Product mix optimization is not simply a question of choosing the products with the highest apparent margin. Production planners must account for capacity, processing time, changeovers, costs and the uncertainty of future prices and operating conditions. A manufacturing case study involving a proposed paper-napkin production line illustrates how deterministic optimization and Monte Carlo simulation can be combined to compare production plans and assess investment risk.
The decision-making objective was to maximize the investment’s net present value (NPV), while meeting a minimum production commitment and staying within available operating hours. The analysis also examined how planners could reduce the uncertainty of the financial forecast without overlooking the effect of those choices on expected returns.
Why capacity constraints complicate product mix decisions
The case involved three product types: white napkins (A), coloured napkins (B) and patterned napkins (C). The company planned a new line partly to replace outsourced production of white napkins, for which at least 700 tonnes per year were covered by long-term orders. Remaining capacity could support products B and C.
Available machine time was a central constraint. Products differed in average production time per tonne, while format changes, cleaning and machine repairs affected usable operating time. Changeovers were especially significant because production could require format adjustments, cleaning between patterns and time to restart the line. A product mix that looks attractive based on selling price alone may therefore use capacity less effectively once processing and changeover requirements are considered.
The financial model evaluated operating cash flow over an eight-year investment life and used NPV as its measure of economic efficiency. In general, NPV discounts projected cash flows over time and subtracts the initial investment cost. In this case, production volumes, prices, rejects, variable costs, labour expenses and other assumptions contributed to the projected cash flows.
Product mix optimization: deterministic and stochastic approaches
Start with a deterministic model
The first method treated model inputs as fixed values. A Generalized Reduced Gradient (GRG) algorithm was used to optimize the production plan, subject to the minimum 700-tonne annual volume for product A and an annual operating-time limit of 7,332.5 hours. The model also required products B and C to each receive a share of the remaining production time, preventing either product from being excluded.
The resulting plan allocated the free capacity between B and C at a ratio of 0.1 to 0.9. Its NPV was approximately EUR 2.31 million, compared with about EUR 1.99 million for the preliminary plan, which divided the free capacity equally between those products. The deterministic result offered a useful benchmark, but it assumed that inputs such as market prices and production times would take their specified values.
Add uncertainty with Monte Carlo simulation
To examine how variable inputs could affect the decision, the analysis assigned probability distributions to selected assumptions. These included triangular distributions for prices, variable costs and investment costs; normal distributions for rejects and several time and cost inputs; and a discrete distribution for the number of format changes per week. The model then used repeated random sampling to generate possible combinations of input values and calculate the resulting NPV.
Monte Carlo simulation does not remove uncertainty or guarantee a particular outcome. Instead, it helps decision-makers see how a financial result may vary when assumptions change. In the case study, the stochastic model was optimized using OptQuest, with the objective of maximizing mean NPV while retaining the capacity constraints. The simulation used 10,000 runs, and the optimization process used 1,000 experiments.
The best stochastic plan again assigned 700 tonnes to product A and divided free capacity between B and C in a 0.1-to-0.9 ratio. Its mean NPV was approximately EUR 2.30 million, close to the deterministic result. However, the forecast standard deviation was about EUR 336,517. The study reported a 68% probability of an NPV within approximately one standard deviation of the mean, and an approximately 81% probability of exceeding the preliminary plan’s NPV of EUR 1,994,379.
Two ways to manage production-plan risk
The case study explored two distinct options for reducing forecast risk. The first was to change the product mix. Because product C had relatively high uncertainty in both price and cost, shifting some capacity toward products A and B reduced the NPV forecast’s standard deviation. The trade-off was a lower mean NPV: as the risk-reduction requirement increased, the optimized plan placed more capacity in product B and less in product C. At the largest tested reduction, the mean NPV was approximately EUR 2.02 million.
The second option was to examine uncertainty in input assumptions. Sensitivity analysis identified product C’s price as the largest contributor to output variance, followed by product A’s price. The analysis also identified variable cost C and production-time assumptions among the influential inputs. Reducing the assumed uncertainty in prices A and C lowered the forecast standard deviation in the tested scenarios without a statistically significant change in mean NPV.
This approach depends on whether the uncertainty can realistically be reduced. Tighter assumptions should reflect better information or control, not simply a more optimistic model. The study based its input ranges on estimates and noted that demand uncertainty was not included, highlighting the importance of grounding probability distributions in suitable data where available.
Applying the method to manufacturing decisions
For engineers, integrators and operations leaders, the practical value is in linking production constraints to financial outcomes. A useful workflow is to:
- Define the decision and constraints: specify required output, available machine time and any product-mix limits.
- Build a transparent baseline: connect production volumes and operating assumptions to cash flow and NPV.
- Represent material uncertainty: assign distributions to relevant inputs and document their basis.
- Optimize and assess risk: compare the expected result with the spread of simulated outcomes.
- Test mitigation options: evaluate whether changing the mix or improving input certainty better fits the organization’s risk tolerance.
Deterministic optimization is valuable for establishing a reference plan, while stochastic optimization reveals how that plan behaves under uncertain conditions. Neither approach makes the decision automatically; the quality of the conclusions depends on the model, constraints and assumptions.
Conclusion
Combining product mix optimization with Monte Carlo simulation gives manufacturing decision-makers a clearer view of both expected investment performance and forecast risk. The case study shows that a plan with a strong mean NPV can still carry substantial uncertainty, and that risk can be addressed either by reallocating production capacity or by reducing uncertainty in influential inputs. For plant and investment planning, the practical implication is to evaluate not just which plan looks best under fixed assumptions, but how reliably it may perform when operating and market conditions vary.