Mint Velvet, like all online retailers, had always recommended cross-sell and up-sell products to their customers. But the process to set these up was time consuming and were these the best recommendations to give?
Dragon Drop was tasked with automating the product recommendations in a way that dramatically reduced set-up time, but gave Mint Velvet's online team control over what was displayed to customers.
Four recommendation algorithms were implemented:
- Bought in the same order
- Bought by the same customer
- Best sellers in the same category
- Best sellers across all categories
The recommendations app consumes historical order data on a daily basis to recalculate the top recommendations for each algorithm, for each product on the site.
The Mint Velvet team decide which order the algorithms should be applied, category by category.
So the Skirts category can have "bought in the same order", followed by "bestsellers in this category" and so on. Accessories can have a different configuration - "bestsellers in the same cateogry", followed by "bestsellers sitewide".
When presenting recommended products on the website, the recommendations app works through the configured algorithms in the order specified until it has 4 in-stock products to display to the customer.
Clicks on recommended products are tracked so that Mint Velvet can analyse which algorithm is working for each category, refining the configuration over time.
Technologies, Platforms + Integrations
- React - for UI
- GraphQL - for API
- Meteor - for app platform
- MongoDB - for data storage
- Mandrill - for internal email alerts
The result?
- Fully-automated product recommendations
- Strong engagement in up-sell and cross-sell products based on industry best-practice algorithms
- Happy customers!
Credits
Jonathan Bowen
Founder and Developer