Robotics in Manufacturing

Manufacturing economy

Preventive versus predictive maintenance: what the savings actually are

Predictive maintenance saves an estimated 8% to 12% over a preventive programme, and preventive saves 12% to 18% over running to failure. Those are the Department of Energy's figures, and they are far more modest than the ones in most vendor material.

The wider question of what maintenance costs has no settled answer. NIST reports Census data of $50 billion spent on outsourced maintenance and repair, a figure that excludes in-house labour entirely, and notes literature estimates ranging from 15% to 70% of the cost of goods sold.

This page assembles the federal estimates and one peer-reviewed measurement, so the savings case can be made with numbers that carry a source.

Data covers US federal maintenance-economics estimates and a peer-reviewed downtime study. Edited by Mike Ramsey / Reliable Media.

The figures and where they come from

Each figure is rated for how safely you can cite it today. Ratings judge current usability, not whether a number was ever correct.

FigureWhat it isSourceCitation ConfidenceNotes
8% to 12% savingsPredictive over preventive[A]HighDOE's estimate of predictive maintenance savings against a preventive programme. Note it is against preventive, not against running to failure.
12% to 18% savingsPreventive over reactive[A]HighDOE's estimate for preventive maintenance against a reactive programme. Stacking the two is where large headline savings claims come from.
35% to 45%Downtime reduction, predictive (surveys reported by DOE)[A]MediumReduction in downtime from a predictive programme. The DOE guide attributes this to independent surveys of industrial average savings rather than presenting it as its own estimate, so cite it that way.
$50 billion (outsourced)US outsourced maintenance and repair spend[B]MediumCensus data for 2016, as reported by NIST. This covers OUTSOURCED maintenance and repair including buildings, and excludes internal spending on labour and materials, so it understates total maintenance considerably.
15%Maintenance as share of cost of goods sold, low end[B]LowNIST reports literature estimates from 15% to 70% of cost of goods sold. A range that wide is a sign the quantity is poorly defined, not a usable figure.
70%Same estimate, high end[B]LowThe top of the same range. Anyone quoting a single percentage for maintenance as a share of cost of goods sold is picking a point inside this spread.
52.7% lessLess unplanned downtime, better-maintained half[C]MediumA measured comparison rather than an estimate: the half of plants relying more on predictive and preventive maintenance had 52.7% less unplanned downtime.
78.5% fewerFewer defects, same group[C]MediumThe same group also reported 78.5% fewer defects. Measured, but observational, so it does not establish that maintenance strategy caused the difference.

Why the numbers disagree

The DOE savings figures look small next to vendor claims because they are incremental and paired. Predictive saves 8% to 12% over preventive, and preventive saves 12% to 18% over reactive. Vendor material often compares predictive directly against running to failure, which stacks both steps, and sometimes against a hypothetical worst case rather than an actual programme.

The denominators are unsettled in a way that undermines any single savings percentage. The $50 billion Census figure covers only outsourced maintenance and repair, including buildings, and excludes internal labour and materials, so it is a floor rather than a total. NIST separately reports literature putting maintenance anywhere from 15% to 70% of cost of goods sold. A saving of 10% means something very different against 15% of costs than against 70%, so a percentage saving without a stated base is close to meaningless.

The peer-reviewed measurement points the same direction but cannot carry the causal claim. Plants relying more on predictive and preventive maintenance had 52.7% less unplanned downtime and 78.5% fewer defects. That is a real, measured association, and it is also exactly the pattern you would expect if better-run plants do several good things at once.

How to cite these figures

Quote the DOE savings as incremental and say what the comparison is: 8% to 12% for predictive over preventive, 12% to 18% for preventive over reactive.

State the base whenever you give a savings percentage. Without knowing maintenance as a share of costs, the percentage cannot be turned into money.

Use the peer-reviewed downtime and defect figures as an association, not a causal effect. The wording that survives scrutiny is that plants leaning on predictive and preventive maintenance reported far less unplanned downtime.

Treat the 15% to 70% of cost of goods sold range as a warning rather than a statistic. It is too wide to plan with.

Where people go wrong

Stacking the DOE savings into one headline number without saying you have. Predictive against reactive is two steps, not one.

Quoting a single figure for maintenance as a share of cost of goods sold. The published range spans 15% to 70%.

Presenting the downtime study as proof that predictive maintenance causes the improvement. It is observational, and better-run plants differ in many ways.

Comparing a predictive programme against a hypothetical worst case. DOE's comparison is against an actual preventive programme, which is a much harder benchmark.

Quoting the $50 billion as total manufacturer maintenance spending. It covers outsourced work including buildings and excludes internal labour and materials.

How we checked

The savings figures come from the Department of Energy's operations and maintenance guidance, the spending and cost-share figures from a NIST analysis, and the downtime and defect figures from a peer-reviewed study accessed through PubMed Central. All three were retrieved and confirmed to contain the figures quoted.

NIST prints its percentages with a space before the sign, as 15 % and 70 %, which is a detail that matters only because a naive search for 15% will not find them. We verified against the text as published.

We rate the cost-of-goods-sold figures Low deliberately. They are correctly reported from NIST, and they are still not usable as a planning number, because the underlying literature spans more than a fourfold range.

We looked for an independent measurement of maintenance-strategy savings, as opposed to a government estimate or a vendor claim, and found one observational study. The absence of controlled evidence in this area is itself worth knowing.

Full source list

Primary sources, with live links. Every figure above traces to one of these.

  1. [A]U.S. Department of Energy (FEMP)Accessed July 2026

    US Department of Energy, Federal Energy Management Program, Operations and Maintenance Best Practices Guide, chapter on maintenance types

    https://www1.eere.energy.gov/femp/pdfs/om_5.pdf
  2. [B]National Institute of Standards and Technology2018

    NIST AMS 100-18, economics of manufacturing maintenance

    https://nvlpubs.nist.gov/nistpubs/ams/NIST.AMS.100-18.pdf
  3. [C]PMC (PubMed Central)2023

    Peer-reviewed study of maintenance strategy and unplanned downtime, PMC

    https://pmc.ncbi.nlm.nih.gov/articles/PMC9890517/

Common questions

How much does predictive maintenance save?
The US Department of Energy estimates 8% to 12% over a preventive maintenance programme, and preventive saves a further 12% to 18% over running equipment to failure. Larger headline claims usually stack both steps or compare against a worst case.
What does maintenance cost a manufacturer?
There is no settled figure. Census data reports $50 billion spent on outsourced maintenance and repair in 2016, which excludes in-house labour and includes buildings. NIST separately notes published estimates of maintenance running anywhere from 15% to 70% of cost of goods sold, a range too wide to plan against.
Is there measured evidence, not just estimates?
One peer-reviewed study found the half of plants relying more on predictive and preventive maintenance had 52.7% less unplanned downtime and 78.5% fewer defects. It is observational, so it shows association rather than proving cause.
Why are vendor savings claims so much larger?
Usually because they compare predictive maintenance directly against running to failure, which combines two separate improvements, and sometimes against a hypothetical worst case rather than a real preventive programme.

More data, traced to source