Starbucks Out of Stock Recommendations
TLDR; Over half of Starbucks ordering sessions face out-of-stock products, leading to a 24% abandonment rate. We hypothesized that recommending alternatives would improve order completion. An A/B test showed these recommendations could add $9 million in annual revenue and led to additional placements and enhancements to our machine learning algorithm.
Duration: Summer 2023
Role: Lead Product Designer
HISTORICAL CONTEXT
COVID-19 highlighted vulnerabilities in supply chains worldwide and the business impact of these issues. When customers encounter an out-of-stock item, they are presented with an error that leaves them at a dead end. Sixty-six percent of customers encounter an out-of-stock message, and 24% of them never place an order.
USER RESEARCH (ROUND 1)
Method: 60-minute moderated interviews
Participants: Five existing Starbucks MO&P customers who identify as "location loyalists"
MOP helps plan daily routines, but late OOS notifications (e.g., in the Order Menu) led to behaviors like ordering differently or visiting stores for inventory. Customers were interested in a recommendation feature, expecting it to reflect in-store suggestions from baristas.
PROBLEM
Customers faced issues when they encountered an out-of-stock message for a specific Starbucks. They were not notified about out-of-stock modifiers until reaching their cart, causing frustration late in the process.
GOALS OF REDESIGN
The project's goal was to reduce customer drop-off when encountering out-of-stock items in the order-ahead flow and to notify them early if their modifier is unavailable. Additionally, it aimed to create a feature that enhances business resilience during future supply chain issues.
COMPETITIVE ANALYSIS
The purpose of this competitive analysis is to capture the trends for product recommendations in the industry and what our customers are currently used to.
​
-
When a product is out of stock, what can the customer do next?
-
If recommendations are provided, what product context is presented?
-
How is it visually presented?
-
How many recommendations were presented?
LEARNINGS
Where do customers go next?
Most companies display product recommendations in a list behind a button after the customer shows interest.
Product context
When showing a product, it often includes pictures, sizes, and prices.
IDEATION
The main problem I set out to solve was how to present recommendations. I aimed to find an interaction pattern that wasn't overwhelming and allowed space for additional product context. I explored options like an evergreen carousel, a bottom sheet, and a push view.
By testing interaction patterns with various out-of-stock items and additional product details, I discovered that the push view was the least overwhelming and offered the most flexibility for future product context.
SOLUTION
This new feature includes a button next to the out-of-stock message that allows customers to "see similar items." Available on Previous, Favorites, in the cart, and on the product detail page, it enables customers to switch products earlier when they are more open to pivoting.
When the button is tapped, users see a list of up to four items to avoid overwhelm. This feature maintains trustworthiness by providing relevant recommendations based on scores from the machine learning algorithm. Each item includes a picture and product name, with space for future details like calories, size, and an indicator if it was previously ordered.
As part of this project, we are adding out-of-stock messaging for modifiers. Our research showed that customers feel frustrated when encountering modifier availability late in the ordering process. Now, they will be informed from the Previous and Favorites sections if a modifier is out of stock and offered an alternative item.
TAKEAWAYS / LEARNINGS
This project was the first project I was able to own from ideation to completion. This project was an exciting way for me to pull together all my learnings from previous projects to create an important update. If I was to do this project again, I would strive to have more, and clearer KPIs at the beginning of the project, to help measure the full impact of this project after it shipped.