Repair or Replace HVAC: Where the $5,000 Rule Breaks

Put two numbers on the table. A $1,180 repair — inducer motor and integrated control board — on a furnace whose rating plate dates it to 2007, against $8,400 to take the whole thing out. Those two figures are a worked example rather than a quote anyone handed me; every calculation below runs the same way on the numbers written on your own estimates.

Run the three rules people will quote at you over those numbers and you get two opposite answers. That is not a close call between good rules. It is a sign that at least two of them are measuring nothing.

Three rules, one furnace, two opposite answers

The 50 percent rule. Replace when the repair costs half of what a new one costs. Here, $1,180 divided by $8,400 is 14 percent. Repair.

The 25 percent version. The Minnesota Department of Commerce home energy guide on furnace replacement lists expensive repairs as a replacement trigger and puts the figure there: spending "25% or more of the cost of a new furnace on repairs may be an indication it is time to replace rather than repair." Same 14 percent. Still repair. (Read 22 August 2026. The same page puts a furnace's useful life at 15 to 20 years.)

The $5,000 rule. Multiply age in years by the repair quote; replace above $5,000. Nineteen times $1,180 is $22,420. Replace, by a factor of four and a half.

Two rules say fix it. One says the decision is not close. Nobody reconciles that for you at the kitchen table, so it is worth knowing what each one stands on.

The chart's residential rows are the ones ASHRAE marks N/A

Somewhere in the conversation a life-expectancy table appears — furnaces 18 years, split systems 15, heat pumps 15. It travels as a PDF headed "ASHRAE Equipment Life Expectancy chart," and it is real in the sense that ASHRAE published those numbers.

ASHRAE's own supplement on economic analyses and life-cycle costs, issued with the ninth edition of Principles of HVAC, says the estimates that circulated for decades were "based on a survey conducted in 1976 under ASHRAE Research Project RP-186 (Akalin 1978)" and that "changes in technology, materials, manufacturing techniques, and maintenance practices now call into question the continued validity of the original estimates."

That document prints Table 4, comparing the 1978 figures against the newer dataset from research project TRP-1237. Look up the residential rows. Window unit: N/A in the new data, 10 years in Akalin. Residential single or split package: N/A, 15. Residential air-to-air heat pump: N/A, 15. The asterisk resolves to a footnote saying there is not enough data yet in the new work, that the 1978 figures for those categories "may be outdated and not statistically relevant," and to use them with caution.

That footnote is easy to never reach. The chart travels as a standalone file with no date printed on it, while the sentences retiring it sit in the supplement above — a different document, found only if you go looking for one. Working from the version that gets emailed around, there is nothing on the page to suggest the 18-year furnace figure is anything but current.

The live version is open to anyone. ASHRAE's service life and maintenance cost database stamps its own front page — 345 buildings, 38,946 pieces of equipment, "database contents as of 8/22/2026". It does hold categories that sound like a house. Query the air distribution system type and "Residential furn. and pkg'ed condensing unit/heat pump" returns 132 units, "Split DX system" 174, "Heat pump, air-to-air" 1,296.

Then read the replaced-units column on those same three rows: 0, 1 and 1. Service life in this database is computed from equipment that has actually been replaced, so rows like these yield no median at all — the 132 residential furnaces are all still running, at a mean age of 31 years, which tells you when the survey caught them and nothing about when they die. The building function filter runs Office, Medical, Lodging, Manufacturing, Retail, Residential: Multi-Family, Dormitory, School, Laboratory, Restaurant, Public Assembly, Warehouse. There is no single-family house on that list, which is what you would expect of a database seeded with 163 commercial office buildings.

One thing matters more than the vintage. ASHRAE defines service life by when equipment gets replaced, "for any reason, including, but not limited to, failure, general obsolescence, reduced reliability, excessive maintenance cost, and changed system requirements." The median is a record of decisions like the one you are making. Using it to predict how long your furnace will run is circular.

A rule that multiplies dollars by years

The $5,000 rule has had a good run — CBS News explained it to a general audience, and it sits on the blog of nearly every contractor who will quote you. Not one page I opened names where it came from.

Look at what it does. It multiplies dollars by years and compares the product to a number in dollars. The units do not resolve, so the threshold cannot be argued with — you cannot ask whether $5,000 is the right cut-off, because the thing compared to it is not money. It is also fixed in nominal dollars, with no index and no base year.

Work out what that implies. At age 20 the rule replaces the equipment at any repair over $250 — a hot-surface igniter and an hour of labour. At age 5 it tolerates $1,000. It is harsher on a cheap fix to an old machine than on an expensive fix to a new one, and it has no idea whether the replacement costs $6,000 or $19,000.

What it gropes toward is real: a repair on old equipment buys fewer years. Put that in a form that survives your two quotes.

The only question underneath: how many years does this repair buy?

Both quotes purchase time. Price each per year of time bought and they become comparable.

Replacement. $8,400 over the 15 to 20 years Minnesota Commerce quotes for a furnace. At 18 years that is $467 a year. At the pessimistic end, 15 years, $560. At 20 years, $420.

Repair. $1,180 over however long the repaired furnace keeps running.

Set them equal. $1,180 divided by $467 is 2.5 years. That is the whole decision, restated as one checkable question: does this repair buy more than about two and a half more heating seasons? Against the pessimistic replacement figure the bar drops to 2.1 years; against the optimistic one it rises to 2.8.

Now the conversation with your contractor has something in it. Not "should I replace it," which invites the answer he sells, but "what else on this furnace is at end of life, and what would each cost?" A heat exchanger inspection result, a burner condition, the state of the blower — those bear on whether three more seasons is plausible. His opinion on a threshold rule is not.

The arithmetic transfers with different inputs. A water heater's tank is the machine; a leaking tank buys zero years no matter what the fittings cost. A compressor on a system running a phased-down refrigerant carries a supply risk in the repair column that a new system does not.

Counting the second failure before it happens

The division above assumes one repair. Nineteen-year-old equipment rarely obliges.

Add the term explicitly. Take a 40 percent chance of another failure inside those three years at an average of $800 — that 40 is a figure chosen to show the shape of the calculation, not one drawn from any dataset. The expected spend on the repair path becomes $1,180 plus 0.40 times $800, or $1,500. Over three years that is $500 a year, above the $467 replacement figure. The decision flips, and it flips on a probability you can state and argue with rather than on a rule.

So where should your version of that number come from? Not from any table — I could not find published age-specific failure rates for residential furnaces I would be willing to print here. What you have instead is your own file: pull the service invoices for this unit for the last three years and count them. Two calls in three years is a different machine from none in ten. State the number out loud even when you are guessing. A guess written down can be argued with; a rule of thumb absorbs the same guess invisibly and hands you a verdict.

What the warranty on each side is actually worth

Both columns have warranty consequences, and both get overstated.

On the replacement side. Trane's residential furnace limited warranty — the document, not a summary of it — covers "only for parts (but not labor)." Its exclusions name labour costs "including, but not limited to, costs for diagnostic calls or the removal and reinstallation of Products and/or Product parts," plus shipping and freight. Registration within 60 days of the commencement date lifts parts coverage from five years to ten. Claims must be filed within 90 days of the failure date. Coverage transfers to a buyer of the house only within 90 days of closing, fee paid, and the buyer gets the balance rather than a fresh term. Goodman runs the same 60-day structure, and exempts residents of California, Florida, Georgia and Quebec from registering. Both read 22 August 2026.

Put a number on it. Say one part failure in years six to ten, with a $400 part and $450 of labour. The warranty pays the $400. Discount that at the 3.0 percent real rate DOE published in the 2025 annual supplement to NIST Handbook 135 — the federal life-cycle cost method under 10 CFR 436A — and $400 eight years out is worth about $316 today. Registering is free and takes fifteen minutes, so do it. But $300 in present value is not the several thousand that "ten-year warranty" sounds like on a bid.

On the repair side, and this is the one people miss. A 2007 furnace is not necessarily out of coverage. Heat exchangers carry far longer terms than parts — Trane's current table lists 20 years base on the furnaces it covers, lifetime on some registered models. If the failure in front of you is a cracked heat exchanger on a nineteen-year-old unit, the part may still be covered while the labour is not. Ask before you agree to anything.

Two cautions. That PDF applies to products manufactured on or after 1 January 2022; for older equipment you need the certificate in force when yours was made, requested by model and serial number. And Trane's eligibility list requires the furnace to be part of an AHRI rated and matched system — the same check that decides whether the replacement being quoted is a certified combination. Looking the system up in the AHRI directory takes ten minutes and applies to both columns.

Bounding the fuel difference from your own gas bill

Efficiency belongs here, and it is the term most likely to arrive with no arithmetic behind it. You can put a ceiling on it from your own paperwork.

For the same heat delivered, fuel input scales as one over AFUE. Going from 80 percent to 95 percent cuts input by 1 minus (0.80 divided by 0.95), which is 15.8 percent. A ceiling, not a forecast.

Apply it to your bills. Take twelve months of therms and subtract twelve times your lowest summer month, which approximates the non-heating baseline — water heating, cooking, the dryer. What remains is heating. On a house burning 780 therms with a 12-therm July, that is 636 heating therms, and 15.8 percent of 636 is about 100 therms a year. Multiply by the per-therm charge on your own bill, not a national average.

Minnesota Commerce says the same upgrade "could save several hundred dollars a year in fuel costs (depending on many factors)," and that parenthesis is doing real work. Whatever you land on, subtract it from the replacement column. At $1.40 a therm those hundred therms are $140 a year, so the $467 becomes $327, and the repair now has to buy 3.6 years instead of 2.5 to win.

That is as far as this page goes on running costs. Comparing fuels, rate structures or an electrification sequence is a different question with different documents behind it.

When the two numbers land within ten percent

Sometimes the arithmetic refuses to decide. $467 against $500 is not a real gap when every input is an estimate. Stop squeezing the numbers at that point and let the non-financial terms break the tie.

Timing is one. A replacement chosen in August is a different transaction from one chosen at 11pm in January, and the fourteen lines a complete replacement bid has to contain are much harder to insist on when the house is cold. A repair that buys a shoulder season to shop in is worth something the division does not capture.

Scope risk runs the other way. A replacement opens ductwork, venting, condensate and electrical connections that a repair leaves alone, and each is a place a fixed price can move once the old unit is on the driveway. Read the change-order clause before you sign: on a tie, the side carrying that clause is the more expensive side.

What none of this settles is safety. A suspected cracked heat exchanger is not an arithmetic problem, and no figure on this page belongs in that decision. I hold no licence and inspect nothing; everything above is documents compared against each other. Settle the condition question first, then bring the numbers back out.