Why people buy what they buy
- Consumer Data Model and Visualization
- Predictive Analytics
Today, 150 million times a day, someone somewhere chooses a product from CPG-giant, Unilever. Marketers at the firm need to identify the core drivers of these sales: why that product and not another one? Will it help scale? What happens if the trend continues? What would stop the growth?
You have a hankering for grape juice, but there’s none in your fridge. How do you buy one? Do you go to the supermarket or shop online? How do you know which one to buy? Do you go for your favourites? Or perhaps you are in the mood for something new? A seemingly simple decision, to buy fruit juice, masks a complex series of interactions, heuristics and influences that drive consumer purchase.
For Consumer-Packaged Goods (CPG) companies this is an unprecedented growth opportunity. They are witnessing a shift in customer habits, fragmented sales channels, the rise of smaller, fast-moving competitors, and a massive move to online purchasing. At the same time, they have access to data marketers in the past could only dream of. In the hypercompetitive CPG world, developing a nuanced understanding of customers is mandatory. Companies must develop effective systems that identify meaningful insights that inform their product development, marketing, and distribution processes.
To help find answers to these questions, we employed advanced data predictive analytics that moves through the marketing and sales cycle and capture a complex network of interactions and connections, unveiling more granular and real-time customer insights. Unilever is one of the first companies in the world to recognize the potential of consumer data, leveraging big data successfully. They launched a high-impact analytics project to develop a holistic model of shopper behaviour. It considered heterogeneity in their demographic profile, as well as past responses to pricing, promotion, and marketing strategies.
Altruistic modelled and analysed a large, disaggregated panel data from supermarket transactions to develop, implement and validate an agent-based model of consumer choice.
Predictive customer analytics takes the guesswork out of market making. They help Unilever create a better customer experience, innovative novel solutions and refine existing ones, optimize supply chain operations, and zero in on effective campaign communications.
Following the development of the computer simulation model, panel data were used in the initialization of agent characteristic preferences, calibration of parameters, and subsequent testing of out-of-sample predictions.
Out-of-sample tests gave the team greater confidence in the ability of the model to predict real-life sales. The validation exercise involved statistical techniques comparing the outputs of the simulation for three months to the real data covering the same period. The reality is, by having expert data scientists dig into valuable data, you can uncover a world of customer behaviours to predict buyer behaviours.