Alexa Shopping List

Motivation

As the customer engagement lead for Ultrafast Grocery at Amazon.com, I had the opportunity to work on a project that aimed to make the transition from meal planning to grocery shopping a seamless experience for our customers. To truly make a great grocery list experience, we would reimagine the voice and online experience, allowing customers to quickly capture their need without having to tie it to a merchant.  We would then optimize the purchase experience to provide the best products at the best prices in a single cart and checkout experience at the time customers want to buy. The goal was to make Alexa a best-in-class shopping list for use at any grocery store with integration throughout the Amazon shopping journey for added convenience when shopping online. It was a challenging project as our designs had to be flexible enough to accommodate multiple stores in the future, and multiple shopping modalities. However, we were determined to make grocery shopping a more approachable and enjoyable experience for our customers.

Personally, I was excited about this project because it was the first time in my career I got to innovate on a product category I was already a customer of. I used a variety of shopping list apps, but each one had its own pitfalls. Now I had an opportunity to fix everything that bothered me about these apps. Should be easy enough to design the perfect shopping list, right?

Approach

As it turns out, it was not easy - in fact, this ended up being one of the most complicated products I worked on in my time at Amazon. As I learned in the discovery phase of this product, everyone uses lists differently. That also meant the stakeholders I worked with each had varying ideas of what the perfect shopping list would entail, and often, the only commonality they shared was their strength of conviction. If getting them all on the same page wasn’t going to be a big enough challenge, the existing list infrastructure certainly would prove to be: at Amazon, we had three different teams working on lists, all of which had a different backend and served different customer needs. Alexa List focused on crossing off one-time needs, Grocery Lists focused on saving commonly purchased products for easy access, and Wish Lists focused on bookmarking high-consideration items a customer may want to buy later. The final challenge would be the breath of fulfillment options this list would have to serve: not only would this list have to work for grocery, but it would also have to excel at helping customers shop in-store, as it was meant to be a flagship feature for our soon-to-launch Amazon Fresh stores.

To start the project off, I did what I normally do when I’m completely overwhelmed: gather as much background information as possible, take the time I need to understand it myself, then make it understandable to others by arranging the information in a more easily digestible way. Next, I simply gather the smartest people I can find into a conference room with a whiteboard, present my findings, and get them to start talking to each other. After I ran a few workshops with the three list teams, we had a good idea of how we wanted our unified list to work, and I had to get that shared understanding out of the conference rooms and into a document we could share and start to bring others on board with.

Implementation

Due to the complexity of the approach, I started with a journey map, outlining how the experience would vary based on the ingress point, where in the journey we might start locking customers into a fulfillment decision, and how easy or hard we wanted it to be to pivot once that decision was made. The following artifact helped us firm up our thinking and made it clear how the customer would discover our feature, and what exit points they would have.

Once we had the shopping journey defined, I worked on outlining the different interactions that a customer could take on the redesigned list. I created a one-page document that would give stakeholders a high-level overview of all the actions a customer could take without having to walk them through several different flows.

Of course, one-page isn’t enough for all the details, but I do find I have more success demonstrating how a feature is supposed to work if I can break interactions down into contained journeys that fit onto one presentation slide. Then, when I’m presenting, I can leave that one slide up for discussion, rather than switching through a bunch of different slides or Figma pages and hoping the audience can follow. I like to make a variety of these one-slide flows that show the journey a customer takes through a feature, and this project gave me lots and lots of opportunities to make lots and lots of of these flows. Most often these were just for stakeholders, but I ended up presenting some of these to a variety of directors, VPs, and even SVPs, which really helped me hone my design presentation skills. By the end of this project, I earned myself a bit of a reputation for being a great presenter and fostering productive discussions.

Out of the dozens of flows I ended up making, I have selected a few of the most interesting ones to share here. Below, you can see some early concepts for all the different ways a customer would be able to add to the unified list, and then how they would be able to shop it. (note: I was only responsible for the app and web UI - a designer from Alexa was responsible for the voice UI and Alexa device UI).

To help customers remember to shop for the items on the list while they were out, we introduced push notifications that were triggered when a customer entered the geofence of a grocery store for over a minute. I thought this would be a great differentiator and sure to increase engagement, but as an early beta tester, I realized just how annoying this could be when I got pinged every time I drove past a grocery store. To avoid bombarding customers with unhelpful alerts, I made sure to include a 60-second dwell time within the geofence as a requirement before launch.

A few other in-store benefits I was proud of: customers who shopped at an Amazon-owned grocery store were be able to see aisle locations for all the items on their list, making it easy to find items in unfamiliar stores. And for everyone who has gotten tired of hearing “if I knew you were going to the store, I would have asked you to get milk!” , we introduced opt-in shared lists, allowing household members to easily add items to the list whenever another household member went shopping.

Launch Experience

One of the toughest challenges we faced during the launch of our redesigned grocery shopping experience for Alexa was figuring out how and when to scope customers into a specific store. Our main goal was to make it easy for customers to quickly capture their needs in the moment with minimal specificity and then bind them to a specific product and fulfillment channel later on. However, we also had to ensure that customers didn't end up splitting their baskets between different stores and paying extra delivery fees. To solve this problem, we created a flow that allowed customers to select a store at the top of the page to narrow down product selections. Customers who accessed the list from within a “walled garden” experience, like the Whole Foods Market storefront, would only ever see items from that store presented to them. In this way, we avoided allowing Whole Foods customers being able to add Fresh items, which would then start a new cart with an unmet free shipping threshold.

One compromise we had to make in the migration to Alexa Shopping Lists as the sole list provider for Fresh and Whole Foods was that each list item had to be stripped of SKU metadata and simplified down to a simple text string. We originally thought we would be able to keep this metadata, but cut the requirement to meet our launch deadline. Instead, we ended up simply using Amazon product titles as the list string if a customer saved to their list from a detail page, which lead to some unwieldily text in the shopping list. However, the upside was that, if the original product went out of stock, our bottom sheet search results would be able to pull up closely related products, so customers could still have their needs met.

Post Launch Results & Modifications

The initial launch was a huge success: based on our A/B testing, this new feature drove $17.7MM in Fresh and Whole Foods online revenue. However, due to resourcing constraints resulting to a descoped MLP launch, we still required customers to click on every item in their list to add them to cart, and we knew we could do better with some post-launch improvements.

Later, we tested a ‘Shop Your List in 1-click’ MLP experiment. When this experiment failed to increase revenue, we did a deep dive on why. We ultimately decided the initial ‘Shop Your List’ ingress was non-intuitive, requiring multiple clicks from the Alexa Shopping List to shop from Whole Foods or Fresh, resulting in a low discoverability rate of the Shop Your List feature. We conducted usability tests and found that customers would get lost in the workflow of picking a store brand, fulfillment option, and address before discovering the value of Shop Your List’s online product recommendations and in-store product locations. We introduced the shopping tabs below to make Shop Your List easier to discover by showing customers product recommendations or in-store product locations in 1-tap, while still providing options to change fulfillment options/addresses and easily compare selection and prices across different store brands.

The introduction of shopping tabs increased Fresh and Whole Foods revenue by an additional $492K per year, due to a +390bps increase in the discoverability of Shop Your List from 2.9% in control group to 6.9% in treatment (measured by: # of customers who view product recommendations or product locations / # of customers who view their shopping list in the Amazon app).