A plant can measure machine utilization without an MES by collecting run time, scheduled time, cycle counts, and downtime records, then calculating productive machine hours divided by scheduled machine hours and multiplying by 100 [S4].
The method works for CNC machines and other shop-floor assets, but the denominator must distinguish calendar time from scheduled production time, and controller signals must be filtered so that idle, setup, and alarm states are not counted as productive operation [S1][S4].
Define the Measurement Before Collecting Data
Machine utilization is the share of a defined time period during which an asset is running and producing, with the standard calculation expressed as productive machine hours divided by scheduled machine hours, multiplied by 100 [S1][S4]. The same calculation can produce materially different results depending on the chosen time basis.
Machine utilization is typically calculated as productive machine hours divided by scheduled machine hours, and the choice of denominator matters as much as the calculation [S4]. For single-shift operation, 6 productive hours in an 8-hour schedule gives 75% [S4]. These figures are not contradictory; they answer different capacity questions.
A machine-utilization overview identifies five time layers and explains which metric measures each one, while the most common formulation divides run time by available time [S1].
Choose the Data Collection Method
Manual operator timestamps are the lowest-cost starting point, but their accuracy depends on disciplined job start and finish recording. One production system can capture a start at 10:47 and an end at 10:59, producing 12 minutes of reported job time [S2]. That method works for short runs only when operators record every event consistently.
Controller-based collection can support tracking machine utilization [S4]. A practical manual collection sheet should include the asset identifier, date, scheduled start and finish, actual start and finish, work-center or job number, planned stop, breakdown, setup, and operator identification. Each row then provides the numerator and denominator for one asset and reporting interval.
Other non-MES options include barcode or QR-code job scans, handheld terminals, relay or current-switch signals on auxiliary equipment, and manual recounts of cycles. Kinetic ERP users can enable the Machine MES module in pilot and consult the Kinetic ERP Mattec Integration documentation to learn how to integrate with their machines [S2].
Separate Running Time from Productive Time
Running is not synonymous with producing. Utilization measures whether the machine is operating during the selected window, whereas OEE combines Availability, Performance, and Quality to determine whether that operating time was effective [S5]. A machine can remain active while running below its intended rate or producing defective parts.
OEE combines Availability, Performance, and Quality, and a machine can show high utilization while having low OEE [S5]. The distinction prevents a high utilization figure from being treated as proof of good output.
The preferred hierarchy is to log in-cycle program execution first, spindle activity second, and general machine-running state third. Where the process requires operator attendance, combine the control signal with operator confirmation. Planned setup, tool changes, inspection and rework should remain separate categories, not be silently included in productive hours.
Compare the Main Measurement Options
The main comparison is not simply manual versus automatic. It is whether the method can supply stable event data, explain loss causes, and fit the maintenance burden. A spreadsheet-based method can track a small fleet without new software, but it depends on manual timestamps and periodic reconciliation. Controller software offers cycle-level visibility with less manual entry, yet it may expose only machine states and not job-specific reasons for downtime. [S4]
Neither approach should be selected solely from machine count; the loss categories and reporting frequency must match the decision being made.
OEE reporting then adds Performance and Quality data that a run-time counter alone cannot supply [S5].
Apply the Method to Common Decisions
Utilization tracking is suitable for scheduling gaps, downtime detection, bottleneck analysis, and capital planning. It can establish whether equipment is available when work is planned and whether additional capacity should be scheduled before another machine is purchased [S3]. A baseline also exposes whether assumptions about workload match machine behavior [S3].
The metric is not suitable as a universal productivity target. Pursuing 100% utilization can create excess inventory, delay maintenance, and reduce reliability, while idle time can be appropriate where output remains stable and defect-free [S5]. For non-bottleneck assets, OEE or a production-output measure is generally more informative than run-time percentage.
Typical loss categories include unplanned downtime, long changeovers, scheduling gaps and operator idle time [S4]. Weekly review should separate these causes rather than presenting one blended percentage. A shop with 55 productive hours out of 80 scheduled hours has 25 lost hours, but those hours cannot guide action until they are assigned to breakdowns, setup, starvation, planned stops or other documented causes [S4].
Control Data Quality and Measurement Limits
At least 3 controls are necessary for a defensible manual measurement system: a published state-code list, fixed reporting boundaries, and weekly reconciliation of machine logs against work orders. Changes to the denominator must be versioned because switching from 80 scheduled hours to 168 calendar hours changes the reported result from 68.75% to 32.7% for the same activity [S4].
Manual systems are vulnerable to late data entry, forgotten job endings and operator interpretation. Automatic systems face different failure modes, including incorrect controller mappings, program aliases, no-part states and communication gaps. A pilot should therefore compare 2 independent weeks of controller or manual output against a sample of observed shifts before the measurement is used for capacity decisions.
Benchmarks also require caution. A 2026 manufacturing overview says that most manufacturers assume their CNC utilization is significantly higher than the average rate of 25.9% reported across more than 12,000 connected machines [S4]. A separate 2026 article cites 25.9% average CNC utilization across more than 12,000 connected machines, with many shops clustering at 17% to 20% [S4]. These figures are context, not universal pass marks; the denominator, industry and measurement definition must remain visible.
Build the First Measurement Cycle
A workable non-MES pilot has 5 stages: define the metric, select 1 asset class, choose the event source, classify the states, and reconcile the result. Use one denominator for the first reporting period, document every exclusion, and preserve raw timestamps rather than storing only percentages. The first useful deliverable is a loss profile showing productive, running, idle, setup, breakdown and planned-stop hours for each asset. [S1]
The MES comparison is mainly about scale and integration. A small plant can begin with spreadsheets, operator scans and controller exports, while an MES evaluation becomes more relevant when the same event model must feed scheduling, quality, work orders and multi-site reporting. The operational case for ERP-versus-MES selection is covered in Small Plant MES vs ERP: When ERP Is Enough and When MES Pays.
The next trackable signal is not a higher percentage, but a stable reconciliation rate between machine events and reported jobs. Establish that rate for 2 consecutive weeks, then review whether the dominant loss remains setup, breakdown, starvation or unplanned idle time before changing schedules or equipment.