Implementing a Demand Forecasting tool is always a critical task when creating your S&OP process. This is a very laborious activity since it will require going back at least three years in time and reviewing sales history numbers to load that information into the new system.

However, the team needs to understand that this is one of those situations where the old saying of “garbage goes in, garbage goes out” is most true.

Over the years, I have found that companies and demand teams are inclined to try to cut corners in this process, which could cost them a lot in the future.

I want to present two specific areas where time investment is required to achieve the best results.

Identifying promotional sales data

When you review your sales data, sometimes the demand team would look for the information, maybe they would try to fill in the missing data, and finally, would try to cross-check the global sales numbers to make sure that they are aligned with finance sales numbers.

Well, this is great but is not nearly enough to help you achieve the goal of having a reliable system to forecast your sales. They are missing a very important part of the process.

In most industries, promotional events are one of the main drivers of an increase in sales. If you offer a special discount to your customers, or create a product kit, or offer a “buy one & get one free” promotion (BOGOF), your sales will be increased in a way that is not related to your regular and seasonal sales. If you feed these sales numbers directly into the system, without taking out the effects of your promotions, then your forecast for that month is going to be completely over-estimated. Also, it will be impossible to evaluate the real seasonal trend of your market.

Therefore, you need to go back in time and review all your notes, meeting minutes, or any resources, to identify all the significant promotional activities that led your company to achieve those sales numbers.

Promotional activities need to be identified as Promotional Events, and these events need to be classified according to their types. It is not the same to give a 30% discount on one product as offering a BOGOF for the same product, each one will generate different responses from your market and customers. Thus, they need to be identified and further differentiated.

Only separating and classifying these events will allow your statistical platform to help you identify how much these events have impacted and will impact your sales.

Once you separate the effects of these promotional events, you can apply statistical tools (monthly contributions, linear regression variations, etc.) to identify the seasonality of your sales, which eventually is going to be the baseline for your forecast.

In my experience, unless you have a very formal promotional activity record at product level, it will take your team a few months to be able to correctly cleanse the sales data. If today, you do not have this kind of information and you are thinking of implementing a forecasting system in the future, you urgently need to build one. So, you better start identifying and classifying your promotional events now.

Identifying supply limitations

Another demand event you need to identify and record to improve your sales data is supply disruptions.

For example, if in a certain month of last year, you identified an inventory limitation for one of your products, and then you needed to limit the number of units your customers could order, or you had several weeks of stockouts and you weren’t able to sale this product, you need to identify those events and their impact in your sales data.

If you feed directly your forecast system with this sales data, the projected sales are going to be underestimated.

Again, as in the case of positive promotional events, these negative supply-related events must be identified and classified to statistically evaluate their specific impact on sales. If you compare your baseline seasonal sales with those specific events, you will be able to quantify their impact. If you use backorders to fulfill all your customers’ requests, then that would be a good place to start looking for those impacts.

As a general recommendation, if your business model does not use backorders, then it is a good idea to be able to capture the initial customer request and compare it to the shipped order values. The difference between them would be your inventory restrictions. The team would need to formally analyze these kinds of restrictions to define which ones are relevant and should be reflected as “lost sales”. This exercise will also allow you to measure your % Fill Rate, which is one of your core supply chain indicators.

In Summary

The decision of implementing a sales forecasting tool is a very important step in improving the results of your S&OP process. However, given the time and financial investment required to execute this kind of project, you better make sure that all this effort is not wasted. For that, you need to validate that the demand team invests enough resources into cleansing the sales data.

Identifying and classifying positive and negative demand events is critical in cleansing data, and also in obtaining a reliable forecasting tool.

If you waste this opportunity to improve the state of your historical data, then you will regret it in the future when your forecast results are unreliable.

On the other hand, if you do it correctly now, then the new system will successfully help you to project both the seasonal forecast baseline, as well as the impact of promotional or supply constraint events. Therefore, the final recommendation is to start, as soon as possible, identifying and registering these events. Then you will be well prepared when you decide to implement this kind of tool.

Juan Pablo Franco

General Manager