A few weeks ago, the consultant Fernanda Monteiro wrote a brilliant essay on the need to monitor an indicator measuring the value added to the sales forecast for each stage of the process, the FVA (Forecast Value Added). I fully share your view, emphasizing that controlling this indicator is fundamental to direct corrective actions both in the construction of the Demand Plan and in the course of the process as a whole.
There is often value destruction during the planning process. This occurs because, involuntarily, when trying to collaborate, some areas end up introducing their own biases, worsening the accuracy of the plans and generating a dysfunctional effort.
In this way, the FVA would be a key indicator to understand both what are the moments of improvement/worsening of the plan, aiming to identify their causes and propitiate the taking of managerial actions, as well as evaluating their magnitude to determine if the efforts of collaboration are being offset by an improvement in predictability e utility of information. It is even possible to reach the limit of assuming that some contribution steps may be unnecessary, when weighing the cost-benefit ratio. This indicator can not only be associated with the added value of each step in the S&OP, but can also be used to help define the best horizon for decision-making.
Weeks ago, I commented that theoretically, the shorter the planning horizon, the closer in time the period that will happen is to the moment of analysis, the better the predictive capacity, given the fact that there is a longer series of demand to be analyzed, and a shorter period available for extraordinary events. It happens that many times this is not what happens in the reality of companies, as shown below.
Figure 1 - Forecast Accuracy by Forecast Horizon
Source: ILOS
In the line graph, we observe the series of sales forecast accuracy measured and adjusted weekly, from 12 weeks ago (S-12) to the week before when the accuracy was measured (S-1). It was possible to identify 3 very different moments in this series: a first moment, referring to the first 4 plotted weeks (from S-12 to S-9), in which the forecast accuracy oscillates at the same level, without a clear pattern of behavior ; a second moment referring to the following 4 weeks (from S-8 to S-4), in which the forecast error reduces progressively and substantially, associated with the incorporation of qualitative information to the forecast, whose horizon coincides with 2 months (8 weeks , therefore); and, finally, a third moment in which, despite the horizon reduction, the forecast does not improve, but worsens.
Figure 2 - Added Value to the Sales Forecast by Horizon
Source: ILOS
When analyzing the proposed FVA indicator, it was possible to identify that despite the company giving up its response capacity, accepting a reduction in the planning horizon by running new forecasts closer to the fact, the forecast worsened. Seeking explanations for such an event, we were able to identify that the closer it got to the current week, the more expectations in relation to the short term brought biases that contaminated the Demand Plan. This hypothesis was confirmed with the following analysis of forecast bias.
Figure 3 - Sales Forecast Bias by Horizon
Source: ILOS
In the analysis above, we noticed that the forecast bias became more and more optimistic over time (negative MPE means that forecast>demand), worsening the accuracy of the Demand Plan in the aggregate. This symptom could be attributed to two main factors.
The first is called wishful thinking, a deviation from rationality that means taking your desires for reality. As the company really wanted to meet its goals, it ended up adjusting the forecast by overestimating the numbers, without any empirical basis for this, so that the forecasted demand was close to the targets. What should have been the most likely number of sales for the next period became the target sales volume. This factor was amplified as the given week approached, due to the progressive increase in distance from the target, as a result of consecutive frustrations in sales volume.
The second factor is related to the mindset from the commercial area, which tends to increase planned volumes to ensure the availability of its products by the Operations areas, without adequately weighing the associated costs, given its more wasteful behavior in relation to resources. In this way, the Sales area, fearing that the Operations areas will not fully meet the requests, becomes more prone to overestimating the volumes, so that the collaborative billing information, instead of having the most probable volume reported as a target, had it as a floor with one more increase in volume to reduce the risk of shortages. Paradoxically, in the integrated planning process there was still a search for the local optimum instead of the global optimum.
Finally, caution should be exercised regarding the successive revisions of the Sales Forecast, which not only can hinder the development of the Operating Plans, but can also end up worsening the predictive capacity instead of improving it. To mitigate these problems, it is desirable to define a forecast freeze, from which no changes can be made, preventing the Plans from being distorted by the heat of execution in the short term. The moment from which the saving associated with improved forecasting is not able to cover the additional cost of reduced responsiveness, it is a good clue for freezing, and the FVA is one of the keys to defining this moment.
References
CORRÊA, HENRIQUE L.; GIANESI, IRINEU GN; CAON, MAURO. S&OP – Sales and Operations Planning, Production Planning, Programming and Control. Atlas publishing house, 2001.
WALLACE, THOMAS F.. Sales & Operations Planning, The How-To Handbook. TF Wallace & Company, 1999.
WANKE, P.; JULIANELLI, L. Sales Forecast: Organizational Processes & Qualitative and Quantitative Methods, Editora Atlas: Rio de Janeiro, 2006.
