← Back to Decision Briefs
Sidy's Intelligence Brief — Decision Briefs

Repair or Replace? Compare the Future, Not the Sunk Cost

2026-10-0117 min read

The repair-or-replace decision should compare two future systems, not defend money already spent. Keeping the asset means future repair cost, expected failures, downtime, energy, quality loss, supportability and residual risk. Replacing it means capital cost, transition and commissioning risk, training, ramp-up and the new asset’s future operating profile. The rational choice is the lower-risk, lower-life-cycle-cost path that still meets the required performance — regardless of how much has already been spent on the old machine.

Asset renewalLife-cycle costReliabilityCapital decisionsIndustrial operations

The Decision in One Sentence

Ignore sunk cost; compare the expected future cost, availability, risk and capability of keeping the asset with the expected future cost, transition risk and performance of replacing it.

Why This Is Not the Same as Maintenance Strategy

A maintenance-policy decision asks how a failure mode should be managed while the asset remains in service: run to failure, inspect, monitor, maintain periodically, redesign or add redundancy. Repair-or-replace asks a higher-level question: should this asset remain in the system at all?

The two interact, but confusing them can trap an organization in endless repair optimization on an asset whose future economics, supportability or capability no longer justify ownership.

The Sunk-Cost Trap

Managers often say: “we have already spent too much on this machine to replace it now.” That statement looks financially disciplined but mixes past and future. Money already spent and unrecoverable is not changed by today’s decision.

The relevant question is whether the next unit of money should go into the old system or the replacement path. A recent overhaul may improve the old asset’s future condition and therefore belongs in the forecast through its actual remaining benefit — but the historical invoice itself is not a reason to keep the machine.

Compare Full Future Cost, Not the Next Repair Invoice

Life-cycle costing exists because the lowest initial cost and the lowest long-run cost are often different. NIST’s life-cycle-cost guidance evaluates alternatives across investment, operation, maintenance, replacement and other costs over the study period. Reliability research adds another requirement: failures, repairs, replacements and downtime affect both cost and availability.

For an industrial asset, the keep/repair case should therefore include expected maintenance labor and parts, failure frequency, production loss, scrap or quality impact, energy and utilities, specialist support, cybersecurity or control-system obsolescence, spare-parts availability and eventual disposal. The replacement case should include purchase, engineering, installation, shutdown, commissioning, training, ramp-up losses, financing where relevant, and the new asset’s future operating costs.

Availability Can Dominate the Economics

A machine can be cheap to repair and expensive to own. If breakdowns are frequent, recovery is slow or the asset is a bottleneck, lost production can dominate the repair invoice. Availability-based life-cycle research explicitly incorporates reliability and maintainability because uncertain failures and repair times create uncertain downtime and cost.

This is why the correct metric is rarely “maintenance cost per year” alone. An old machine with modest maintenance spend but repeated bottleneck outages may be economically worse than a replacement with higher capital cost.

Obsolescence Can Retire a Machine Before Wear Does

NIST’s equipment pre-purchase guidance warns that equipment whose useful support life is shorter than the required operating period can create reliability and cybersecurity risk as it becomes obsolete. In practice, a machine can remain mechanically repairable while controls, software, sensors, vendor support or critical components disappear.

That changes the replacement threshold. The question is no longer “can we repair it?” but “can we reliably keep restoring it inside the business’s required time window for the remaining horizon?”

Replacement Also Has Failure Modes

Replacement should not be romanticized. New equipment can arrive late, fail site acceptance, require utilities or foundations that were missed, underperform at the real product mix, create software-integration problems, or take months to reach stable output.

A fair comparison therefore gives the replacement case its own transition risk: engineering uncertainty, shutdown duration, commissioning defects, operator learning, spare-parts setup, supplier dependence and the possibility that promised performance is not reproduced under plant conditions.

Evidence Map

  • Observed / NIST: life-cycle costing compares alternatives over a study period rather than on first cost alone.
  • Observed / NIST equipment guidance: maintenance effectiveness, expected life, reliability and end-of-life/obsolescence should be considered when selecting equipment.
  • Observed / reliability literature: failures, repair/replacement behavior, downtime and maintainability materially affect life-cycle cost and availability.
  • Observed / repair-replacement research: asset-management models treat repair versus replacement as a lifecycle decision under uncertainty, rather than a one-time invoice comparison.
  • Inference: an organization that excludes downtime, obsolete support and transition risk can systematically bias the choice in either direction.
  • Unknown: no generic formula can set the replacement threshold for a specific asset without plant-specific failure data, production economics, remaining demand, technology horizon and transition constraints.

Sidy’s Synthesis — Compare Futures, Not Histories

The asset decision begins today. Everything before today is evidence; only what changes after today belongs in the choice.

My extension is a two-future comparison. Build one forward-looking system for Keep/Repair and one for Replace. Use the same horizon, output requirement and risk standard.

Two-future test
Required future performance→Keep/Repair future↔Replace future→Choose lower total risk-adjusted burden

Keep/Repair future: repair now + expected failures + downtime + maintenance + energy/quality penalties + obsolescence risk − residual value.

Replace future: purchase + engineering + shutdown + commissioning + training/ramp-up + future maintenance/energy + transition risk − recovery/residual value.

Decision rule: replace when the expected future burden of keeping the asset exceeds the expected future burden of replacement for the required service, after transition risk and uncertainty are included. Past spend is evidence about the machine, not a vote for keeping it.

This is Sidy’s synthesis, not a framework named by the cited sources.

Five Triggers That Should Force a Replacement Review

  1. Failure burden is accelerating: more frequent or longer outages despite competent maintenance.
  2. Supportability is collapsing: parts, software, vendor expertise or compatible controls are disappearing.
  3. The asset no longer meets the process: capacity, quality, flexibility, energy or compliance requirements have changed.
  4. Recovery time exceeds business tolerance: a single failure now threatens customer or production commitments.
  5. A replacement changes the operating economics materially: not just a nicer machine, but a demonstrably lower future burden or higher required capability.

Build From This

  • Build a 3–5 year forward cost and availability history for the keep case using actual failure data.
  • Separate repair invoice from downtime consequence.
  • Quantify parts/support obsolescence as a restoration-time risk.
  • Require vendors to provide commissioning, ramp-up and support assumptions for the replacement case.
  • Run sensitivity cases for production volume, failure rate, energy price, installation delay and residual life.
  • Set a review trigger before the next major repair so the organization does not re-decide under emergency pressure.

Remember This

  • Sunk cost should not choose the future.
  • A cheap repair can preserve an expensive downtime problem.
  • Mechanical repairability is not the same as long-term supportability.
  • Replacement has commissioning and ramp-up risk and must be charged honestly.
  • Use the same time horizon and service requirement for both alternatives.
  • The decision is not old versus new. It is future burden versus future burden.

Primary sources

Facts, figures and quotations should be traceable to the sources below. Sidy's synthesis is labeled as synthesis and does not replace sourced facts.

  1. NIST — Life Cycle Costing Manual for the Federal Energy Management Program, Handbook 135e2025
  2. NIST — Manufacturers: Pre-Purchase Guide for Equipment
  3. NIST — Economics of Manufacturing Machinery Maintenance
  4. Ntuen & Moore — Approaches to life cycle cost analysis with system availability constraints
  5. Availability-based life cycle cost model: A simulation approach
  6. Sustainable asset management: A repair-replacement decision model