Skip to main content
RepairOrReplace

Methodology

How the math works

Weibull survival curves, localized labor rates, and net-present cost — every step is open and citable.

Model reviewed July 19, 2026 · BLS OEWS 49-9031 · NAHB Housing Survey · EIA residential rates

  1. Step 1 — You provide the basics

    Appliance category, rough age, and the technician's quoted repair price. Brand tier, failed component, and your metro area sharpen the estimate but are optional.

  2. Step 2 — Weibull survival probability

    We fit a two-parameter Weibull distribution to NAHB appliance lifespan surveys. The survival function S(t) = exp(−(t/η)^β) gives the probability the appliance is still running at age t. Scale parameter η is the characteristic life; shape β controls whether failures accelerate with age.

  3. Step 3 — Net-present-cost comparison

    We compute two 10-year cost streams: (A) repair cost + expected future costs over the remaining life, discounted at 5 %/yr; (B) replacement cost now. Repair costs include parts, localized BLS OEWS 49-9031 labor, and a repeat-failure probability premium. If NPC(repair) < NPC(replace), the model recommends repair.

  4. Step 4 — Energy penalty

    When an Energy Star flag is available, the annual energy cost delta between the current appliance and a replacement is folded into the NPC replace stream using EIA state residential rate data.

  5. Step 5 — Confidence scoring

    We score how many optional fields were filled, weight them by their effect on the NPC delta, and report a 0–100 confidence figure. A low score means the verdict is directionally correct but the margin may shift with more data.

Limits and assumptions

  • Labor rates use metropolitan area means from BLS OEWS. Rural rates may differ by ±15 %.
  • Lifespan curves are fitted to NAHB survey medians. Individual units vary widely by usage and maintenance.
  • Energy savings assume replacement with a mid-tier Energy Star model. Actual savings depend on the unit chosen.
  • Repeat-failure probabilities are derived from appliance-specific failure mode prevalence data, not per-unit condition.