At a middle eastern grocer, a RETAILIGENCE POC was commissioned to find out the customer decision tree followed by customers when they buy meat. Advised by external consultants experienced in the developed markets of Europe, business had setup the CDT as follows:
This was reflected in the way the product was merchandised in the stores and was generally accepted as common-sense. However, RETAILIGENCE machine learning algorithms, which go through customer baskets and convert each basket to a mathematical vector, seemed to suggest that for meats, customers tended to shop by brand. Quite understandably, this was dismissed by the leadership as nonsense.
Nevertheless, they agreed to conduct a customer survey among the customers visiting the store in Red Sea Mall, Jeddah. Much to everyone’s surprise most customers said that if the beef of their brand was not available, they would switch to chicken of the same brand. This view was supported by several store staff members who also confirmed that the brand they trust for buying meat is non-negotiable.
Sometimes the most compelling conventional wisdom can be found to be out of date or out of context
The above demonstrates that sometimes the most compelling conventional wisdom can be found to be out of date or out of context. It is best to let the data speak for itself rather than make assumptions. The more we let machine learning run unsupervised, the greater are our chances of overcoming human biases and preconceptions.
To see how a revised assortment and display where products were assorted by brands rather than protein, helped increase sales, please get in touch at
admin@retailigence.com