Consensus versus Average Forecasting1
Appendix A: Simulation Comments4
Appendix B: Simulation Results6
Consensus versus Average Forecasting
The consensus forecasts worked well for quick insight into estimated demand for each month. In our first year we used the consensus demand because we did not know the dynamics of the group, and we were relying on their expertise to guide us toward a more accurate forecast. As we progressed through the simulation we came to the realization that the consensus forecasts were often much different than the average estimated demand. After we analyzed the results of the first couple years, we noticed that the average demand was generally more accurate. This led us to the conclusion that the dynamics of the forecasting team were likely distorting the estimates for the consensus number. There were strong personalities within the team that seemed to sway the opinion of the team members to agree with them, thus lowering the accuracy of the estimation. For the final two years we spent more time looking at the individual opinions of the team and tried to exclude estimates that were exceedingly high or low compared to the rest of the team. This gave us a number that was close to the average estimate of the team and allowed us to make a more accurate forecast. Overall, the consensus forecast was a good tool to use to quickly see whether demand was expected to be high or low. The biggest downside is the lack of accuracy. The average demand was a more accurate tool as long as we took the time to check each individual opinion to see how the number was made up. Our most successful forecasts seemed to come out of a combination of both the consensus and average. Options
In general we tried to look at the overall benefits the option would provide. First we looked at the effect the option had on the consensus demand forecast. If the demand forecast went up, we decided to add the option to the phone because we believed we would be able to sell more units. The second deciding factor was whether or not the option increased profit for the model. We decided to add any option that would increase profit for the model. If the option increased both profit and consensus demand, we checked the graphs to look at the standard deviation. We avoided options that had a high standard deviation because we believed that there was more risk that the estimate would be inaccurate. A lower standard deviation indicated a higher level of confidence with the effect the option would have on demand. In retrospect we relied fairly heavily on the consensus estimates to choose the options. We realize that the consensus demand estimates were often inaccurate, and could have led us to choose options that were not beneficial to the models we were trying to sell. Just as we switched to using average demand estimates for our forecasting, we should have switched to using individual estimates when we were choosing options. It would have provided a greater level of accuracy and could have helped us make better decisions. Demand Forecast
Initially we were reluctant to order too much stock because we were focused on the high cost of inventory becoming obsolete. This led to us experience a number of stock outs in our first year. We realized that this was costing us a substantial amount in lost profits. Due to this we changed our strategy for the following years to focus on ensuring we had products available to sell. Example of this would be in the first year, where we found the inventory cost of $4 per month and markdown loss was $17 per unit, but the profits were $70 for Model A and $90 for Model B. Using this as an example and a lesson learned from first year, we concluded that it would beneficial to carry excess inventories to compensate for fluctuations in demand, and that it is better to have more inventory and have the holding...