Targeted advertising can be a big challenge for businesses especially retailers. For any business, advertising is a necessary evil. They need it to retain the customer, but they have to pay the heavy costs of advertising and irritating customers by promoting them products they might not need. Targeted advertising, which ensures that customers get advertisements / promotions for only those products they need or are most likely to buy, helps businesses on two ways. One, it reduces the advertising cost – since now you are sending your promotions to lesser but more relevant number of people, and second, it reduces the chances of irritating customers since you are sending them the promotions they really want.
In this project, we have tried to solve this challenge of targeted advertising for the retailer. We have tried to predict the most popular dairy level item in a basket based on customer’s demographics information and past purchase patterns. We have used ‘K- nearest neighbours’ and ‘Categorical and regression trees’ Methods to create a classification model. Both the models have helped us achieve prediction with around 27% overall error.
We recommend the retailer to use this prediction model to run targeted promotions such as discount coupon campaigns for the predicted product. The predicted intelligence can also be used to run promotions for new product launches for the item a customer is most likely to buy.