← Back to Field Notes

June 9, 2025 · The Bloomfield Team

Manufacturing Capacity Planning: A Practical Guide for Job Shops

A 45-person machine shop in Michigan quoted a $280,000 aerospace program last year. Won the bid. Then missed the first delivery by three weeks because their 5-axis department was already running at 96% utilization when the PO landed. Nobody checked. The data to check existed in their ERP. Nobody pulled it together in a way that made the constraint visible before the commitment was made.

Capacity planning in job shops is hard because the work mix changes constantly. A production facility making 10,000 of the same bracket can plan capacity with a calendar and a calculator. A job shop running 40 different part numbers across 15 machines in a given week needs a more dynamic approach.

The Basic Formula

Available capacity for any machine group equals the number of machines multiplied by available hours per shift, multiplied by the number of shifts, multiplied by an efficiency factor that accounts for setup time, maintenance, and unplanned downtime.

For a shop running two Mazak 5-axis mills on a single 10-hour shift, five days per week, with an 82% efficiency factor (typical for high-mix job shops), the math looks like this: 2 machines x 10 hours x 5 days x 0.82 = 82 available spindle hours per week.

The efficiency factor is where most shops get the math wrong. They assume 90% or higher because that is what the machines are capable of in theory. In practice, high-mix environments with frequent changeovers, first-article inspections, and programming time at the machine run between 75% and 85%. Pull your actual data from the last six months and calculate the real number. It will be lower than you think.

Demand Loading

Once you know available capacity, the next step is loading demand against it. For every open order and confirmed quote, map the estimated hours by machine group and by week. This produces a load chart that shows where demand exceeds capacity and where capacity sits idle.

Most ERP systems can generate a basic load report. The problem is that the data feeding those reports is only as accurate as the estimated setup and cycle times in the router. If your routers were built five years ago and never updated with actual floor data, the load chart will be wrong by 15 to 30%. This is where connecting shop floor data back into your planning system produces immediate improvements in planning accuracy.

The 85% Rule

Run a machine group above 85% utilization for more than three consecutive weeks and lead times will start expanding. This is a well-documented principle in queuing theory, and it holds true on every shop floor we have worked with. At 85% utilization, any disruption (a quality hold, a machine down for maintenance, a rush order from your best customer) pushes jobs back because there is no buffer.

The practical implication: when you are quoting new work, check the load on your constraint machine group for the delivery window. If it is already above 80%, either adjust the delivery date, plan overtime, or decline the quote. Making that decision at the quoting stage prevents the cascading late shipments that erode customer confidence.

Rolling Horizon Planning

Static capacity plans fail in job shops because the work mix changes weekly. A rolling horizon approach works better. Each Monday, update the load chart with new orders received, completed jobs removed, and any schedule changes from the prior week. Look out four to six weeks. Beyond six weeks, the data is too uncertain to plan against with precision in a high-mix environment.

The weekly update should answer three questions. Which machine groups are overloaded in the next two weeks, and what is the plan to address it (overtime, subcontracting, or rescheduling)? Which machine groups have available capacity that could absorb new work? What is the earliest realistic delivery date you can commit to for a new order?

Where AI Fits

The manual version of this process works for a shop running 20 to 30 open jobs. Above that volume, the number of variables exceeds what a human can track in a spreadsheet. Machine availability, operator skill matrix, tooling constraints, material lead times, quality hold risks based on historical data for similar parts. Each variable interacts with every other variable.

AI-powered capacity planning pulls data from the ERP (open orders, routers, job history), from the shop floor (actual cycle times, machine status), and from the quoting system (pending quotes with probability-weighted demand), then produces a load forecast that updates continuously. The scheduler still makes every decision. They make those decisions with a complete picture of demand and capacity that would take hours to assemble manually.

For a broader view of how these tools connect into your existing systems, see our guide to production visibility in manufacturing.

Start This Week

Pull open orders from your ERP. Map estimated hours by machine group for the next four weeks. Calculate available capacity using the formula above with a realistic efficiency factor. Compare load to capacity by machine group. The first time you run this analysis, you will find at least one machine group that is overloaded and at least one that has more capacity than you thought. Both of those findings are actionable today.