OEE-Adjusted Capacity Planning: How to Model Planned and Unplanned Downtime

Here is how most manufacturers build their capacity plan: take the number of machines, multiply by available shift hours, and call that capacity. A 10-machine shop running 8-hour shifts 5 days a week has 400 machine hours per week. The plan gets loaded to 380 hours — a comfortable 5% buffer — and everyone feels good.

Then the deliveries start slipping. A machine goes down for 6 hours on Tuesday. A changeover runs 90 minutes longer than planned. Three parts from the morning run fail inspection. By Thursday, the schedule is 22 hours behind and the expediting starts. By the following Friday, two customers are waiting and the buffer is gone.

The problem isn't the machines or the operators. The problem is that 400 hours of rated capacity and 400 hours of actual available capacity are completely different numbers. Overall Equipment Effectiveness (OEE) is the bridge between them, and most capacity plans don't use it.

This post covers OEE-adjusted capacity planning from first principles: how to build a realistic capacity model, how to treat planned versus unplanned downtime differently, and how RMDB integrates OEE data so your schedule reflects what machines can actually deliver.

For a broader view of how capacity feeds into finite scheduling, see our guide to finite capacity planning.

OEE as a Capacity Multiplier

OEE is defined as the product of three loss factors:

OEE = Availability × Performance × Quality

Each factor erodes rated capacity:

• Availability = (Planned Production Time − Downtime) / Planned Production Time. Captures time the machine was scheduled to run but didn't — breakdowns, unplanned maintenance, material shortages.
• Performance = (Actual Output × Ideal Cycle Time) / Operating Time. Captures speed losses — running slower than rated speed, minor stops, idling.
• Quality = Good Parts / Total Parts Started. Captures yield losses — scrap, rework, and non-conforming product that consumes machine time but produces no shippable output.

A machine running with 90% Availability, 85% Performance, and 94% Quality has an OEE of:

0.90 × 0.85 × 0.94 = 71.9%

That machine has a rated capacity of, say, 160 hours per month. Its OEE-adjusted demonstrated capacity is:

160 × 0.719 = 115 hours per month

Your capacity plan that assumes 160 hours per month is overcommitted by 45 hours — 28% — before you've scheduled a single job. That 28% overcommitment is the structural source of your chronic schedule slippage.

Theoretical, Rated, and Demonstrated Capacity: Using the Right Number

Three capacity definitions are in common use, and only one of them belongs in a production schedule:

Theoretical capacity is the physical upper limit — 24 hours per day, 7 days per week, zero downtime, zero setup time. Useful for engineering design calculations. Never appropriate for scheduling.

Rated capacity is the engineered throughput during planned operating hours. It assumes the machine runs at its design speed during all scheduled production time, with zero unplanned interruptions. This is the number quoted by machine manufacturers and used in most capital justification analyses. It's appropriate for evaluating machine specifications, not for scheduling real production.

Demonstrated capacity is what the machine has actually produced over a recent historical period, net of all OEE losses. It is calculated from shop floor data, not from engineering specifications. It accounts for the specific product mix, operator skill level, maintenance program quality, and downtime patterns in your specific facility.

Demonstrated capacity is the only number that should drive production scheduling. The gap between rated and demonstrated capacity is not a problem to be solved — it is a reality to be planned around. If you're scheduling to rated capacity and expediting to demonstrated capacity, the gap is your scheduling error, not your operations team's failure.

A practical benchmark: in job shops and mixed-manufacturing environments, OEE of 55-65% is typical for well-run operations. World-class OEE of 85%+ applies primarily to high-volume, dedicated equipment in continuous manufacturing. Capacity planning to an 85% OEE target in a job shop that's actually running at 60% creates a 42% structural overcommitment. Use your actual trailing 90-day OEE, measured, not assumed.

Separating Planned and Unplanned Downtime: Two Different Modeling Problems

Planned and unplanned downtime require fundamentally different approaches in capacity modeling because they have different predictability profiles.

Planned downtime — preventive maintenance windows, scheduled changeovers, tool qualification runs, shift change overlap, planned trials — is deterministic. You know it's coming. You can put it on the calendar. It should be subtracted from gross available time before you calculate capacity.

If a CNC machining center gets a 4-hour preventive maintenance window every other Friday, that's 2 hours per week of planned non-production time. In a 40-hour shift week, gross available time is 40 hours. After planned maintenance, net available time is 38 hours. This subtraction happens before OEE is applied to the remaining 38 hours.

Unplanned downtime — equipment failures, emergency maintenance, material shortages causing machine starvation, unexpected quality problems requiring rework loops — is stochastic. You know it will happen, but not when, not how long, and not which machine. It cannot be scheduled away. It must be modeled as a probability distribution.

The correct approach for unplanned downtime is to use historical MTBF (mean time between failures) and MTTR (mean time to repair) data to build an availability distribution for each critical work center. Instead of a single availability number (say, 88%), model it as a distribution: 88% median, 95th percentile at 76%, 5th percentile at 97%. This distribution is what drives the OEE-adjusted capacity range rather than a point estimate.

For scheduling purposes, this means acknowledging that your capacity plan is probabilistic, not deterministic. You have a high probability of achieving 115 hours of output on a machine rated at 160 hours. You have a low probability of achieving 135 hours and a low probability of delivering only 90 hours. Planning to the median (115 hours) is correct. Planning to the optimistic tail is how you create chronic schedule slippage.

Monte Carlo Thinking for Stochastic Downtime

Sophisticated capacity planning at facilities with high downtime variability uses Monte Carlo simulation to model the probability distribution of total available capacity across the planning horizon.

Monte Carlo simulation runs thousands of random scenarios using the probability distributions for each machine's availability, performance, and quality. Each scenario draws a random downtime event profile consistent with historical MTBF/MTTR patterns and calculates total capacity under that scenario. After thousands of runs, you have a capacity distribution: P10 (10% probability of being below this capacity), P50 (median), P90 (90% probability of being below this capacity).

For a practical scheduling rule, plan to P50 capacity (the median). Carry a capacity buffer that covers the P10-P50 gap for critical customer commitments. Accept that 10% of weeks will see capacity fall below your plan and have a contingency protocol — overtime authorization, subcontracting, expedited outsourcing — that deploys when the week's actual downtime exceeds plan.

Most job shops don't run formal Monte Carlo simulations. The practical equivalent is a rule of thumb based on historical downtime patterns: reduce your OEE-adjusted capacity by an additional 5-10% as a planning buffer. A machine with demonstrated OEE of 72% (115 hours on 160 rated) gets scheduled to 105-110 hours to account for the stochastic tail of downtime variability. This is empirically effective for shops without statistical analysis infrastructure.

The Cost of Capacity Planning Without Downtime Modeling

The business impact of ignoring OEE in capacity planning is measurable across three dimensions:

On-time delivery loss. A shop scheduling to rated capacity while operating at 65% OEE is chronically overcommitted by 35%. Every job released under that plan has a structural probability of lateness that is disconnected from anything the production team does well or poorly. On-time delivery performance in these environments is typically 60-75% — not because operations is performing poorly, but because the plan never matched reality.

Expediting cost. When the plan overcommits capacity, expediting is the operational response. Expediting has direct costs (overtime, premium freight, emergency outsourcing) and indirect costs (schedule disruption for non-expedited jobs, operator fatigue, quality risk from rushing). In shops without OEE-adjusted capacity planning, expediting costs typically run 8-15% of direct labor cost annually.

Customer relationship damage. Systematic schedule slippage erodes customer confidence faster than almost any other operational failure because it's the most visible. A job that's consistently 2-3 days late on a 10-day lead time is a 20-30% service level failure, and customers notice. In defense and aerospace — industries User Solutions has served for 35+ years — late delivery triggers formal corrective action plans that consume significant management bandwidth.

Building an OEE-Adjusted Capacity Model: Step by Step

A practical OEE-adjusted capacity model for a 15-machine job shop can be built in 4 steps:

Step 1: Collect 90-day trailing OEE by machine. Gather actual run hours, downtime hours, speed-loss hours, and quality yield data from your ERP or shop floor system. Calculate OEE = Availability × Performance × Quality for each machine. If your system doesn't track this automatically, use manual job traveler data — it's slower but the data is available.

Step 2: Separate planned downtime from OEE availability losses. Planned maintenance and changeover time should be excluded from the gross available time before OEE is applied, not absorbed into the availability factor. Recalculate availability only for unplanned downtime events.

Step 3: Calculate demonstrated capacity per work center. Demonstrated Capacity = Net Available Time × OEE. Net Available Time = Gross Shift Hours − Planned Downtime. This is the number that goes into your capacity planning model.

Step 4: Set capacity buffers based on downtime variability. Calculate the coefficient of variation in weekly downtime for each critical work center over the 90-day period. High-variability work centers (CV > 0.4) warrant larger planning buffers (10-15% below demonstrated capacity). Low-variability work centers (CV < 0.2) can be scheduled closer to demonstrated capacity (3-5% buffer).

How RMDB Integrates OEE Data into Capacity Plans

RMDB allows capacity planners to configure OEE factors at the work center level, separate from shift availability calendars. The separation is important: shift availability (planned downtime, shift patterns, holidays) is deterministic and goes into the work center calendar. OEE factors (availability efficiency, performance efficiency, quality yield) are applied as multipliers to the resulting available time.

When a planning run executes, RMDB applies the OEE-adjusted capacity to each work center before assigning operations to time slots. Jobs are not scheduled to time that doesn't exist in the OEE-adjusted capacity model. The result is a plan that reflects what machines can actually deliver — not what they would deliver if they ran at rated speed with zero downtime.

RMDB also integrates with EDGEBI for real-time OEE dashboards by work center. When actual OEE deviates from the planned factor — a machine has a worse-than-normal week for breakdowns — EDGEBI flags the variance and planners can recalculate available capacity mid-period rather than discovering the shortfall at the end of the week when the damage is already done.

For an overview of how OEE connects to Theory of Constraints scheduling, see our post on Theory of Constraints scheduling.

Your capacity plan is only as good as the OEE data behind it. Contact User Solutions to see how RMDB builds OEE-adjusted capacity models that match what your machines can actually deliver. Trusted by GE, Cummins, BAE Systems for 35+ years of finite capacity scheduling.