Customer demand forecasting for food delivery client
Multi-level time series models built to plan resourcing at market, store and hourly levels.
The challenge
The client was a new competitor in the food delivery marketplace – with a very strong growth trajectory. They had an existing forecasting tool but wanted to extend the capabilities and improve the accuracy of predicting customer demand within their two key markets UK and Netherlands, primarily to allocate staffing and resource more effectively.
The solution
We combined various data sources including customer sales, marketing, CRM, plus 3rd party data such as weather to build predictive models per market, per store and per hour within each store. The multi-level models achieved very high accuracy of 97% and 96% at daily levels respectively per market, utilising an ensemble of time-series and regression-based modelling techniques.
The result
The outcome was two simulator tools, which hosted the time series models and allowed the client to forecast sales at the various levels of depth with high levels of accuracy. It also allowed the client to test various “what if” future scenarios particularly useful for marketing spend allocation going forward and CRM planning. A deep understanding of the causal mechanisms driving sales was delivered as a full report, giving direction on what the key drivers of sales are and how these can be influenced to drive sales.
If you have any data analytics questions or challenges please get in touch with:
Unbelievable impact in a short space of time. Analytics across diverse data was an eye-opener, providing clear direction to take to the CEO
Craig Milligan
Global Head of Consumer Analytics & Insights | Decision Sciences, Diageo
