If you read last week’s MRP post, you got an answer to "what do we need to buy and build, and by when?" That answer assumed your factory could actually do the work in the time available. For most SMEs, on most days, this assumption quietly fails. MRP says "machine the casing assembly by 8 July." Your one CNC mill, already booked for two other orders, will be free on 12 July. The slip happens. The dispatch date moves. The customer is unhappy.
This is the gap capacity planning fills. The good news: for an SME, the gap is small. You don’t need a Rs 50-lakh advanced-planning system. You need an honest answer to "can my floor make this by then or not?" and the discipline to act on it when the answer is no.
What is capacity planning, really?
Capacity planning compares two numbers at every work centre:
- Required hours — how many hours of work MRP has stacked onto this work centre across all live work orders.
- Available hours — how many hours this work centre can actually run, given its calendar (shifts, working days, holidays), efficiency, and historical downtime.
When required ≤ available, the schedule is achievable. When required > available, something has to give: extend shifts, outsource the work, push the customer delivery, refuse the next order, or invest in more capacity. Capacity planning’s job is to surface the overload early enough to do something about it, not after the dispatch date slips.
The ratio between required and available hours is the load factor. A work centre at 60% load is comfortable; 80-90% is tight but workable; above 100% is a problem to solve, not a fact to accept.
Finite vs infinite capacity — the fundamental choice
Every scheduling system makes one core architectural choice that decides what its output means.
Infinite-capacity scheduling assumes any work centre can handle any load you give it. The system computes the ideal sequence based on routing and due dates, regardless of whether the work centre physically has the hours. Output is a wish list, not a schedule. Almost all "ERP scheduling" in practice is infinite-capacity. Pros: fast, cheap, simple to implement. Cons: it lies to you about feasibility.
Finite-capacity scheduling respects the hours actually available at each work centre. When a work centre fills up, additional work pushes later in time — sometimes cascading downstream operations with it. Output is a schedule that could actually happen. Pros: honest. Cons: computationally harder, and a small change upstream can ripple loudly downstream.
Most SMEs don’t need finite-capacity scheduling across every work centre. They need it at the bottleneck. Apply infinite scheduling everywhere else (it’s usually right anyway, since the non-bottleneck centres have slack). Apply finite scheduling at the one or two work centres that are always full. This is the 80/20 of SME capacity planning.
The bottleneck principle (Theory of Constraints in 200 words)
In 1984 Eli Goldratt published The Goal (ISBN 978-0884271956) — a business novel that became one of the most-read books in manufacturing operations. Its central insight, the Theory of Constraints — today curated by TOCICO — is this:
The throughput of any production system equals the throughput of its weakest link — its bottleneck. Improving any non-bottleneck work centre delivers zero extra output. The only way to increase factory throughput is to increase the bottleneck’s throughput.
This sounds obvious until you watch how SMEs actually invest in capacity. The CNC mill is the bottleneck. Most weeks it has a queue of work in front of it. The factory owner, frustrated, buys a faster welding machine because that vendor was selling. Welding now runs at 80% utilisation instead of 60% — and the factory produces exactly the same number of pumps as before, because welding wasn’t the constraint. The CNC queue is unchanged.
Five steps from the book, simplified:
- Identify the bottleneck (which work centre always has a queue?).
- Exploit it (no idle time, fastest setup, dedicated operator).
- Subordinate the rest (other work centres slow down to match — don’t produce ahead of what the bottleneck can absorb).
- Elevate the bottleneck (add capacity: second machine, third shift, outsource overflow).
- Repeat (when the old bottleneck is no longer the constraint, find the new one).
This is the single most useful framework for SME capacity planning. Almost everything below in this post is mechanics; this is the strategy.
Worked example: MRP says make 200 pumps but you have 5 CNC machines
Continuing from last week’s pump factory. MRP has stacked work for the next two weeks. Your bottleneck is the CNC mill bank (5 machines). Each machine runs 8 hours/day, 6 days/week. Total CNC capacity per week: 5 × 8 × 6 = 240 hours.
What MRP has loaded onto CNC:
Week 1 is overloaded (129%). Week 2 has slack (82%). The infinite-capacity scheduler would silently let week 1 slip and nobody would notice until the dispatch date. The finite-capacity scheduler flags it now — before the work orders are released.
Four options, in increasing order of cost:
- Re-sequence within week 1. Pull some of the Pump-1HP-INLINE work into week 2 where there’s slack. Net: zero extra cost, but only works if the customer due dates allow.
- Extend shifts. Add a 4-hour overtime evening for 4 days. Adds 80 CNC-hours (5 machines × 4 hrs × 4 days). Now: 310 - 80 = 230 hr needed against 240 available. Solved, with overtime cost.
- Outsource the overflow. Send 70 hours of work to a job-work vendor (see job-work / ITC-04 guide). Cost: vendor rate + 2-day round-trip. Solved.
- Push delivery dates. Call customers and renegotiate dispatch on the lower-priority orders. Worst option from a customer-relationship perspective; sometimes the right one.
The point is: all four options are visible before the slip. Without capacity planning, only options 3 and 4 exist, and only after the dispatch date has been missed.
The four levels of capacity planning
Different decisions need different horizons. Mature factories run four:
RCCP · Rough-Cut Capacity Planning
A monthly sanity check at the major work-centre level (CNC, welding, painting, assembly) before the Master Production Schedule is finalised. Coarse: usually accurate to within 10-15%. Goal is to catch a "we don’t have the CNC hours for next month’s sales target" two weeks before the sales target hits the floor.
Horizon: 4-12 weeks ahead · Frequency: monthly · Granularity: work-centre levelMPS · Master Production Schedule
The signed-off weekly plan: which finished goods get made in which week, in what quantity. The output of RCCP and the input to MRP and detailed CRP. For an SME, the MPS often lives in a planner’s spreadsheet today; in an ERP it’s a first-class object you can drag-and-drop.
Horizon: 2-8 weeks ahead · Frequency: weekly · Granularity: finished-good level by weekCRP · Capacity Requirements Planning
The detailed view that takes MRP output and explodes it against per-operation routings to compute load on every individual work centre, by day. This is where finite-capacity scheduling does its work. For SMEs with one clear bottleneck, you can usually skip CRP and run RCCP + dispatch. Above 100 employees or with multiple competing bottlenecks, CRP starts to pay back.
Horizon: 1-3 weeks ahead · Frequency: daily · Granularity: operation level by dayDispatch · today’s machine queue
"Which job runs on which machine next, in what order, with which operator." Updated hourly on the shop floor. Driven by today’s priorities (the rush order, the QC-failed rework, the machine breakdown). The ERP’s job here is to give the supervisor the right information in the right order, not to over-schedule.
Horizon: next shift · Frequency: hourly · Granularity: machine and job levelRCCP at level 1 + dispatch at level 4. Run RCCP monthly against your bottleneck work centre only. Run dispatch on a digital board on the shop floor (or a printed sheet). Skip MPS and CRP until you have outgrown this combination — usually past the 100-employee mark.
5 things that destroy capacity plans (and how to handle)
1. Setup and changeover time ignored
Most factory operators count run-time-per-piece in their capacity numbers but forget the 45-minute setup at the start of each batch. A schedule with 10 batches a day loses 7.5 hours to setups alone. The plan is theoretical; the floor doesn’t hit it.
Fix: store setup time as a separate field on the routing operation, and include it in capacity calculations. Then minimise batch count by sequencing similar jobs together.2. Operator-machine mismatch
The CNC machine is free. The only operator trained on it is on leave. Effective capacity at that work centre, that day: zero. Capacity planning that treats the machine as the resource (without the operator constraint) over-promises.
Fix: model both machine and operator as constrained resources. Cross-train at least two operators per machine for the bottleneck work centre.3. Quality rework not modelled
10% of work-orders fail QC and come back for rework. The rework consumes the same machine hours as the original operation, but capacity plans rarely allow for it. Result: every Friday the floor falls a half-day behind.
Fix: inflate work-centre demand by your historical rework rate. If 10% of jobs come back, plan capacity at 110% of the nominal MRP demand.4. Downtime not factored
A CNC mill spec says 8 hours/day. The real number, after planned maintenance, unplanned breakdown, tool changes, and lunch, is closer to 6.5 hours/day. Capacity plans against the spec hour number over-promise by 20%.
Fix: track OEE (Overall Equipment Effectiveness) per work centre. Use OEE-adjusted hours, not nominal hours, in capacity planning. For an SME, even tracking simple uptime % for the bottleneck is a big step.5. Plan never updated when reality changes
The Monday capacity plan is right on Monday. By Wednesday a machine has broken, a rush order has been booked, an operator is on leave. The plan is wrong and nobody updates it. By Friday the plan is folklore.
Fix: a capacity plan should be a live view, not a snapshot. Either update it daily (15 minutes of the planner’s morning) or have it auto-update from the ERP. A weekly-only plan that never updates mid-week is worse than no plan.What an Indian SME actually needs (vs SAP-grade APS)
The capacity-planning software market is segmented by factory size. Roughly:
| Factory size | What they need | What they typically buy |
|---|---|---|
| Under 50 people | RCCP + dispatch (often manual) | Whiteboard + Excel |
| 50-200 people | RCCP + dispatch in an ERP, finite-capacity at bottleneck only | Manufacturing ERP with capacity module |
| 200-1000 people | Full MRP II with CRP, multi-bottleneck finite capacity | Mid-market ERP (NetSuite, Acumatica, Dynamics) |
| 1000+ people | APS layer on top of ERP, multi-site optimisation | SAP IBP, Oracle ASCP, Kinaxis, dedicated APS vendor |
The mistake we see most often is mid-segment factories (50-200 people) trying to install an APS, getting a 3-6 month implementation, blowing Rs 30-50 lakh on it, and discovering 18 months later that the planner still runs the floor from her spreadsheet because the APS demands more discipline than the factory has the maturity to provide.
Don’t buy APS until you have outgrown a good manufacturing-ERP capacity module. Most SMEs never need to. The factory that doubled output didn’t do it with optimisation algorithms — it did it by finding the bottleneck and adding a second machine.
How OEMup handles capacity planning
OEMup’s production module ships with finite-capacity scheduling at the work-centre level. The minimal setup:
- Define each work centre with: hours-per-day, days-per-week, operator count, OEE %, the operations it handles.
- Define a calendar per work centre: working days, shift pattern, holidays, planned-maintenance windows.
- Define operation times per BOM level (see multi-level BOM): setup time + run time per piece + work centre.
That’s it. When MRP releases manufacturing orders, the system schedules them against work-centre calendars. The capacity dashboard shows:
- Load percentage per work centre per week (colour-coded: green < 80%, amber 80-100%, red > 100%).
- The bottleneck highlighted automatically (the work centre with the highest sustained load).
- A drag-and-drop view to re-sequence manufacturing orders when an overload is flagged.
- What-if mode: "what if we add a Saturday shift on CNC for the next 2 weeks?" updated live.
No APS-grade optimisation. No sequencing algorithm trying to minimise total setup time across 50 machines. Just an honest view of where the squeeze is, and the levers to fix it. For 95% of Indian SMEs, that’s the right level of sophistication — and the capacity module is included in every OEMup plan.
See your CNC load this week, before it slips
Bring your top-3 product BOMs and your CNC capacity. In 20 minutes we set up the work centre, schedule a week of MRP-driven manufacturing orders, and show you exactly which day the bottleneck overflows — and the levers to fix it.
Book a Demo →The bottom line
Capacity planning is what makes MRP’s output trustworthy. Without it, the schedule is a wish list and the dispatch date is a hope. With it, the schedule is a commitment and the slips are visible early enough to act on.
The single highest-leverage thing an Indian SME factory owner can do this week, regardless of software: walk the shop floor with your supervisor and identify the bottleneck. Compute its weekly hours of capacity. Compare to MRP’s weekly demand on it. If demand > capacity, you have your first capacity-planning decision to make — and the right answer is usually obvious. Everything else — finite scheduling, OEE tracking, dispatch boards — is mechanics on top of that one insight.
Companion reads: MRP Explained for the layer above this (what to make, when); Multi-Level BOM for the data foundation routings hang off; Job Work and ITC-04 for outsourcing the overflow when the bottleneck’s overloaded.
FAQ
What is capacity planning in manufacturing?
Comparing what your factory has been asked to produce (the demand) against what it can actually produce in the available time (the capacity) — and adjusting one or the other when the two don’t match.
Finite vs infinite capacity scheduling?
Infinite assumes any work centre can handle any load (output is a wish list). Finite respects the actual hours available (output is a schedule that could happen). Most SMEs need finite at their bottleneck only.
What is a bottleneck?
The work centre with the lowest throughput in the production sequence. By the Theory of Constraints, factory throughput equals bottleneck throughput — improving any non-bottleneck delivers zero extra output.
What are RCCP, MPS, CRP, dispatch?
Four levels at different horizons. RCCP: monthly sanity check. MPS: weekly master schedule. CRP: detailed work-centre load. Dispatch: hourly machine queue. SMEs usually need RCCP + dispatch only.
Does my factory need APS?
Almost certainly not under 200 people. APS costs Rs 30-50 lakh and takes 3-6 months; a manufacturing-ERP capacity module covers 95% of SME needs.
Simplest useful capacity-planning step this week?
Walk the floor with your supervisor and find the work centre with the queue at end of shift. Compute its hours-per-week. Compare to MRP’s weekly demand. Solve any gap.
How does OEMup handle capacity planning?
Finite-capacity scheduling at the work-centre level. Define work centres with hours, OEE, operations and calendar. MRP-driven manufacturing orders schedule against availability. Capacity dashboard shows load per work centre with the bottleneck auto-flagged.