Geest bagged salad case study
As part of our initial research we were tasked
with investigating the weather effect on a range of bagged salads. This
is an example of one of those products.
Stage 1: Specific raw sales data and other information was collected and aligned with a range of specific weather data (both individual and combinations). The data was cleaned from specific "noise".
Stage 2: Using the 1st Clean data we built a correlation matrix of sales vs a range of weather parameters to identify the Key Weather Driver (the weather element or combination of weather elements with the highest correlations).
Stage 3: Using the Key Weather Driver (KWD) we built a model (V1). By comparing the difference between Model V1 and Actual gave the ability to question the reason for major anomalies.
Having identified "Explainable Noise" the data was then cleaned again.
Stage 4: This process was repeated until all explainable noise was dealt with - in this instance with Model V4.
Stage 5: Cost benefit analysis of the Model. Using the Feel Good Factor as a very accurate measure of consumer behaviour we were able to use this model as the core driver of sales with the understanding that issues like Price, Promotion and Events were just influences on the sale. By comparing the model with total sales we were able to identify many Out of Stocks and partial daily sales. In this case 3.5% OOS which represented £560,000 lost sales. Had our Model V4 been running at the time it would have recovered £456,000 (2.85%).
Once a core driver model has been built it is possible to reinsert data for Promotions and Events to evaluate their effectiveness leading to cost analysis of Promotion expenditure against uptake.
