Preventive Maintenance: Only 54% Gets Done, Limble Survey Finds

On 8 October 2026, Manufacturing Dive reported the results of a new survey by Limble, a maintenance software vendor, conducted among 686 maintenance and operations professionals in the United States. The key finding: teams complete only 54% of their scheduled preventive maintenance (PM).

The key numbers

According to the publication, the survey found that:

• 47% of maintenance managers report more than 10 hours of unplanned downtime in a typical month;
• missed PM can lead to downtime costs of $100,000 or more;
• 65% of respondents believe reactive maintenance, compared with preventive maintenance, increases risks for technicians;
• more than half of teams use or are piloting artificial intelligence tools, and 92% agree that AI needs "good and complete" maintenance data to work.

Limble attributes the gap to three causes: little time available for hands-on work, skills that are never documented, and inconsistent use of maintenance data. The article also notes that few manufacturers actually follow a data management plan; we have not seen the full development of this point, so we only flag it here.

Why it matters for OEE

OEE is made up of availability, performance and quality. Unplanned stops hit the first component directly: every hour of sudden breakdown is planned production time that disappears. If PM is postponed because time or people are lacking, the risk is trading a short, scheduled intervention today for a longer, unexpected stop tomorrow, with higher costs and greater danger for those working on the machine.

A preventive maintenance plan that exists only on paper does not protect availability: what counts is what gets carried out and recorded.

The data problem before AI

For anyone running a plant, the most interesting finding may be the link between AI and data quality. If most teams are evaluating predictive algorithms but knowledge stays in the heads of the most experienced technicians and records are inconsistent, models risk learning from incomplete information. Predictive maintenance presupposes reliable histories of failures, interventions, causes and machine operating conditions.

Caveats when reading the results

Before drawing conclusions, three limitations should be kept in mind:

• the sample is exclusively American, and the European or Italian picture may differ;
• the survey was commissioned by a maintenance software vendor with a commercial interest in the topic;
• the data is self-reported, so it reflects perceptions rather than field measurements.

That said, the figures are consistent with a problem many manufacturers recognize: PM is the first activity to be skipped when production is under pressure.

What to do in practice

A few concrete steps, independent of the vendor you choose:

• measure your own PM completion rate (work orders completed versus planned) and compare it with the survey's 54%;
• link recorded machine stops, with reason codes, to maintenance work, so you can see how much downtime stems from skipped PM;
• plan PM based on actual production load, for example through an MES that schedules and tracks interventions alongside machine data;
• document currently informal procedures and skills, turning them into accessible checklists;
• define minimum data quality rules before launching AI projects.

In short, the survey is a reminder that production efficiency is not regained through new technology alone: it starts with maintenance carried out consistently and with data that faithfully describes what happens on the shop floor.