Introduction to Demand Planning

As put forth in the first article, demand planning seeks to predict future buying behaviour by drawing inferences from customer buying behaviour. The output of a demand planning exercise is estimated sales of a product in a particular geography.

Demand Planning is a structured way to arrive at a enterprise wide agreement on the expected sales of a product. The following are the key steps of a demand planning process
Data collection & cleaning
Statistical forecasting
Final Consensus forecasting

The first step of demand Planning is to determine the horizon of previous sales to consider. In typical cases, 24 months of rolling data is used to create a forecast. However in many industries such as garments and fast moving consumer goods, the product lifecyle is very short. In such cases, a concept of like profiling is used. In like profiling various products (SKUs) which have the same attributes in terms of product characteristics, price points and target segments are treated as same products for the purpose of forecasting.After the data has been collected, outliers are cleaned. Outliers are identified based on various business rules. For example, one organization defined outliers as those demand points which were more than 2 times the averaeg. This led to the elimination of 2 historical data points.

Once the data has been identified & cleaned, various statistical methods are used for demand planning. The most popular statistical methods are listed below. For a more detailed explanation of each of these methods, please refer to the article on statistical methods.
> Simple average
> Moving average
> Weighted Moving average
> Trend model
> Seasonality model
> Winters model
> Holtz winters model

The statistical forecast generates a forecast that may or may not be agrreeable to all stakeholders. In cases where the sales has seen significant variability, the forecast may be quite inaccurate. In other cases, where the marketing team wants to run a promotional campaign during an otherwise lean season, the forecast may need to be pumped up. However, all inputs from the sales & marketing teams may not be correct. The supply chain planing team , ideally, should be in a position to challenge the input numbers by referring to the current inventory position and historical sales patterns. It is important to note that in most organizations, sales & marketing have their own revenue targets to achieve and hence they prefer plugging in higher numbers. Supply Chain, on the other hand, is responsible for the inventory and hence can find itself in a difficult situation if it does not sufficiently challenge the numbers from sales.

Once a consensus is reached between sales, marketing & supply chain, the demand planing process is said to be complete. The above sequence is followed typically for old products. For new products, marketing has a go to market quantity for different markets. The go to market quantities are merely added on top of the forecast for other products. In case of phaseout SKUs, the forecst is adjusted to reflect the liquidation timeline. For example, a product has stock of 300 units and needs to be liquidated in 3 months. Even if the product has a forecast of 400 units, it will be adjusted to reflect 150,100 and 50 units over 3 months.