Amazon Stores is a self-service platform that allows brands to create a digital shop within Amazon. I developed Recommmendations — a new feature within the Stores Dashboard that offers intelligent tips in context.

Understanding Structure

To kick off the project, I started by diving into the Stores dashboard, seeing the type of data and affordances that existed for Sellers.

Creating Recommendation Types

With clarity on types of data available, I developed a round of  possible recommendations: combining multiple sources of data to offer intelligent insights to Sellers. With these tips I also prototyped an array of MVP components for these Recommendations. 

Adding “Opportunities”

This was the first entrance of this recommendation feature.

Evolving into “Tips & Insights”

To make the nature of the insight feel more personalized and in context, we evolved into adding a badge style icon to these suggestions.

Making Recommendations

The team was looking to start exploring Recommendations that would feel more like a full feature, take up its own real estate and visual style, and feel like a guided onboarding affordance at each phase of designing the store. We also wanted to start  incoporating the new Storm design system.

Putting Recommendations in context

After the component was made, I explored the ingress opportunities for users to see and act on these tips. 

Building a Recommendations Dashboard

To give Sellers an opportunity to interact with recommendations beyond their in-context appearance, I prototyped dashboard ideas.

A first pass at a dashboard sorted Recommendations by a Seller’s goals. It only had a view-functionality for previously-shared tips.

Another pass at a dashboard was given more affordances for long-term continued engagement with the feature. It was arranged into 3 focuses: seeing new recs, results from recs that were tried, and revisiting previously dismissed recs. A filter affordance was also added to quickly find relevant tips.

A timeline was brought into context, showing a history of tried recommendations and their associated results.

Remaining questions on data collection, the parameters of the Recommendations algorithm, and how to accurately show results from tried recommendations were being explored for a future pass at this feature.