Business Problem:
Instacart wants to understand which of products are bought together by their repeat users to more intelligently drive their recommendations and offers.
Approach:
Network analysis was used to understand how each product relates to one another. This allows us to understand which products are most "well connected" and which group types exist.
In this graph each product is a node and every time two products were purchased in an order counts as a weight in the edge.
Data:
https://www.kaggle.com/c/instacart-market-basket-analysis
Takeaways
- Instacart repeat users buy mostly produce, with 76% of all orders having produce is them
- Looking at products, Bananas both single or in bag are the most well connected products by the number of degrees. Bananas alone connect with 40% of all other products.
- Looking at products that are bought together by department, produce is most bought with itself and with egg dairy
- 79% of all users bought bananas and they have higher order size (~ 13 products per order) than users that did not buy bananas (~9 products per order)
Recommendations
- For new users, the advantages of buying produce from Instacart can be displayed in ads or in the app as first recommended.
- For current users to stay in the platform, the price sensitivity of produce have to be kept into account. Produce is the reason they stay and buy other products.
- Realize the importance of bananas in the Instacart platform in order to run promotions or keep the price low.
Next Steps
- Get prices for each product in order to see which departments earn more money
- Get more data to compare new customers vs returning customers, most data here was returning customers