Recommender system: FreshDirect
The online grocer FreshDirect has several features from a recommender system that appear at different places on the website. These different features seem to be designed for different phases of the shopping experience.
Some of the features include the following:
On the home page, “Easy Reorder”: this seems to present the most frequent items in my purchasing history. This makes it easy to re-order items that I’ve bought in the past with only a couple keystrokes from the home page.
When viewing a product category, “You Might Also Like”: this seems to suggest items from prior purchases that might go with the product category being viewed. For instance, when I viewed the whole fish category, this feature presented several items that I bought recently, including eggs, blueberry scones, and chocolate croissants.
When viewing an individual item prior to purchase, “Like That? You’ll Love This”: this seems to suggest new items that are similar to, or that might accompany, the item being viewed. For instance, when viewing a rib eye steak, this feature suggested a NY strip steak, organic boneless skinless drumsticks, and Wensleydale cheese with cranberries (?).
When viewing your cart and checking out, “Did You Forget” and “Your Top Items”: the first feature seems to suggest items that were viewed but not bought, while the second feature shows common items from my purchasing history. These are both nice reminder features that might prompt last-minute purchases prior to checking out.
Scenario design analysis
For the organization
- Who are your target users?
- FreshDirect is the client, i.e., the organization that benefits from the recommender system.
- What are their key goals?
- Generally to maximize customer sales and revenue.
- More specifically, to solicit and trigger incremental purchases that might not have occurred without the recommender system.
- Also to deepen and broaden the customer relationship by enhancing the online buying experience.
- How can you help them accomplish those goals?
- By customizing the recommender system to their needs and objectives.
- By effectively increasing customer sales and purchase activity, in a meaningful and measurable way.
For the customer
- Who are your target users?
- Viewers and shoppers on the FreshDirect website.
- What are their key goals?
- To have an efficient and smooth shopping experience in general.
- To see value-added suggestions that streamline the shopping process. Examples include:
- Filling out a shopping list with likely other purchases, e.g., if the customer is viewing breakfast items, then suggest other breakfast items like milk, eggs, or orange juice.
- Rounding out a menu with complementary items, e.g., if the customer is viewing steak, then suggest spinach, asparagus, and potatoes.
- How can you help them accomplish those goals?
- Provide features in the recommender system that will make suggestions pertinent to their viewing and purchase history, and that accounts for the purchasing behavior of other similar shoppers.
Suggestions to improve the recommender system
- The existing recommender system seems to:
- Rely mostly on the customer’s prior purchasing history, and make suggestions from items frequently purchased in the past.
- To a much lesser extent, recommend items based on what other shoppers of that item have purchased in the past.
- Suggestions to improve the system include:
- Leverage collaborative filtering or other algorithms to make more suggestions based on:
- Other items that are similar to the item being viewed
- Other items bought by other shoppers who have bought this item (similar to the Amazon approach of “item-to-item collaborative filtering”).
- Provide features or user-defined settings that allow the shopper to dial or customize what kind of suggestions are made by the recommender system.
- Suggest wine pairings when a shopper views steak or fish!! This seems to be an easy one (unless there are legal / regulatory reasons they can’t solicit wine sales online).