runtime creeping up for the same weather, the temperature drop across the coil narrowing, current draw rising for the same output. The comfort system hides the fault, and the utility meter absorbs it.
The compounding path
Left long enough, the cost stops being a percentage and becomes a part. The mechanical sequence:
- Airflow restriction. The clogged filter chokes air across the evaporator coil, so less heat reaches the refrigerant and the unit runs longer for the same cooling.
- Coil icing. With too little warm air passing over it, the coil can drop below freezing and ice over. Ice blocks airflow further, which makes more ice. Capacity falls while the unit keeps drawing power.
- Compressor stress. The compressor now runs long cycles against abnormal conditions, and refrigerant that should have evaporated in the coil can return to it as liquid. That is the wear path that shortens compressor life, and the compressor is the most expensive component in the system.
The same neglect that wastes 5 to 15% on the energy line is the front end of the failure that takes the unit down on the hottest day it runs.
What a baseline sees that a calendar misses
The standard defense is a schedule: change the filters every one to three months. A schedule knows the date. It does not know that one rooftop unit sits over a dusty renovation, that a filter collapsed and is bypassing air, or that the filter changed last week is already loading. Between visits, the schedule is blind.
A learned per-unit baseline watches the symptoms instead of the calendar. AirPulse learns what each unit's runtime, temperature drop, and cycling look like when healthy, in each band of outdoor weather, then compares every hour against it. A loading filter shows up as a drift from that baseline (a compressed delta-T, runtime creep) with a dollar estimate attached, weeks before any human would notice, and the alert arrives whether the filter is three months old or three weeks old. The mechanics are on How it works, and pricing shows how the subscription compares against exactly this class of waste. Actual savings depend on your equipment, runtime, and utility rate.
Sources
- U.S. Department of Energy, Air Conditioner Maintenance (dirty, clogged filter: 5 to 15% of an AC's energy consumption): energy.gov/energysaver/air-conditioner-maintenance
- Downey, T. & Proctor, J. (2002), “What Can 13,000 Air Conditioners Tell Us?”, ACEEE Summer Study (improper refrigerant charge: 10 to 20% of unit efficiency): aceee.org/files/proceedings/2002/data/papers/SS02_Panel1_Paper05.pdf
- Dollar baselines: AirPulse shown-math derivation, tons × ~1.0 kW/ton × ~1,600 equivalent full-load hours × $0.18/kWh, detailed in What HVAC really costs a NYC building per year.