Demand Profiling – A small but pragmatic approach in Data Analytics

Introduction

Demand Profiling is an important strategic tool used in predictive analytics domain of supply chain management. Demand profiling at an item level (using volume and variability classifiers) is the primary driver of supply chain strategy. It is the basis helps in inventory and logistics profiling. At the component level, demand profiling serves as an important tool in supply planning, demand planning and forecasting. The complexity of the supply chain is dependent on customer demand profile, geographic locations and lead time and with the help of demand profiling, we can turn data into valuable and executable information. It addresses the problems related to complex supply chain, large inbound component, high finished goods volume, dual and parallel sourcing, longer lead time and high inventory levels. It is a modular and tactical tool often considered as the watchdog of traditional planning engines (supply inventory, demand manufacturing etc.), thus identifies the most relevant basis for supply chain segmentation.

Complexity of Supply Chain = f (Customer Demand Profile, Geographic Location, Lead Time)

Customer Responsiveness = f (LT, Order Visibility, Order Forecast ability)

Customer Demand Requirements= f (Volatility, Variety, Product Complexity)

The Methodology

Demand profiling has got two dimensional objectives: –

1. To profile the supply planning and forecasting of each of the components.
2. To profile the monthly demand of consumer finished goods requirement, in the form of a Consumer FG demand profile of each of the component to be used on monthly basis for the entire year

At an initial level, demand profiling includes the inputs in the form of Value, Volume, Velocity, Variety, Volatility of the existing demand from the data. Thus, actual past transactions will be an input to the tool. From the transactions data, the required Vs are calculated as the first step. Demand profiles forms the basis of a logic calculates different coefficients for each of the part numbers based on the input transactions. The customer demand is in the form of value, variation, volatility, volume and velocity. The data is churned using the basic steps in excel that helps in part level segmentation. After receiving the demand data, the Analyst needs to understand the demand data and find out the dependencies of volume, variation, volatility on each other.

Then the customer requirement index and customer demand requirement index is calculated for each SKU & each SKU is placed on a simulated graph. Constraints are applied to the models and the required constraint demand profile is generated. Predictive modeling is used for data profiling which involves the following algorithms: –

1. Statistical Models – Usage of simple time series, linear regression & advanced time series.
2. Machine Learning – It is directed by the use of artificial neural network, decision trees & deep drilling of the available data

The Challenges

The various challenges to cater to in the supply chain domain while profiling the demand can be summarized under the following heads: –

• Complex supply chains
• Large inbound component
• High Finished goods volume
• Dual and parallel sourcing
• Longer lead times from ‘make to order’
• High inventory levels

The Output

The output of demand profiling work is in the form of a suggested supply chain fulfilment model for each of the SKUs. The various models in this output are: –

1. Local Sourcing
2. Global Sourcing
3. Make to Order/ Engineer to Order
4. Supply from Stock
5. Delayed Differentiation

The Work Flow of the Demand Profiling

The flow of the work for demand profiling as conceived by the experienced data analytics team at SS Supply Chain Solutions is as follows:

Demand Profiling based on Descriptive Analytics

To build a demand profile based on Descriptive Analytics, the following steps need to be followed: –

1. Pick a product, product family, customer, customer-item pair, or business unit of interest.
2. Determine an appropriate time unit – hourly, daily, monthly.
3. Gather the true demand as best you can.  Be careful about using promise dates instead of requested dates, and be doubly cautious of schedules which are often smoothed, filtered, or otherwise manipulated.
4. Create the simulated & detailed graph or time series plot.
5. Now calculate some simple descriptive statistics – range, minimum, maximum, standard deviation, to proceed further for the demand profiling calculations.

Demand Profiling & Product Life-cycle

Once the demand profiling is done considering the various spikes, levels, trends & skewness at the SKU level, it reduces the stress on analysts to cope with the various stages of PLC starting from the launch to the end-of-life phase. Since the demand profiling is done quarterly or half yearly or annually, it is more likely to consider all the disruptions in demands & hence cater to the requirements without causing any extra mile shortage.

Sample Demand Profiling Plots

In the following example, the demand profiling has been carried out on the basis of monthly sales data, by taking its average the coefficient of variation has been calculated. The graph plotted with average monthly sales in x-axis & coefficient of variation in y-axis depicts the following: –

• As average monthly sales go up, the coefficient of variation goes down
• The slope of coefficient of variation is not very significant, indicating there is demand is not stable at even at higher volumes

Demand Profiling on Average Monthly Sales

Demand Profiling on Average Weekly Demand

The above graph depicts the plot of normalized for average weekly demand versus average aging for a period of 12 months carried out through demand profiling. The highlight of the graph is:-

• Around 70% the products have average ageing less than 20 days
• Average ageing is less for products with higher average weekly demand
• Profiling graph shows area of improvement for products with average weekly demand from 200- 2000 and average ageing greater than 20 (encircled)
• For most of the high demand items, ageing is less than 7 days.

Advantages of Demand Profiling

The various advantages of demand profiling in various segments can be summarized under the following heads: –

• Local Manufacturing

1. High manufacturing flexibility to respond to customer demand
2. Almost zero FG inventory and high component inventory
3. High agility in upstream supply chain

• Global Supply

1. Stable demand, so high LT can be afforded, high delivery frequency
2. Initiatives on logistics optimization as volume and frequency of delivery are high
3. Aggregate component demand planning for global supply chains

• Delayed Differentiation

1. Postponement set up closer to customer
2. Less FG inventory and more sub-assemblies and component inventory

• Supply from Stock

1. Warehouse acts as buffer to absorb customer demand variations
2. Number of supply chain legs (echelons) required for each SKU will vary

Conclusion

In most of the organizations, demand profiling techniques have evolved on an ad hoc basis, over time, from the ground level without any pre-defined supply chain metrics. The result is an ambiguity of functionally specific metrics, calculated in different ways in different levels of supply chain as and when the need is perceived that limits visibility required for demand forecasting. The demand profiling technique helps in moving to a portfolio with aligned supply chain metrics at three levels – the executive level, the end-to-end supply chain view and the deeper functional metrics beneath it, thus enabling the companies to effectively manage end-to-end supply chain performance.