Capacity Planning for Small Job Shops: How to Calculate Available Hours by Work Center and Stop Overloading Your Bottleneck

Many small job shops do not have a sales problem. They have a capacity visibility problem. Work keeps coming in, quotes go out fast, and the schedule looks full, but nobody can say with confidence whether the laser, brake, mill, weld cell, or paint booth actually has enough time next week to absorb another order. The result is familiar: one work center becomes the choke point, jobs stack up in queue, due dates slip, and expediting becomes a daily routine. For a complete overview, see our manufacturing execution system software guide.

The good news is that you do not need advanced APS software or a full-time industrial engineer to fix this. A simple work-center capacity model, built from routing times, shift calendars, expected downtime, and queue data, is enough to make much better decisions. When you know how many hours are truly available at each work center, you can quote more accurately, set realistic schedules, and stop overloading the bottleneck that is driving chronic late jobs.

Why capacity planning fails in small job shops

In many shops, available time is estimated too loosely. Owners may assume a machine has 40 hours available because it runs one shift, five days a week. But that ignores setups, breaks, meetings, preventive maintenance, material delays, changeovers, scrap, rework, and unplanned downtime. It also ignores the jobs already waiting in line.

That gap between theoretical hours and usable hours is where late jobs are born. Capacity planning gets distorted when:

• Routing times are outdated or incomplete.
• Setup time is buried inside run rates or not tracked at all.
• Shift calendars do not reflect vacations, holidays, or planned maintenance.
• Downtime is treated as random instead of measured by work center.
• Queue is invisible, so new orders are promised against time that is already consumed.

If you already struggle with rush-order prioritization, this article pairs well with our guide to production scheduling for small job shops.

The simple work-center capacity model

Your model only needs four inputs for each work center:

1. Gross available hours from the shift calendar.
2. Planned losses such as breaks, meetings, and preventive maintenance.
3. Expected unplanned losses based on actual downtime history.
4. Load hours from jobs already queued and jobs you plan to release.

The basic logic is simple:

Net available hours = Gross scheduled hours - planned losses - expected unplanned losses

Remaining capacity = Net available hours - current and planned load

Build this by work center, not just by department. “Machining” is usually too broad. Separate CNC mill, CNC lathe, manual mill, saw, deburr, inspection, welding, and packing if they behave differently and have different queues.

Step 1: Define your work centers clearly

Start with the places where work waits. Those are usually the true planning points. For a small fabrication or machine shop, your work centers might be:

• Saw
• Laser
• Press brake
• CNC mill
• CNC lathe
• Welding
• Paint
• Final inspection
• Packing and shipping

Do not overcomplicate this. If two machines are interchangeable and staffed the same way, they can be grouped into one work center. If one machine is specialized or often becomes a queue point, give it its own line in the model.

What to include in a work center record

• Work center name
• Number of machines or stations
• Standard operators required
• Shift pattern
• Days worked per week
• Typical setup and run characteristics
• Downtime history
• Current queue in hours

Step 2: Calculate gross available hours from the shift calendar

Begin with the calendar, not with the order book. Ask: how many hours could this work center run during the planning period if nothing went wrong?

For each work center:

1. Count the number of machines or stations.
2. Multiply by shift length.
3. Multiply by number of shifts.
4. Multiply by working days in the period.

Example for one CNC mill work center with 2 machines, one 8-hour shift, 5 days:

Gross available hours = 2 machines × 8 hours × 5 days = 80 hours

If one machine is offline for a full day due to a planned service, adjust the calendar before you go further.

This is also where operator constraints matter. If you have two machines but only one qualified operator on second shift, your capacity is limited by labor, not equipment. Capacity should reflect the true limiting resource.

Step 3: Subtract planned losses

Now remove time you already know will not be available. Planned losses usually include:

• Breaks and lunch if machines do not run through them
• Daily startup meetings
• Cleaning and end-of-shift housekeeping
• Preventive maintenance
• Scheduled calibration or inspection activities
• Training time
• Known absenteeism or vacations if no coverage exists

Example:

The same CNC mill work center has 80 gross hours for the week. You know there are 5 total hours of breaks not covered, 2 hours of team meetings, and 4 hours of preventive maintenance.

Planned losses = 5 + 2 + 4 = 11 hours

Available after planned losses = 80 - 11 = 69 hours

Many shops skip this step because it feels obvious. But if you do not account for it explicitly, you will keep promising work into time that does not exist. If maintenance is one of your biggest planning blind spots, see our preventive maintenance scheduling guide.

Step 4: Estimate expected unplanned downtime from real history

This is the part that turns a rough estimate into a practical model. Every work center loses time to setup issues, tool breakage, material shortages, waiting on first-article approval, machine faults, and other disruptions. You do not need perfect forecasting. You need a realistic average based on recent history.

Look back over the last 6 to 12 weeks by work center and total the downtime hours. If you track downtime by reason code, even better. A basic system is often enough to reveal hidden losses; see how to use a simple downtime code system.

For example, if your press brake averaged 6 hours of unplanned downtime per week over the last 8 weeks, use 6 hours as the expected unplanned loss for next week unless conditions have changed materially.

Net available hours = Gross hours - planned losses - expected unplanned downtime

Using the CNC mill example:

Gross hours = 80

Planned losses = 11

Expected unplanned downtime = 6

Net available hours = 63

If you do not yet track downtime, start now. A simple sheet or digital log will do. For reference on the broader concept of equipment effectiveness, see Wikipedia’s overview of Overall Equipment Effectiveness. You can also estimate losses with FactoryOS’s free downtime cost calculator.

Step 5: Convert routings into load hours

Once you know net available hours, compare them to the load created by jobs in queue and jobs you are considering releasing or quoting.

For each open job, pull the routing steps that touch each work center. Then calculate load hours:

Load hours = setup time + (run time per piece × quantity) + expected rework or inspection time if applicable

Example job at the press brake:

• Setup: 1.5 hours
• Run time: 3 minutes per part
• Quantity: 120 parts

Run time in hours = 3 × 120 = 360 minutes = 6 hours

Total load on press brake = 1.5 + 6 = 7.5 hours

Repeat for every released job and every quote under serious consideration if you are checking future promise dates.

Common routing mistakes that distort capacity

• Setup time missing from the router
• Run rates based on best-case instead of normal conditions
• Inspection or first-article time omitted
• Outside processing not reflected in queue timing
• Batch sizes different from how the shop actually runs
• Rework loops ignored

If quality losses regularly consume hidden hours, that should be visible in your planning. Our article on first-pass yield by work order explains how to measure that loss in a practical way.

Step 6: Add queue data so you stop double-booking the bottleneck

This is where many small shops go wrong. They calculate the hours for the new job but ignore the work already waiting. Capacity is not just what is on today’s machine. It is what is already committed to the work center.

For each work center, maintain a simple queue bucket:

• Released jobs waiting to start
• Jobs currently running but not yet complete
• Planned jobs expected to be released in the period
• High-probability quotes if you use capacity checks during quoting

Then total the load hours by week.

Simple weekly capacity table

In this example, the press brake is overloaded by 6 hours. That does not sound huge, but if this happens every week, backlog compounds quickly. The brake becomes the bottleneck, upstream work keeps feeding it, and due dates become fiction.

For a solid explanation of bottlenecks and constraints in production systems, the National Institute of Standards and Technology is a credible resource for small manufacturers, especially through its MEP network and manufacturing guidance.

Step 7: Use the model for quoting, scheduling, and release decisions

For quoting

Before you promise a date, check the future capacity buckets for the relevant work centers. If the bottleneck week is already full, either:

• Quote a later ship date
• Split the order into partial deliveries
• Subcontract a constrained step
• Run approved overtime if margins justify it
• Sequence lower-priority work out of the period

This makes quoting more honest and more profitable. The cheapest quote is not the one that wins the order and destroys the schedule.

For scheduling

Use the bottleneck as the anchor. Schedule around the constrained work center first, then align upstream and downstream operations to support it. If you release more work to non-bottleneck areas than the bottleneck can absorb, you only increase queue and confusion.

If your current system relies heavily on paper packets and tribal knowledge, replacing paper work orders can make capacity data far easier to keep current.

For daily release control

Not every job that can be started should be started. Release work based on bottleneck capacity, material readiness, tooling readiness, and due date priority. A shorter queue is usually easier to manage than a huge pile of half-started jobs.

How to identify the real bottleneck

The bottleneck is not always the most expensive machine. It is the work center with the least available capacity relative to demand. In your model, it will usually show up as one or more of the following:

• Consistently negative remaining capacity
• Longest queue in hours or days
• Highest schedule volatility
• Most frequent expediting activity
• Late jobs clustering after that step

Do not assume the bottleneck is fixed forever. It can shift by product mix, staffing, season, or maintenance condition. Review the model every week.

What to do when a work center is overloaded

Once overload is visible, you have choices. None are perfect, but all are better than pretending the hours exist.

• Improve standards: Clean up routing times, especially setups.
• Reduce downtime: Attack the biggest recurring causes first.
• Run selective overtime: Use it surgically, not as a permanent crutch.
• Move work: Shift to alternate machines or qualified operators.
• Split batches: Smaller transfer batches can reduce downstream waiting.
• Subcontract strategically: Reserve this for constrained or low-margin work.
• Change quote dates: Protect due-date performance instead of overpromising.
• Elevate the constraint: Add tooling, fixturing, staffing, or another machine only after confirming the true cause.

Often the fastest gains come from reducing hidden losses rather than buying more equipment. Setup delays, missing tools, poor traveler instructions, and waiting for approvals can consume enough hours to mimic a capacity shortage.

A simple weekly review rhythm

You do not need a long planning meeting. A disciplined 20- to 30-minute weekly review can be enough. For each work center:

1. Confirm next week’s shift calendar.
2. Subtract planned losses.
3. Review recent downtime and estimate expected losses.
4. Total queued and released job hours.
5. Compare load to net available hours.
6. Flag overloaded work centers.
7. Adjust quote dates, release timing, outsourcing, or overtime.

Make one person accountable for keeping the numbers current. Capacity planning fails when everyone assumes someone else updated the queue.

Keep the model simple enough to maintain

The best capacity model is the one your shop will actually use every week. Start with a spreadsheet if necessary. Track by work center and by week. Use rough-but-honest standards first, then improve accuracy over time.

You do not need perfect data to get value. If your current promise dates are based on instinct alone, even a basic model will improve quoting discipline and scheduling realism. Over time, digital work orders, downtime tracking, and better routing data will make the model stronger and easier to maintain. That is exactly the kind of operational visibility we built FactoryOS to support.

Capacity planning is really about telling the truth early: how many usable hours each work center has, where the queue is building, and which orders your bottleneck can realistically absorb. When you know that, late jobs stop being mysterious. They become manageable.

If you want a simpler way to track work centers, routing times, queues, and schedule risk in one place, start a free trial of FactoryOS.