ISE 754 (Fall 2026) 6.3 Production System Design
Overview
Michael G. Kay compares design models for service, make-to-order, and make-to-stock production systems, showing how capacity, cycle time, inventory, and delayed discounting shape profit. Service and make-to-order models optimize capacity, while make-to-stock adds a maximum inventory decision; cycle time is estimated by simulation for services and an analytical Factory Physics approximation for make-to-order.
Key takeaways
- Service and make-to-order systems optimize capacity R_E, while make-to-stock systems require a second decision, maximum inventory Q_max.
- The service-system model cannot rely on steady-state cycle-time formulas because queues clear daily and demand rates can change by hour; FIFO simulation accommodates both conditions.
- A Little's Law backlog calculation captures waiting caused by demand exceeding capacity but omits random-arrival waiting, so it can understate cycle time even when capacity exceeds peak demand.
- The make-to-order Factory Physics approximation estimates cycle time from the squared coefficients of variation of arrivals and production, avoiding simulation when steady-state assumptions apply.
- In make-to-stock design, delayed discount affects value only during stockouts, while inventory carrying cost trades off against the stockout probability π₀.
- Equipment capacity cost should reflect annualized capital expense, operating hours, and production rate so underutilized machinery is not assigned an unrealistically low cost per productive hour.
Chapters
- Service and make-to-order designs choose production capacity R_E; make-to-stock also chooses maximum inventory Q_max.
- Service-system profit weighs marginal unit value, price minus cost, against delayed discount G based on cycle time and a capacity cost K times capacity.
- Because service queues empty daily and hourly arrival rates vary, a FIFO simulation with period-specific Poisson arrival rates estimates average cycle time without assuming steady state.
- A Little's Law recursion estimates waiting from demand exceeding capacity, but misses stochastic waiting when arrivals bunch randomly; it is better as a rough check than as the primary design method.
- The make-to-order objective has the same basic form as the service model, balancing delayed discount against capacity cost.
- Instead of simulation, cycle time can be estimated with the Factory Physics approximation introduced in Lecture 1.3.
- The approximation uses the squared coefficients of variation for both arrivals and production, and applies when the system can be treated as steady state.
- Make-to-stock design optimizes both production capacity and Q_max; the probability of stockout π₀ and average inventory Q̄ depend on those choices.
- Delayed discount applies only when an item is out of stock, while inventory carrying cost accounts for holding stock; the inventory cost includes both unit cost C and capacity cost K.
- For expensive equipment, annualize the initial investment into an effective capital cost, then divide by operating hours and production rate to estimate unit capacity cost.
- A million-dollar machine illustrates why equipment cost must be allocated to actual production time rather than treated as a simple hourly operating expense.
Transcript excerpt
0:02 This lecture discusses production system design. Models are presented for three types of production systems, service and then make-to-order and make-to-stock. The difference between service and the other two systems are that services typically deal with people. People are things that essentially empty every day in the sense that you can't carry items being produced over from shift to shift. That's different when you're making goods. The difference between make-to-order versus make-to-stock, make-to-order because every item is custom-made. It doesn't really make sense to carry any type of inventory because you're likely to have to change the item. Whereas make-to-stock, you're making
0:49 more standard types of product. In terms of what the design is actually solving for, in this case of service and make-to-order, the one single design parameter or knob that is being decided is what is the service or production capacity, which is R sub E. And then in the case of make-to-stock, you actually have a second knob or parameter, the production capacity and then also the maximum queue level. This is the same Q max that we saw in the earlier lecture when we looked at base stock inventory models. This is the formulation for the service
1:52 system design model. And this basic format is an extension of what we saw earlier for the inventory models, the periodic safety stock models. What's different here is that we add a capacity term. So we're again maximizing profit by changing the capacity. We're looking at the marginal value per unit of product produced, which is the price minus cost. And then we're also adding to this the delayed discount factor G. And then here we're adding an estimate of what is the cycle time that the unit being serviced spends in the system. And the idea is the longer it spends,
This is the start of the transcript. Save this video free for the whole transcript, searchable and time-coded.
In this course
Related notes
Summary, takeaways, and chapters were generated by AI from the video's transcript and may contain errors. The video belongs to its creator, Michael G. Kay.