Choosing process and layout from volume and variety

Physical layout follows an understanding of what actually flows through a plant. A process layout groups similar resources—cutting in one area, drilling in another—so different products can take different routes, at the cost of travel and interdepartmental waiting. A product layout places stations in a repeating sequence, supporting flow when volume is high and the mix stable, but it is less flexible when orders need different routes. A cell collects operations for a related family, potentially reducing movement without pretending every product is identical. Do not turn the product-process matrix into 'more units means a line.' Ask for volume by family, the share following a common sequence, changeover time and mix variation. High aggregate volume with many small routings may leave a dedicated line idle while shared departments queue.

Separate nominal capacity from deliverable capacity. An eight-hour station needing six minutes per unit has an arithmetic ceiling of 480 ÷ 6 = 80 units per shift before losses. With a 60-minute changeover, only 420 ÷ 6 = 70 remain before breakdowns and breaks. Do not add the capacities of stations in series: the slowest constrains sustained output. When comparing a line with a functional route, multiply each changeover by the number of batches, then compare achievable output with demand for each family rather than total demand alone. State assumptions on shifts, availability, skills and inspection. One machine's capacity does not guarantee materials, an operator or testing capacity. A choice based on unit cost alone can break a delivery promise when a small order waits behind a large batch.

Sensitivity testing prevents a worked case from becoming a permanent recipe. If B demand rises from 20 to 40, testing load becomes (60 + 40) × 5 = 500 minutes, 20 beyond the shift, while cutting needs only 400. Moving A to its own line does not solve the problem if testing remains shared. Testing capacity must increase, its actual time must fall, or backlog and a revised promise must be acknowledged; drawing another line on a layout does not remove the bottleneck. If A volume falls to 30 while line setup still takes an hour, setup consumes a larger share of the dedicated line's time than when A volume is 60. Compare decisions over demand and mix scenarios, not one average day. Document measurement quality as well: are times standards or observed averages including rework? Mixing the two can create a false bottleneck.