Capacity Utilization and Throughput Analysis

Executive summary

Capacity utilization and throughput analysis answers two questions that plant tours and monthly reports tend to blur. How much of what the plant could make is it actually making? And which single resource decides how much the whole plant can make? The second question matters more. A factory behaves like a chain, and the output of the whole line is set by its slowest or most overloaded step, the constraint. An hour lost at the constraint is an hour lost for the entire plant. An hour lost at a step with spare capacity costs very little.

That changes where improvement money should go. A company can raise utilization at every machine and ship the same number of units, because the extra output at non-constraint steps just piles up as work in process. A company that finds the constraint and protects it, by removing downtime, scrap, and starved or blocked time, ships more with the same equipment.

This guide gives you a working method. Define capacity and utilization in a way that cannot be gamed. Find the constraint with data from your own plant. Compute overall equipment effectiveness (OEE) from availability, performance, and quality. Put a dollar value on an hour at the constraint so improvement ideas can be ranked by money rather than by volume of complaints. Then decide whether to fix, offload, or add capacity, and what it costs. All formulas use your own data, and the worked example is labelled illustrative.

When to use this

Use this guide when:

• The plant reports high utilization but cannot ship more or ship on time.
• Overtime and expediting are routine, yet some machines sit idle.
• A request for new capital equipment is on the table and you need to know whether the constraint justifies it.
• Lead times are stretching, and work in process is building ahead of one or two operations.
• An acquirer, lender, or sponsor is asking whether the plant can support a volume plan.
• Your sales and operations plan shows a supply gap and you need to size the real constraint. See S&OP process setup.

If you do not record downtime, scrap, and run counts by work center, begin by capturing a few weeks of that data, even on paper. The analysis below depends on it.

Diagnostic questions

1. Which work center has the longest queue of waiting work, week after week?
2. Which resource, when it stops, causes the whole line to stop within an hour or two?
3. How is capacity defined at each work center: nameplate, demonstrated, or planned? Does anyone know which one is used in the reports?
4. What is the planned production time per shift, after breaks, meetings, and scheduled maintenance?
5. What are the top five downtime reasons at the suspected constraint, by minutes lost over the last month?
6. What is the actual cycle time at the constraint compared to the standard or ideal cycle time?
7. What share of units produced at the constraint are scrapped or reworked?
8. How long does the constraint wait for material, tooling, labor, or a changeover? Is the constraint ever starved by the step before it?
9. Does the schedule load the constraint first and build everything else around it, or is every work center scheduled independently?
10. What is the contribution margin per unit that passes through the constraint?
11. Which demand do you turn away or ship late because of the constraint, and what is it worth?

The framework: measure, locate, value, decide

Define capacity and utilization so they cannot be gamed

Capacity needs three definitions in the report, or people will argue past each other.

• Theoretical capacity: output at the ideal cycle time for every calendar hour. Useful only as an upper bound.
• Planned capacity: output at the ideal cycle time during planned production time, after scheduled breaks and planned maintenance.
• Demonstrated capacity: the best sustained output the work center has actually achieved over a period you name, such as its best four consecutive weeks.

Capacity utilization = Actual output ÷ Capacity (state which definition)
Load percentage = Scheduled hours ÷ Available hours at the work center

Utilization against theoretical capacity makes every plant look poor. Utilization against planned capacity is the useful number for scheduling. Demonstrated capacity is the best anchor for a target, because it is something your plant has already done.

A high utilization number is not a goal in itself. A non-constraint work center at 95 percent utilization is probably producing inventory nobody needs yet. A constraint at 95 percent is a good place to be. Report utilization by work center and mark which one is the constraint.

OEE: availability, performance, quality

OEE separates the losses at a work center into three types that have different owners and different fixes.

Availability = Run time ÷ Planned production time
Performance  = (Ideal cycle time × Total units produced) ÷ Run time
Quality      = Good units ÷ Total units produced
OEE          = Availability × Performance × Quality
            = (Good units × Ideal cycle time) ÷ Planned production time

Availability loses time to breakdowns, changeovers, and waiting. Performance loses time to slow cycles and minor stops. Quality loses time to units that are scrapped or reworked. The last form of the formula is a useful cross-check: good units times ideal cycle time, divided by planned time, must give the same answer as the product of the three.

Use OEE at the constraint first. OEE at a non-constraint work center tells you about that machine's efficiency, but not about the plant's output.

Throughput accounting in one paragraph

Throughput accounting asks what each hour at the constraint is worth. Throughput per unit equals selling price minus truly variable cost (materials and other costs that vary with each unit, such as a per-unit freight or commission). Throughput per constraint hour equals units per constraint hour times throughput per unit. Operating expense (labor, rent, depreciation) is treated as fixed in the short run. This lens is deliberately blunt. It stops you from approving a project that saves labor at a non-constraint, which saves no money unless you also remove the cost, and it highlights projects that add output at the constraint, which add margin immediately if the demand is there. Check the result with your finance team and your costing policy, because the throughput view is a decision aid and not a replacement for your accounting.

Step-by-step process

Weeks 1-2: Establish the data

1. List the work centers in the routing for your top three product families. Owner: Head of Operations. Output: a flow map with one line per step.
2. Define planned production time per shift and per week for each work center, after breaks and scheduled maintenance. Owner: Plant manager. Output: a planned time table.
3. Record ideal cycle times from engineering standards, or from the best demonstrated cycle at each work center if there is no standard. Owner: Industrial engineering or the lead supervisor. Output: an ideal cycle time per product.
4. Start a downtime and scrap log at each suspected constraint, with reasons from a short, fixed list. Owner: Shift supervisors. Output: four weeks of reason-coded minutes and units.

Weeks 3-4: Locate the constraint

5. Compute load versus capacity by work center for the coming four to eight weeks, using the schedule and planned capacity. Owner: Planner. Output: a load table.
6. Collect the physical evidence. Where does work wait? Which downstream step is starved when one stops? Walk the floor at three different times of day. Owner: Plant manager with the planner. Output: a one-page note naming the constraint and the evidence.
7. Check outside the plant. Test supplier lead times, shipping capacity, inspection queues, and engineering release times for the same queue pattern. Owner: Head of Operations. Output: a statement of whether the constraint is internal.

Weeks 5-6: Measure and value

8. Compute OEE at the constraint by shift and by product for the logged period. Show availability, performance, and quality separately. Owner: Planner or analyst. Output: an OEE loss tree.
9. Rank the losses by minutes (or unit-equivalents) at the constraint, using a Pareto of reasons. Owner: Plant manager. Output: a ranked loss list.
10. Value an hour at the constraint. Get unit price and variable cost from finance, units per hour from the log, and compute throughput per constraint hour. Owner: CFO or controller. Output: dollars per hour.
11. Convert each loss to dollars. Minutes lost per week, times throughput per minute, assuming there is demand to absorb the extra output. Owner: Controller. Output: a dollar-ranked loss list.

Weeks 7-12: Decide and act

12. Protect the constraint first. Cover breaks with relief operators, stage material and tooling before the machine needs them, inspect upstream so that the constraint does not process bad work, and keep a small buffer of work in front of it. Owner: Plant manager.
13. Attack the top loss. Typical tools are quick changeover work, preventive maintenance on the constraint, standard work for the operator, and error-proofing. Owner: The loss owner named in the ranked list.
14. Compare options in dollars per hour: fix the loss, offload some work to a non-constraint machine or a vendor, add a shift, or buy capacity. Owner: CFO with the Head of Operations. Output: a ranked option table.
15. Reschedule around the constraint. Load it first, release material to the rest of the line at the constraint's pace, and stop pushing work into non-constraint queues. Owner: Planner.

Templates

Work center capacity sheet (copy into a spreadsheet)

Downtime and loss log

OEE calculation sheet

Improvement decision table

Metrics and KPIs

Setting targets by method. The best anchors are your own: the demonstrated capacity from your best sustained four weeks, your best shift's OEE at the constraint, and the contractual delivery lead times you have promised customers. For an outside reference, use disclosed operating data from a named public peer or an equipment maker's stated rate for your machine, and cite it. Avoid taking a generic "world class" OEE number and imposing it on a plant that makes different products on different equipment. A plant that runs short batches of many parts will not look like a plant that runs one part around the clock.

Worked example (illustrative)

The figures are made up to show the method. They are not benchmarks.

Finding the constraint

A plant makes one product through four steps. Each step's capacity per hour:

The line can make no more than 80 units per hour, the rate of the weld step. Weld is the constraint. The utilization of the other steps, if the line runs at the pace of weld, is:

Pushing cut harder than 80 units per hour would not raise shipments. It would grow the pile of parts waiting for weld.

OEE at the weld station for one shift

The shift is 480 minutes. Breaks total 30 minutes, so planned production time is 450 minutes. Logged downtime is 45 minutes, so run time is 405 minutes. The ideal cycle time is 0.5 minutes per unit. The station produced 700 units, of which 665 were good.

Both routes agree. The loss tree shows where the 26.1 percent went: 10 percent of planned time to downtime, about 13.6 percent of run time to slow cycles or minor stops, and 5 percent of output to scrap.

Valuing an hour at the constraint

Assume the weld station runs at 80 units per hour when running well, selling price less variable cost is $35 per unit, and there is unmet demand. Throughput per constraint hour is 80 × $35 = $2,800. Recovering 15 minutes of downtime per shift is worth 0.25 × 80 × $35 = $700 per shift, as long as the extra units are sold and the downstream steps can handle them. They can, since their utilization is well below 100 percent at this rate.

Now compare two improvement ideas. A labor-saving fixture at the cut station saves two operator-hours per shift. Cut is not the constraint, so shipments do not change, and the saving is real only if the labor cost actually leaves the plant. A quick-changeover project at weld that recovers 15 minutes per shift is worth $700 per shift in throughput. Whichever is cheaper per dollar of benefit should lead, but the two should not be confused. The first is a cost saving that depends on a staffing decision. The second is added output.

Common mistakes

1. Maximizing utilization everywhere. It builds work in process and lengthens lead times. Keep non-constraint utilization at the level that serves the constraint.
2. Using nameplate capacity. A work center rated for a rate it has never achieved produces a plan that cannot be met. Use demonstrated capacity for scheduling.
3. Counting scrap as output. A unit that is scrapped at the constraint consumed constraint time and produced no throughput. Count good units.
4. Ignoring changeovers. At a constraint with many product changes, setup time is often the largest availability loss.

AI prompt assistance

AI can compute and tabulate. It cannot see your floor. Give it the data, and compare the result with what you can observe.

1. Compute OEE from a shift log.

I will paste, for each shift over four weeks at one work center: planned production time, downtime minutes by reason, ideal cycle time, total units, and good units. Compute availability, performance, quality, and OEE per shift and for the period. Show the formulas, and confirm each OEE using the cross-check (good units × ideal cycle time ÷ planned time).

Check: recompute one shift by hand and verify that the units match the production report, not an estimate.

2. Rank downtime reasons.

Here is a downtime log with minutes and reason codes. Produce a Pareto: total minutes by reason, share of total, and cumulative share. List the reasons that make up the first 80 percent. Flag entries with no reason or with a reason code that is not on this list: [list].

Check: ask the shift supervisors whether the top three reasons match what they experience.

3. Load versus capacity.

Here is a schedule of hours by work center by week and the planned hours available. Compute load percentage per work center per week and flag any week above 100 percent. Do not recommend changes yet.

Check: confirm that available hours exclude breaks and planned maintenance.

Implementation guide

Downloadable assets

The capacity sheet, loss log, OEE calculation sheet, and decision table above are designed to be copied into a spreadsheet. The OTD root cause calculator can help separate capacity-driven lateness from other causes.

Related resources

• Operations guides: the full operations library.
• S&OP process setup: testing demand against the constraint.
• Quality cost and scrap reduction: the quality component of OEE.
• Lean visual management for plants: making constraint status visible on the floor.
• Inventory optimization: turns, safety stock, and excess: the inventory side of work in process and buffers.
• Margin expansion playbook: where added throughput lands in margin.

Frequently asked questions

What is capacity utilization?

Capacity utilization is actual output divided by the capacity available, expressed as a percentage. The result depends on which capacity you use: theoretical, planned, or demonstrated. State the definition in every report. For scheduling, planned or demonstrated capacity is more useful than theoretical capacity, which makes every plant look underused.

How do you find the bottleneck in a plant?

Find the work center with the longest waiting queue in front of it, whose stoppages stop downstream steps within hours, and whose load is highest against its capacity. Confirm with the routing data and a few floor walks at different times. Also check outside the plant, because the constraint can sit with a supplier or in shipping.

How do you calculate OEE?

OEE equals availability times performance times quality. Availability is run time over planned production time. Performance is ideal cycle time times total units, over run time. Quality is good units over total units. As a cross-check, good units times ideal cycle time, divided by planned production time, must give the same answer.

Is higher utilization always better?

No. Higher utilization is valuable at the constraint, because every lost hour there is lost for the plant. At a non-constraint step, pushing utilization higher mostly builds work in process and lengthens lead times. Report utilization by work center and mark the constraint, so that effort goes to the step that sets output.

When should we buy more capacity?

Buy capacity when the constraint is already running near its demonstrated best, losses at the constraint are small, offloading work is not practical, and there is demand to absorb the output at a throughput value that covers the cost. If OEE at the constraint is far below your own best, fix the losses first.

How this works in a System of Performance

Throughput analysis fails when the downtime log, the schedule, the quality record, and the finance valuation live in four places and meet only in a slide. A Performis routine can take on a recurring job like this, with the scope and fit agreed with you: it works from the records your team supplies, prepares the work for your named reviewer, and keeps each approval on record. If you want to see where your own plant's output is being decided, bring it to us as a launch partner.