Efficiency / US / Retail & Consumer Goods
Using predictive analytics to anticipate changes in coffee consumption
"Our automated approach to forecasting consumption behaviours improved the clients' results at a rate of 50x faster than manual approaches."
Business issue
Our food and beverage client had been tracking attitudes and behaviours related to coffee products through surveys for several years. They asked Ipsos if we could leverage this data – the how, where, and why people drink espresso – to anticipate potential changes in consumption patterns.
The client sought automated predictive analytics; in this case, insights allowing them to understand how to position their products to sell more to customers and stay ahead of competitors.
Our solution
Ipsos implemented automated time series forecasting to predict sales outcomes. This method involved testing scenarios – “what ifs” – such as “What happens to home consumption if restaurants sell 10% more of my products” and “If people spend 50% more time in the office, how does that impact my sales?”
We then took a data science approach as it made forecasting scalable: testing and analysing results for 3,000 scenarios in just two days.
This approach and final report relied on multi-source data integration, including text analytics on survey data and consumer feedback on social media and news sources, to explain the “why” in the forecasting results.
For more information on this case study, get in touch here.
Impact
Our automated approach to forecasting consumption behaviours improved the clients' results at a rate of 50x faster than manual approaches, and with predictive accuracy falling within the 5% margin of error. Our client noted the forecasts were “spot on” for the year.
Analytics on non-survey data also provided context for forecasted behavioural shifts, such as flavour profile preference.
The findings stimulated ideas for business planning, including aligning production and distribution to anticipated preferences and consumption behaviors.
