The most common version of this question doesn’t come with a repair estimate attached — it comes with a verdict already reached: “they said it’s not worth fixing.” Sometimes that’s the right call. Often the math behind it was never actually shown, just asserted.
The short answer
The real comparison is repair cost against remaining expected service life, not repair cost against a flat percentage of a new system’s price. A widely used shorthand — “if repair costs more than 50% of replacement, replace” — is a reasonable starting heuristic, not a rule, because it treats a 5-year-old system and an 18-year-old system as if they were the same decision. They aren’t. The number that actually matters is what you’re paying per remaining year of expected life either way.
Why the shortcut breaks
The 50% rule was built as a fast filter for a specific, common situation: an old system near the end of its expected life with an expensive repair. In that case it usually gives the right answer, because there isn’t much remaining life left to protect. It breaks down at the edges, in both directions. A young system with one failed, inexpensive part can get an inflated repair quote that clears the 50% threshold on a cheap system, making replacement look justified when it isn’t. An old system near end-of-life with a genuinely cheap repair can fall well under 50% and still be a bad deal, if that repair is likely the first of several as other original-equipment parts reach the same age in the same conditions.
What the rule leaves out entirely is how much life the repair buys you. A $500 repair that restores 10 remaining years is a very different trade than a $500 repair that restores 18 months before the next failure.
A second shortcut floats around alongside the 50% rule: multiply the unit’s age in years by the repair estimate, and treat anything over roughly $5,000 as a signal to replace. It fails from the opposite direction. Age is only useful as a stand-in for remaining life, and this rule uses it directly instead — so it can push a long-lived 15-year-old system toward replacement over a $350 repair (15 × $350 = $5,250) without checking whether that specific system’s condition says it has years left, and it can wave through a 6-year-old system with a real $900 problem (6 × $900 = $5,400, right at the line) without ever asking whether $900 buys 10 more years or a repeat visit next season. Both shortcuts are trying to approximate the same thing — age × repair cost is a proxy for a proxy, and 50%-of-replacement skips straight to a dollar comparison that never asks about time at all. Cost per remaining year is the number both are reaching for, calculated directly instead of guessed at sideways.
What a repair like this actually costs
The repair estimate itself deserves a sanity check before it goes into either calculation above. Real current prices for the parts that fail most often, checked directly against a public HVAC parts retailer on 2026-08-19:
| Component | Parts price (typical) | What installed labor adds |
|---|---|---|
| Igniter (hot surface, universal) | ~$18–$25 | Usually a same-day, single-visit job |
| Dual-run capacitor | ~$3–$10 | Usually a same-day, single-visit job |
| Blower motor (1/2 HP, direct drive) | ~$94–$132 | More labor — mounting, wiring, balancing |
| Integrated furnace control board | ~$160–$255 | More labor and diagnostic time; one of the more commonly misdiagnosed parts |
Parts-only, sourced from a representative sample of listed prices on SupplyHouse.com within each category — not every SKU, and not an installed price. Labor is left out on purpose rather than guessed at: it’s the number that varies most by contractor and region, and a made-up range would be worse than none. What the table is useful for: if an estimate comes in well above the parts price plus a couple hours of labor, that gap is either real added complexity or margin — a fair, specific thing to ask about.
Two categories from the usual list are deliberately left out. A refrigerant leak repair is priced by leak location and diagnostic time, not one part — a single number here would flatten a $200 fix and a $1,500 one into the same line. Compressor replacement is close to a full system decision in its own right, and is often factory-authorized-technician-only — a bare parts price would undersell what that job actually involves.
A worked example (illustrative, not a real quote)
Two hypothetical furnaces, same $4,800 replacement cost, to show how the same repair price reads differently depending on what it’s buying:
| Furnace A | Furnace B | |
|---|---|---|
| Age | 6 years into an 18–20 year expected life | 17 years into an 18–20 year expected life |
| What failed | Igniter | Heat exchanger |
| Repair cost | $400 | $2,100 |
| Repair as % of replacement | 8% | 44% |
| Remaining expected life if repaired | ~12–14 years | ~1–3 years |
| Cost per remaining year, repair path | ~$30/yr | ~$700–2,100/yr |
Furnace A clears the 50% rule easily and should obviously be repaired — nobody needs the full math for that one. Furnace B is the case that actually needs it: 44% is comfortably under the 50% shortcut, which would say “repair,” but the cost-per-remaining-year number tells a different story once the short remaining life is priced in. This is a constructed example to show the shape of the calculation, not a real submitted quote — the actual numbers for your system depend on your equipment’s rated life, your repair estimate, and your climate.
What to ask
Ask what specifically failed, and whether it’s a component with a known high failure rate industry-wide (igniters, capacitors, blower motors, control boards) or a core, non-serviceable component (heat exchanger, compressor) whose failure often signals the rest of the system is at a similar age-related risk. Ask for the repair estimate as its own number, separate from any replacement pitch, so you can do the comparison yourself rather than being handed a conclusion. And ask what remaining service life a neutral source — not the person selling the replacement — would estimate for your equipment’s age and model; service life varies more by category than most people expect, which deserves its own piece.
What a bad answer sounds like
“It’s old, so it’s not worth fixing” without naming what failed or its cost is a conclusion with the math withheld — the age is real information, but it isn’t the whole calculation, and stated alone it’s doing the job of evidence without being any.
“The repair is almost as much as new” deserves a follow-up: almost as much as new what, exactly, and compared against how many remaining years either way? A repair that’s 40% of replacement cost on a system with over a decade of remaining life is a very different trade than the same 40% on a system with one season left in it.
“Parts are hard to find for this model” is sometimes true and sometimes an unverifiable claim used to close a decision on the spot — ask for the part number and take a day to check yourself before treating it as settled.
One to look at yourself
The comparison above works fine as a table with real numbers substituted in — a homeowner’s actual repair quote, replacement quote, and equipment age dropped into the same format. No real example collected yet for this one; the illustrative version above stands in until a real one is available to publish alongside it.