In the post-COVID19 world, many aspects of business have changed in unique ways ­– and so has the consumer response to goods and services. Companies are increasingly witnessing consumer attention disparity, changing rules of purchase criteria and different consumer responses to promotional tactics. These widespread changes have necessitated a refinement in the analysis of shopper data (both structured and unstructured), choice and purchase behavioral patterns, online activity (both for the consumer and manufacturer) and consumer activism. Fortunately, marketing, sales and behavioral science experts across industries have recognized this opportunity to improve promotional and marketing intelligence and respond to change.

In the fiscal year 2021-22, leading consumer-packaged goods (CPG) companies spent an average of 20-23% of revenue on trade promotions. But the inability to link trade promotion spending and retail execution impact marketing ROI ­– and the customer experience. Many marketers struggle to manage the promotion budget and fail to increase marketing ROI due to ineffective resource allocation to marketing tactics. To resolve this challenge, adopt data-driven intelligence. It can promote the right product, at the right place and time.

Approaching AI-driven Sales and Promotion Planning

The size and scale of the CPG industry are enormous – multiple brands, billions of global consumers, and the supply-chain engine that supports the industry. This entire process generates vast amounts of transactional data. In the past, marketers and business development analysts have analyzed this data with statistical methods that are cumbersome and time-consuming, yielding mixed results. But Wipro is pioneering the art of possible with AI-based trade promotions and marketing driven by data science.

It starts with a dataset consisting of 104 weeks of reasonably contiguous data points to generate and train the AI algorithm. If less data is available, it is possible to start with a smaller data set, and the algorithm can be trained and improved as the repository of data points grows and becomes more integral. Training the algorithm is a phased approach with multiple other factors contributing to its prediction and forecasting accuracy. It is important to note that Data Integrity is crucial and primary for model success.

Data inputs range from basic (sales, scan, promotion history, promotional sales, shipments), mid-complex (marketing tactics, black swan events, unexplained anomalies) to complex (customer reviews and feedback, customer behavior, transaction documents, etc.).

Figure 1: The Promax approach to collaborative data modeling

The process begins with a consolidation of the available data. Next, the data undergoes a process called Data Harmonization, or the method of unifying disparate data fields, formats, dimensions, and columns into a continuous and composite dataset. Data harmonization and validation require significant human intervention.

The harmonized data runs through a modeling engine to provide a base algorithm. Training the base algorithm is ongoing until achieving an acceptable level of accuracy (∼75%). Learning repeats until the level of accuracy (> 90%) arrives at the initial coefficients. These initial coefficients and other base values are the basis of the model file and drive forecasting and prediction capabilities. 

If data for the entire range of products and customer hierarchy is scarce, start using the most critical data. The recommendation is to identify the most impactful contributors to spending and revenue that meet the prediction accuracy.For example, one of the world’s largest brewers in Australia wanted to determine the success of its trade promotion activities. The process was flawed at the outset due to a lack of identification of correct and accurate baselines (less than 60% accuracy) and the associated causals/features affecting the data. Wipro deployed its Promax AI models that looked at challenges like data accuracy to quality and addressed these issues through automation and advanced machine learning algorithms. Using techniques such as feature engineering, trends in the historical data were identified, and new causals addressed the patterns. These improvements in the model helped the company improve its revenue growth by 2%+ and forecast accuracy by 80%+ across product groups, retailers and stock.

Figure 2: Graph depicting the model accuracy of approximately 95% achieved in a beverage business

Extending AI to Marketing Mix and Revenue Growth Management

With revenue growth management (RGM), data-driven models use the same approach described above. Significant achievements in mapping unstructured data sources accurately predict key RGM drivers such as Price Elasticity, Consumer Preferences, Brand Development, Competition Actions, Price Pack Architecture, Product Assortment, Supply and Demand Planning and Promotion Optimization. The best prediction uses the difference between considerable sums of trade spending that reaps exceptional/negative returns and intelligent and connected systems. AI can be a valuable link and a rare and non-imitable resource that gives CPG organizations a competitive edge.

The Wipro Promax solution offers a transparent, collaborative, and expert-assisted model of transforming data into valuable insights. The model accounts for global supply-chain disruptions like the COVID19 pandemic or geopolitical conflicts. These events impact commodity pricing, which will, in turn, affect production planning and promotion forecasts. Wipro’s Promax considers these events, causals, or features and incorporates them into the modeling equation. Factoring in the causal effect makes predictions more realistic and accurate. AI-based models reduce turnaround time and manual effort required for precise planning and allow companies to respond to market changes in near real-time.

Abhishek Banerjee
Global Sales Manager, Wipro Promax

Abhishek is a Sales Consulting Lead for the Wipro Promax business unit focused on Trade Promotion Management and optimization for the Consumer Goods industry. He has worked in the Trade Promotions Expenses (TPx) sector with numerous CPG customers in the Americas, Europe and APAC, helping to identify and solve challenges across TPx landscapes.

Arvind Sachidev C
Lead, Data Science & Machine Learning, Wipro Promax

Arvind is the Data Science & Machine Learning Lead for the Wipro Promax Business Unit. His focus is on Trade Promotion Management & Optimization. Arvind has worked in multiple industry verticals for almost two decades, specializing in the field of AI/ML and Engineering. He has worked with Engineering, F&B and CPG customers to build AI solutions.