The retail and consumer world is volatile, uncertain, complex, and ambiguous (VUCA) due to changing consumer behavior and product preferences. These cause dynamic changes in the consumers’ shopping carts. The vicissitudes happen in product categories, volumes, brand preferences, and even channels.

Companies are trying to understand, analyze and adapt to the unpredictability of the consumer’s world, which has been further exacerbated by the COVID-19 pandemic. To maximize profitability, companies have transformed from decision-making in silos and sequential planning (that lacked agility) to ‘integrated planning’ across functions. Demand forecasting is of paramount importance, sensing near accurate demand is the foundation on which strategic and operational plans are built. 

Based on the demand forecasts, companies develop and create financial plans, pricing policies, manufacturing schedules, sales and marketing strategies, capacities and infrastructure expansion plans, manpower plans, and capital expenditure forecasts. 

The volatile consumer behavioral attributes have made accurate demand forecasting a nightmare. Consumers are more likely to stockpile and hoard goods (especially essentials) due to the pandemic restrictions and unpredictability of supply. This, in turn, leads to ‘stock outs’ at retail stores and fulfillment centers for specific categories; thereby creating a bull-whip effect - a pseudo surge in demand. In order to respond to this wrongly sensed demand, companies put undue pressure on their production lines, compelling the supply planning function to reduce the lead times of raw materials. This is a daunting task considering the already disrupted supply chains due to travel restrictions, unavailability of logistics and labor shortages. 

Dynamic pricing and omni-channel servicing of consumers in a short turnaround time is a standard practice in the new retail and consumer landscape. Demographics and location intelligence parameters have also become very important, as there is a shift towards local buying. 

Keeping these changes in view, the traditional ways of demand forecasting (based on historical data and exponential smoothing application on time-series analysis) which involve manually analyzing data from across the supply chain, do not make the cut for accurate forecasting. The traditional ways are inaccurate and time consuming.  The dynamic changes require adding variables and sources into the forecast on a continuous basis. A sophisticated forecasting model (enabled by machine learning algorithms and artificial intelligence) combined with ‘big data’, could be the answer for timely and accurate forecasts in the consumer landscape. 

A hybrid model that combines the old methods and new age technology is the need of the hour as it effectively bundles internal data such as historic trends, data from points in a value chain, distribution centers, payment gateways etc.  It also incorporates external data such as geo location based news feed, social media analysis, ongoing promotions, GDP, weather conditions etc. The data can be structured or unstructured. An eclectic combination of digitalization, structured operations and appropriate platforms make it possible to combine internal and external as well as structured and unstructured data points to create more realistic and timely forecasts.

Data can be broadly classified into 4 data pools as under:

• Structured and internal: This data pool is easiest to capture and analyze. For example: sales transactions and inventory
• Structured and external: With the correct tie ups and using appropriate databases, this data pool can be leveraged through minimal effort. For example: government census and paid market reports
• Unstructured and internal: This is the data bank that usually goes unnoticed and hence remains untapped. For example: reviews received from clients on emails and contact center call recordings
• Unstructured and external: This data pool, being most difficult to capture and analyze, is more often than not under leveraged but it holds undeniable potential to create actionable insights. For example: social media and data from internet of things

If the forecasters can accurately capture and analyze all the above data pools, it is like achieving business nirvana.

Figure 1: Envisioned forecasting model

Accurate forecasts as per stock keeping units (SKU), channels and regions help facilitate relevant inputs to other functions, which can help create function wise plans to meet the demand effectively.

A near accurate forecast answers fundamental questions like - what to make, how much to make, how to sell, where to sell, and whom to sell to. This boosts overall operational profitability, enables efficient production planning leading to efficient alignment of resources and workforce, allows better control over damaging events like stock-outs that cause customers to switch to other brands, avoids production of unsaleable stock that have high holding, and helps mitigate the bullwhip effect. This also ensures an effective response to the demand volatility enabling top line and bottom line growth.

Anubhav Billore

Manager, SCM Practice, Digital Operations & Platforms

Anubhav has around 12 years of industry experience and currently focuses on the retail, CPG and logistics sectors in his supply chain practice management role. He has worked for several customers on process and performance improvement projects across the supply chain space. He has diverse experience in management consulting engagements, technology strategy, business process re-engineering, customer experience enhancement, and change management programs.

Rishi Reejhsinghani

Deputy Manager, Presales and Solutions, Digital Operations & Platforms

Rishi has 4 years of business strategy consulting experience with expertise in business study, business process re-engineering solution designing and project management for customers in the retail and consumer domain.