Unlocking the value of a demand driven supply chain by integrating demand sensing analytics with Supply Chain Planning

Print This Page

April 28, 2017 - In the quest to build a responsive and profitable supply chain, companies have long been trying to predict demand accurately, and synchronize supply (production, purchase and inventory) with market demand. In-spite of spending millions of dollars in best of breed planning platforms, forecast inaccuracy is still plaguing the industries.

Figure-1 depicts the evolution of forecasting methods and practice in three stages:

  • Stage-1: Traditional forecasting took top down approach to forecast sales (and not demand) at an aggregate level and then disaggregating the forecasts to lower levels, resulting in poor forecasts
  • Stage-2: Consensus forecasting took bottom-up approach to forecast demand using statistical models and collaborating with cross-functional teams to arrive at ONE plan. This was the first major step towards a demand driven supply chain
  • Stage-3: Demand sensing takes the short term demand data to improve the accuracy of forecasts based on the current realities of the supply chain

Figure 1: Evolution of demand forecasting 

SeeClose detailed view

The need for a demand-driven supply chain stems from the fact that markets are becoming highly competitive, with a high degree of variability in demand, low customer loyalty, changing needs of personalization across markets, and increasing supply chain complexities. This is calling out for higher velocity of planning (from months to weeks to day to hours) and tighter integration between planning and execution, in order to respond faster to changing market needs.

Integrating demand-sensing analytics with supply chain planning holds the promise of linking planning and execution by sensing real-time demand signals from markets. Figure-2 depicts how this data can be harnessed to generate insights for course correction on decisions related to demand-supply gaps, promotion timing, allocations etc. using a “Sense -> Analyze -> Respond" model.

Figure 2: Demand sensing model

SeeClose detailed view

Sense - At the simplest level, sensing would mean capturing POS or sell-through data and key demand-shaping drivers, such as trade promotions and marketing campaigns, from internal data sources. Sensing can also include external data sources (market data), such as social media, weather data, and competitive intelligence.

Internet of Things (IoT) and big data technologies have made it possible to capture and store real-time granular data, which can generate effective demand-sensing insights by analyzing the latest demand patterns and shifts. For example, IoT can be used for inventory reallocation decisions, by capturing real-time POS data and analyzing the latest rate of sale across channel partners. IoT can be used to prevent costly out of stock situations by analyzing this rate of sale and tracking real-time inventory levels across the supply chain for deploying inventory to the right distribution centers and markets. 

Analyze – Once you have the data, you can test various hypothesis using statistical analysis to better understand demand drivers by markets and product portfolios, demand shifts (structural break points) over a time period, and predict short-term demand for forecast adjustments. Questions that could be answered include:

  • Is there a significant misalignment between demand and supply?
  • How can we align/adjust forecasts, production, deployment, allocation, etc., based on the latest demand signals from the market?
  • Is there a correlation between demand and the level of customization or personalization offered?
  • Is there a significant shift in demand across product portfolios and markets?
  • What are the key drivers of demand in each market and across product portfolios?

Portfolio alerts can be set to monitor key drivers influencing demand or demand-supply gaps, based on the results of statistical analysis (studying causal relationship between demand and level of personalization offered, type of promotion, depth of promotion, promotion placement, pricing, forecast accuracy, forecast bias, etc.). For example, an alert can be set to monitor stock coverage (Stock coverage increases or decreases beyond established control limits), demand variability or shifts (Actual vs forecasted values for last the three periods falls outside of two times calculated demand standard deviation) or Forecast Bias (Tracking signal for the last three periods is greater than 4.5 i.e. Over-forecast, or lesser than -4.5 i.e. Under-forecast). In addition, daily (or weekly) changes in demand and supply components can be studied to generate demand-sensing insights.

Respond – Portfolio alerts, demand-sensing insights, and short-term forecasts can be used by different teams to respond faster to demand signals.

  • Sales and Marketing – By understanding the demand drivers and demand shifts, sales and marketing teams can run or tweak demand-shaping programs.
  • Supply Chain – By predicting short-term demand, the supply chain can adjust forecasts, production, deployment, and allocation plans to satisfy demand and push out excess supply.
  • R&D or Product development – By studying the co-relation between product attributes, demand, and consumer behavior, products can be developed or modified to fit market needs.

If you are planning to start your demand-driven supply chain journey, start with what you have. If you are collecting POS data (e.g., top-channel partners) or end-customer orders, this could be great starting point for demand sensing. If you are collecting POS/orders data monthly, the next step could be to collect them weekly, then daily, and so on. The more granular the data, the better it is.

The concept of demand sensing can be borrowed by any industry; however, if you are operating in a fast-moving industry with a dynamic supply chain (e.g., consumer products, consumer electronics, automotive), you should start your journey now.

Typical benefits that you can realize using demand-sensing analytics are:

  • Up to 50% improvement in forecast accuracy
  • Up to 70% reduction in excess or unproductive inventory
  • Up to 15% improvement in customer service level

It's time to explore demand-sensing analytics to build a responsive, demand-driven, and profitable supply chain.

This blog is co-authored by Sudhanshu Singh, Senior Vice President and Chief Operating Officer, Analytics and ResearchMani Iyer, Assistant Vice President, Analytics and Research; Vikram Sethi, Assistant Vice President, Analytics and Research at Genpact

Continue Reading

Ready to