"Yeah, we can do that by Friday"

Those are the most expensive words spoken in a job shop. Someone — the owner, a salesperson, a customer service rep — says them on a Thursday afternoon, the customer hangs up happy, and the commitment lands on the floor like a dropped fixture. Now comes the question nobody asked before the promise was made: can the shop actually keep it?

Most shops find out the answer the hard way. The job gets slotted in, something else gets bumped, a setup runs long, and by the following Wednesday you're resequencing three machines to protect one due date you never should have quoted. A scheduling conflict that reaches the floor costs $250–$1,000 per incident in machine restart, resequencing, and lost capacity (Product Brief §2). String enough of those together and you've quietly handed back a chunk of your margin to fix problems you created at the moment of the quote.

Machine capacity planning is the discipline that answers the question before you commit, not after. It isn't scheduling — scheduling is the exact sequence of operations once the work is already yours. It's the layer above that: knowing how much real productive capacity you have, how much of it is already spoken for, and how much is left to promise against.

This is a guide to building that picture for a real job shop. How to calculate available capacity without kidding yourself, why your ERP keeps saying yes when the honest answer is no, and how to turn a capacity number into a due date you can actually keep.

Capacity planning, scheduling, and utilization are three different questions

These three terms get used interchangeably on the floor, and the confusion costs shops real money. They answer different questions, and they point in different directions in time.

Utilization looks backward. It tells you how well you used the capacity you already had — actual run hours against available hours, last week or last month. It's a report card. Overall Equipment Effectiveness (OEE) — Availability × Performance × Quality — is the refined version of the same backward-looking question, and the world-class benchmark for it sits at 85% (Nakajima/TPM literature). Utilization is worth tracking, but it tells you nothing about whether you can take the order that's on the phone right now.

Scheduling looks at the present. Given the work you've already committed to, what runs on which machine, in what order, starting when. It's the sequence of what's already yours.

Capacity planning looks forward. Before any of that, it answers a different question: do I have room? Can I commit to this order and hit the date without breaking the dates I've already promised? That is the question that protects your due dates, and it's the one most shops never formally answer.

Here's the trap that confusing them creates. A shop with great utilization numbers can still be incapable of taking on the next order, because utilization describes the past and capacity describes what's left. And a shop staring at three idle machines can be genuinely full, because the idle machines aren't the constraint. That second case is the most common capacity misread in a job shop, so it's worth slowing down on.

"We have machines sitting idle" is the wrong way to read capacity

Walk almost any shop floor and you'll see machines not running. The instinct is to read that as free capacity: we've got three idle mills, so we can take more milling work. That instinct is usually wrong, and it's wrong for a specific, structural reason.

A job's path through your shop is gated by its slowest, most contested resource — the bottleneck. If every job in the queue needs the wire EDM, and the wire EDM is booked solid, it does not matter that three mills are idle. Your shop's real capacity is the capacity of the constraint, not the sum of all machine-hours on the floor.

This is why "total machine-hours" is a vanity number. Add up every machine across every shift and you get an impressive figure that has almost nothing to do with how much work you can actually promise. The number that matters is the capacity at the constraint, adjusted for everything that eats into it.

And plenty eats into it. Start with the calendar math, which is more sobering than most owners expect. A single machine on two shifts, Monday through Friday, runs at most 80 hours against the 168 hours in a week. Before you account for a single breakdown, changeover, or absent operator, a two-shift operation has structurally capped itself below 50% of calendar time. That's not a performance problem — that's arithmetic. Honest shop floor capacity planning starts by accepting that ceiling instead of quoting against a fantasy of round-the-clock availability that doesn't exist.

Machine capacity planning in three numbers

Forward capacity planning comes down to three numbers, tracked per work center or per constraint machine:

• Available capacity — the productive hours you can realistically count on in a given window.
• Committed load — the hours already consumed by work you've promised.
• Remaining capacity — what's left to promise against. Available minus committed.

That's the entire model. The difficulty is never the subtraction; it's getting available honest. Most shops that miss due dates aren't bad at arithmetic — they're optimistic about the first number. They quote against capacity they don't actually have, then spend the week absorbing the difference in overtime, expedite fees, and apologetic phone calls.

How to calculate available capacity without kidding yourself

Start with gross calendar hours, then subtract reality in layers. Take a single work center of 10 machines running two 8-hour shifts, five days a week:

10 machines × 2 shifts × 8 hours × 5 days = 800 gross machine-hours per week.

That's the fantasy number — the one an infinite-capacity calculation quietly assumes you have in full. Now subtract the things that are always true on a real floor:

• Planned maintenance and PMs. Time the machine is down on purpose. You know roughly how much; subtract it honestly rather than pretending it's zero.
• Setup and changeover. In a high-mix job shop, this is often the single largest bite out of available capacity. Sequence-dependent setups make it worse — the changeover time depends on what part ran immediately before.
• Unplanned downtime. Breakdowns happen, and they compound: unplanned downtime runs about 35% more expensive than planned downtime (Arda Cards 2026), and every unplanned hour is capacity you'd already sold to someone.
• Operator availability. A machine with no qualified operator on shift is a zero-hour machine for that shift. If you're short an operator or running thin on cross-trained people, your machine-hours and your operator-hours diverge fast — which is why operator and shift coverage belongs inside the capacity calculation, not beside it.
• Scrap and rework. Hours that produced nothing sellable still consumed capacity. They count against you.

Apply your own real numbers to each line — they'll differ by shop, by machine, and by season, and the entire point is to use yours rather than an industry average that doesn't describe your floor. Suppose, after those subtractions, your 800 gross hours come down to 520 genuinely available hours. Then 520 is your capacity number, and 520 is what you quote against. Quote against 800 and you'll miss dates structurally, every single week, and never quite understand why the floor is always behind.

This is also exactly where manual scheduling quietly bleeds money. When available capacity lives in a manager's head or a spreadsheet that's already a day stale, every quote is a guess, and the guesses are biased optimistic because nobody wants to tell a customer no. The cost of that guessing is measurable: manual scheduling inefficiency runs an estimated 5–10% of revenue in a typical job shop, which for a $2M shop works out to roughly $128,000–$276,000 a year (Qlector 2025).

Rough-cut versus detailed: pick the right altitude

Not every capacity question needs the same precision, and treating them as if they do is its own kind of waste. Manufacturing capacity planning generally operates at two altitudes, and knowing which one you're answering keeps you from over-engineering a quick yes/no or under-thinking a major commitment.

Rough-cut capacity planning is the fast altitude. It answers the strategic question: across the next quarter, do we have the broad capacity to take on this new account, this seasonal surge, this second large customer? It works at the level of work centers and weeks, not individual machines and hours. You're checking whether the shape of the demand fits the shape of the shop. It's the right tool when an owner is deciding whether to chase a contract, and the wrong tool when a customer wants a date for one specific part.

Detailed capacity planning is the precise altitude. It answers the question on the Thursday phone call: can this specific job hit this specific date on this specific constrained machine, given what's already loaded? This is machine-level and day-level, and it's the altitude where due dates are actually won or lost.

Most shops have a rough intuition for the strategic question and almost no real system for the detailed one. That's backwards. The strategic question gets revisited a few times a year and tolerates being approximate. The detailed question gets answered dozens of times a week, every answer becomes a promise, and approximate promises are how a shop earns a reputation for being late. Good machine capacity planning means running both altitudes deliberately, and not mistaking a confident answer at one for a real answer at the other.

Why your ERP says "yes" when the honest answer is "no"

Here's the part that catches shops by surprise. Most ERP and MRP scheduling logic is infinite-capacity by default. You enter a due date, the system back-schedules from that date using standard run times, and it cheerfully tells you the job fits — because it never checked whether the machine was already full during that window. It assumes capacity is unlimited. It is answering "can this be back-scheduled cleanly from an empty calendar," not "is the machine actually free."

That's the gap between an MRP module and real finite-capacity scheduling. Finite-capacity logic does the thing the infinite model skips: it checks the actual committed load on each machine across the planning window and refuses to double-book the same hours. It's the difference between a system that confirms your wish and a system that tells you the truth — and the truth is what a customer is actually buying when they ask for a date.

You don't necessarily have to rip out your ERP to close that gap. But you do need a capacity layer — visual or otherwise — that models finite constraints, so that when something tells you a date is achievable, it means the constrained machine is genuinely open during that window, not merely that the arithmetic back-scheduled without complaint.

The capacity killers that don't show up in the run-time field

Standard run time is the number every system tracks. The capacity that vanishes lives in the numbers most systems don't.

Setup and changeover, especially sequence-dependent. Running five different parts on one machine isn't five run times — it's five run times plus four changeovers. And if those changeovers are sequence-dependent, your effective capacity swings based purely on the order you run things. Good sequencing recovers hours you didn't know you had; bad sequencing burns them, and nobody sees the loss because it never appears as a line item.

Rush orders. Every expedited job is capacity borrowed from jobs already on the books. A rush order doesn't add capacity, it reallocates it — and if you're saying yes to rush work without seeing exactly what it bumps, you're quietly manufacturing the next batch of late deliveries while feeling responsive in the moment.

A moving constraint. In a high-mix shop, the bottleneck doesn't stay put. This week it's the EDM; next week a big production run shifts it to the surface grinders; the week after, a tooling delay parks it at inspection. Capacity planning that assumes a fixed bottleneck will be wrong roughly half the time. You have to watch where the constraint actually sits for the current mix of work, because that's the machine whose capacity is the only one that matters.

None of these show up if you plan capacity off total machine-hours and standard run times. All of them show up the moment a due date slips — at which point it's a fire, not a plan.

Turning a capacity number into a promise you can keep

The payoff for all of this is a single capability: being able to look at an incoming order and answer, honestly and fast, yes by this date — or no, but I can hit this date instead.

That's the available-to-promise question, and answering it well is what separates shops that quote confidently from shops that quote hopefully. With a real capacity picture in hand — available hours per constraint, committed load already mapped against the calendar, remaining capacity visible at a glance — the Thursday-afternoon phone call stops being a gamble. You can see whether Friday is real before you say the words.

This is the practical case for planning capacity visually. A drag-and-drop schedule that shows each machine's committed load against its genuinely available hours turns capacity from a number you recalculate once a month into a picture you can read in the second a customer asks for a date. When a new job lands on the board, you see what it touches, what it bumps, and whether the date holds — before you commit, not after the commitment has already started costing you.

That's the entire point of capacity planning for a job shop: moving the question from after the promise to before it.

Where to start

If your shop currently quotes dates off gut feel and a whiteboard, the first move isn't software — it's the three numbers. Pick your constraint machine, calculate its genuinely available hours for the next two weeks, map your committed load against it, and look hard at what's actually left. Do that by hand once and you'll see your real remaining capacity for the first time, probably lower than you assumed.

You don't need a perfect model to start. A single sheet for one constraint machine — available hours this week, hours already committed, hours remaining — is more machine capacity planning than most shops have ever done, and it will change which jobs you say yes to by Friday. The discipline matters more than the tooling at this stage.

Then, when you're tired of recalculating it by hand every time the mix shifts or a machine goes down, that's where a visual scheduling tool earns its keep — it keeps the capacity picture live instead of stale.

Want to see your shop's real capacity against your committed load, on a board you can actually drag? Start a free trial of Visual Machine Scheduler. No credit card required, 14-day trial. Bring your real jobs and quote your next due date against the truth instead of a guess.