Forecasting for Aviation Spares Inventory – Part 2
In Forecasting: Part 1, we examined the complex, multi-dimensional problems of forecasting for airline and MRO inventory. We considered how the nature of this industry impacts on inventory planning, and in particular, how it impacts our ability to forecast demand. We acknowledged the difficulties of uncertain and stochastic demand, but showed, through a short example, that while the demand is difficult to predict, it can still be quantified statistically and we are still capable of anticipating and planning for it.
Our example illustrated that what is forecast, what is most likely to occur and what we would provision are three very different things. Where demand patterns are highly stochastic, we cannot expect nice, clean trend lines against which to forecast. The demand is too uncertain, and what is “on average” likely to occur is practically immaterial to the reality. In the example from the previous article, it is most likely that we will require 0 or 2 parts for the upcoming 2 checks, but, because the demand could potentially spike to 4, we need to hold between 4 and 8 parts. The uncertainty and erratic nature of the demand, coupled with our natural aversion to risk, can cause us to sometimes provision higher levels of stock than we will probably need. The stochastic nature of the demand is a feature of our business and we must embrace and work with it. We can also apply smarter methods to optimize the cost of provisioning against this uncertainty.
Of the four dimensions we named initially, three were discussed in the previous article and have been briefly summarised above. The final dimension is the Cost Dimension, which is laid out below.
The Cost Dimension
Our objective is to provision against forecast uncertainty to deliver a level of service. We are spending to mitigate uncertainty, in an economically efficient manner. Obviously, it does not make economic sense to make these risk-mitigating decisions on a part by part basis in isolation. This myopic approach opens the door to inefficiency and waste. We could find ourselves in a situation where we are increasing investment in one part, while a lower incremental investment in another part could provide greater service level risk mitigation. Take question 3 in the above example; should we hold 6 or 8? If the part costs $10 then the answer is clearly to hold 8; if the part costs $10,000 we would need to reconsider even holding 6. Logically we should not make this decision by looking at parts in isolation. In an optimal situation, we would provision holistically across the entire inventory considering each individual part’s forecast uncertainty to deliver high availability at the lowest cost.
To successfully plan inventory to support aircraft maintenance, we need to go beyond traditional forecasting methods. We are aware that current forecasting techniques – or rather, current applications of forecasting techniques – are inadequate. We must accept that there will be a high degree of forecast uncertainty and that the trend-lines will not be meaningful because of the multiple demand drivers and high demand variability. Understanding this, we must find a more sophisticated, multi-dimensional approach to provision against forecast uncertainty holistically across all parts, at minimum taking the 4 dimensions of Forecasting, Uncertainty, Risk and Cost into account. Only by doing so can we maximize availability at the lowest total cost.