A material plan is feasible only when every component and process can meet the date

The net requirement is an unconstrained quantity until supplier and factory constraints are applied. A supplier may require a minimum order, sell in fixed packs or limit production within the required window. The factory may lack the machine hours, approved process capability or inspection capacity needed to convert the kit. EOQ extensions address lot and replenishment trade-offs, while MRP time-phases requirements; neither fact makes every calculated quantity and date physically achievable.[S03][S07]

Order rounding comes first. In the illustrative net kit, a 300-metre fabric pack would turn the 260-metre net need into a 300-metre purchase and leave 40 metres before later consumption. A 200-piece zip minimum would turn the 110-piece need into 200 pieces and leave 90 pieces. If thread comes in 5,000-metre order multiples, the 12,400-metre need would round to 15,000 metres and leave 2,600 metres. These residuals are arithmetic consequences of teaching assumptions, not recommended purchase quantities. They need an owner, an age and a credible route into later compatible orders.[S03]

Capacity reconciliation then asks whether the acquired kit can be converted in the planning window. For each process, required work is the sum of order quantity multiplied by the applicable work content, plus explicitly modelled setup and rework. Available work time is the approved calendar time after breaks, planned maintenance and existing commitments, using the same process boundary. Detailed line balancing belongs to WP15. The material-plan test here is coarser but essential: do not release a customer promise whose aggregate process load already exceeds feasible hours.[S07]

Consider an illustrative capacity check for the 1,000-garment order. Assume cutting work content of 0.8 minutes per garment, sewing of 24 minutes and finishing of 4 minutes, with setup and ordinary allowances already included. Required time is therefore 800 cutting minutes, 24,000 sewing minutes and 4,000 finishing minutes. After other commitments, suppose the window offers 720 cutting minutes, 22,800 sewing minutes and 4,800 finishing minutes. Cutting is short by 80 minutes, sewing by 1,200 minutes and finishing has 800 minutes unallocated. The material kit could be numerically complete while the order remains infeasible in its current slot.[S07][S19]

That result calls for a decision, not an average utilisation percentage. Cutting and sewing cannot borrow finishing's spare minutes unless a genuinely qualified alternative process exists. Possible responses include moving to another approved slot, splitting the order across feasible deliveries, using an approved alternate resource, resequencing other orders or renegotiating the date. Overtime is a separate operating and labour decision, not free capacity that the planning engine may insert automatically. Each option changes cost, other orders or risk and should appear in the approval record.

Lot size also interacts with flow. A larger batch may reduce repeated setup, but it can occupy a process for longer and delay other complete kits. Splitting a batch may improve delivery sequencing while adding setups and handling. Time-phased lot-sizing methods provide a way to compare these choices over a fixed horizon, but the objective must include due dates and terminal stock rather than setup cost alone.[S04] The planner should test the actual candidate lots against process calendars rather than assume that the supplier's rounded quantity is also the best production batch.

Supplier capacity deserves the same test. An accepted purchase order for the full net requirement does not prove that the supplier can produce and release it by the required usable date. Partial deliveries may cover an initial operation while leaving later operations exposed. The plan should therefore retain dated receipt tranches and their accepted quantities. A feasible promise uses the sequence of usable supply and conversion capacity, not the order total alone.

A production buffer has a different purpose from replenishment safety stock. Safety stock protects a defined material service outcome against demand and supply uncertainty. A production buffer decouples operations whose processing times, failures or setups would otherwise stop one another immediately. Both are inventory, but they sit at different boundaries and require different evidence. Moving fabric into the sewing hall does not prove that more protection has reached the customer.

Three events should remain distinct in the line record. Starvation occurs when an available operation has no released input to work on. Blocking occurs when an operation cannot pass completed work onward because the next buffer or process cannot accept it. Quality hold occurs when material or work cannot advance because disposition is unresolved. A single idle-time total mixes these causes and can lead to the wrong remedy. More upstream release may relieve starvation, worsen blocking and increase the amount of work exposed to a quality problem.

A deliberately small buffer before a genuine bottleneck can protect that resource from short upstream interruptions. Its purpose is to keep scarce conversion capacity doing useful work, not to maximise work in progress everywhere. The buffer should have a defined location, unit, maximum quantity, release authority and review rule. Its size needs observed variability and recovery behaviour. Without those inputs, GFT would treat any proposed quantity as illustrative and test it before granting it policy status.

Finite-buffer simulation offers one way to conduct that test. NIST's Simantha method reference explicitly represents asynchronous production lines with finite buffers and includes machine degradation and maintenance functions.[S19] This supports the modelling choice, not a claim that a garment line has been simulated here. A credible model would need the actual route, processing-time distributions, setups, breakdown and repair behaviour, rework loops, release rule and buffer capacities. It would also need enough replications to distinguish a policy effect from random variation.

Controlled release and batch release should be compared under those common conditions. Controlled release admits work when a defined line or buffer condition permits it. Batch release admits a larger order quantity according to the production lot. Neither rule is universally superior. A large batch may be appropriate when setup cost dominates and downstream capacity is available, while controlled release may limit queues when disruption and long waiting dominate. The comparison must hold demand, route, material availability and failure scenarios constant so the release rule is the actual difference being tested.

An illustrative scenario can show why raw loading is a poor objective. Suppose a batch release places 1,500 pieces inside the boundary while a controlled alternative produces a measured average of 1,000 pieces, and both later demonstrate stable throughput of 500 good pieces per day. Little's Law would imply average flow times of 3 days and 2 days respectively. Those outcomes are constructed for arithmetic, not predicted effects of either rule. The test is whether a real pilot can sustain service and throughput with less waiting, not whether the line initially looks busier.[S28]

A large queue can also conceal failure for longer. Work waits without an immediate shortage signal, quality defects may be discovered only after more pieces have been produced, and order identity becomes harder to preserve as batches move. Buffer caps and age alarms make those conditions visible earlier. They do not remove the need to repair the underlying process. GFT would review starvation, blocking, quality-hold age and completed orders together before changing a buffer, because improving one local measure can damage the whole promise.