The third option also carries a high cost associated with R&D on the hardware.
Data/Assumptions
Our decision will be based on the following criteria: costs, payback and probability of success. The cost savings data can be easily calculated since the majority of costs are known in advance. We attempted to identify the cost drivers early in the process, which allowed us to detect escalating development costs so that we could determine when to abort. Payback is readily available through forecasts derived from historical data and in-hand purchase orders. Assumptions regarding the probability of success are derived from historical data and our in-depth knowledge of the industry as gathered through discussions with company management and other internal stakeholders. In calculating the probability of success we took several factors into account such as schedule slips, development/testing failure, management decision-making and customer demand.
It should be noted that in order to satisfy the company's requirement of non-disclosure regarding confidential information, we have scaled the numbers in order to disguise them. The end result was verified internally with the model, using the actual numbers. Anytime numbers are forecast, there is a percentage of error involved. To address this, we conducted a sensitivity analysis to determine the allowable margin of error.
Model and Analysis
Our company's model was best designed by using a Palisade Precision Tree. Since it was very difficult to get an accurate assessment of costs and probabilities, the sensitivity analysis played an important role in this project. At every step, we ascribed some value to the "abandon project" option since we felt that this was a distinct possibility for which we should account. We also needed to ensure that the company would be aware at each step the sunk costs and future costs of each course of action. Some of the major sources of errors, such as the schedule slippage, were difficult to account for but we attempted to account for them as best as possible.
However, management felt this was a risk that we needed to incorporate into the model. We agreed that "do nothing" does not guarantee the continuation of the status quo.
We conducted a sensitivity analysis on the cost savings per unit. We also performed a sensitivity analysis on the forecasted number of units sold. Management felt that these analyses would be critical to their final decision, in particular with regards to determining a reasonable breakeven point. We also wanted to determine the amount of money that should be paid for the market research. The model determined that maximum the company should pay for perfect information was $X.
Conclusion/Future Work
One of the key areas for future work would be to create a model to help better forecast sales. Management expressed that sales was a major concern for them, so better sales forecasting would add value to my model. I also feel that there may be more derivations of possible actions that management provided me with. Another potentially valuable addition would be Oracle Crystal Ball or Palisade @RISK support in order to simulate X number of runs using different probability distributions. The assumptions and costs for which we have the highest confidence could be done with a triangular distribution and for the others we can use a uniform distribution. Lastly, we could have added support for the input of actual live data from our backend servers with the ability to run different scenarios.
That said, the results that the Palisade Precision Tree yielded for our company have allowed the company to choose the appropriate course of action with confidence. The model allowed for a scientifically robust synthesis of the information available. The decision tree also helps confirm the costs and other assumptions that were presented so that the entire organization can understand the model. We felt it important to ensure that the key decision makers agreed with all of the assumptions we made.
