Rufus

Duration

April 2023 - Current

Rufus is an expert shopping assistant trained on Amazon’s product catalog and information from across the web to answer customer questions on shopping needs, products, and comparisons, make recommendations based on this context, and facilitate product discovery in the same Amazon shopping experience customers use regularly.

Roles

UX Design

Partners

UX designers, conversation designers, product managers, engineers, scientists

You may have also heard about it from these folks

What did I do exactly?

The two biggest things I’m responsible for is being a design owner for both product comparisons and a shop-able component. Prior to launch I was selected to be part of various “swat” teams, where a small team of designers had to generate solutions to critical problems under shortened timelines. Also, when I first joined the team and the project was in the early phases, I also worked on multiple presentation decks that were reviewed with the CEOs at Amazon.

Who were the types of partners you worked with and how?

Designers: Day to day I worked closest with the other UX & conversational designers within Rufus team to review explorations and build various CX’s.

Product managers: Within my verticals of ownership (comparisons & shop-able components), I worked closer with product managers to prioritize action items, roadmap changes by key delivery dates, and align on the final design treatments to present to other partners.

Engineers: As CEO, VP, or director level feedback would be communicated through bi-weekly reviews prior to launch, I would work with various front end engineers to modify components or CX's based on the changes requested.

Scientists: For a larger vertical like product comparisons, I would review various proposed CX explorations with scientists to gauge the feasibility and create a work backwards plan to eventually land on an ideal customer experience.

Here are a few of the bigger things I contributed

1. Comparisons

May 2023 - Current

Myself and my manager were the two UX design owners for comparison responses for project Rufus. The two core CX’s we delivered were 1. broad entity comparison, and 2. product comparison.

Broad entity comparison is higher up in the shopping funnel when customers are trying to understand the differences between larger groups or types of products. Queries like “What the differences between a drip and pour-over coffee makers?” or “Compare air fryers vs toaster ovens”. This are comparison based questions, but are not referring to one singular product.

Product comparison is lower down within the shopping funnel, once the customer has identified specific products they are interested in. When a customers enters in a question like "Compare the Apple Watch Series 7 vs the Series 8" they want to understand the differences, tradeoffs, and pros & cons between those two or more products.

Highlights

Ownership as a junior designer: Shortly after I joined the Nile design team, my first senior manager who was also owning comparisons with me at the time suddenly left the company. Which left me in the position as an L4 junior designer to learn on my feet what it meant to be a design owner for an important vertical within a high visibility project. During the few months where I was the sole owner, I collaborated with the user research team to test multiple product comparison concepts via usability studies to gain insights from customers. This knowledge gained helped inform me and my product manager partner build our proposed CX's for launch and roadmap for the year.

Collaborating with base page teams: After a new senior designer was brought in to work on comparisons with me, we made it a priority to collaborate with the personalization team at Amazon. This team owned a comparison widget on the product detail page, where customers can compare via a table the product they are currently viewing with similar recommended ones. Together we explored a number of integrated solutions to avoid a duplicative or contradicting comparison experience between their widget on the base page, and a Rufus response within the conversation layer. This was an ambiguous space to design for that challenged me to effectively communicate between partner teams and present to leaders. This effort led to a deck that featured a refined list of near term and future state integration concepts that was reviewed with the org’s VP and other Rufus leadership.

2. Category Cards

May 2023 - Current

I created, defined, and delivered one of the core shop-able design components within project Rufus called category cards. These would be included in responses to upper shopping funnel research based questions like (“What are the different types of coffee makers?”, “Compare air fryers vs toaster ovens”, etc).

When a customer tapped on one of these cards, they would be taken out of the conversation layer, and on to a new search results page of the relevant product category shown. Giving customers these actions to be able to move farther down the shopping funnel is what we considered “shop-able”. These were only 1 of only 3 shop-able components we had within the initial launch for Rufus.

Highlights

From ideation to creation: While working on a CEO presentation for project Rufus back in June 2023, I created the initial idea for what would eventually become category cards. At the time we didn’t have a way to easily allow customers to ingress into those various different product categories that are mentioned within a response. The initial concept of text links for each categories imbedded in the response were not ideal, making it very easy for the customer mis-click and tap on another product category. This created a need for a design component that would more easily allow customers to ingress into each product category mentioned in the response. I went out of my way to explore a few different options for visual components, which ended up making it into the final deck that was reviewed by CEOs in June of 2023. Since that point I’ve been the sole owner of the component. I’ve farther refined it based on the feedback from senior leaders, and worked with front end engineers and scientists to rapidly build these requested changes into the internal beta app and initial launch.

Card variations & framework: Beyond tactical improvements to the design, I discovered that a one-size-fits-all approach to category cards was no longer optimal. The initial design allowed for a long text description beneath each product category. However, some categories don’t require a long description, and some may not need a one at all since they're self explanatory. I then created two more iterations of category cards beyond the initial design, in order to have flexibility within the component to best present the content included in the response. This led me to create a framework that outlined to partners, which of the three card iterations we would use for different types of responses. For example, the response for "What are the different types of coffee makers?" we would use the category card variation with a shorter description to provide a glanceable high level overview of the category. Where for a response to "Are cabin tents or dome tents easier to setup?", we would use a different category card variation that allows for a longer description to be able to explain the nuanced differences between the two brands. I drove cross functional reviews of this framework to drive alignment between designers, engineers, product owners, and scientists.

3. Second input field

Dec 2023 - Jan 2024

I was selected to be a part of a small team of designers to help create a second text input field for Rufus. This is what customers tap on within the conversational layer to be ask any of their shopping related questions. It operates separately from the regular search field found at the top of the app, and can only be seen when customers pull up the conversational layer to the full height.

Context

At the time we were considering having the standard search field at the top of the Amazon app to be the only input to have a conversation with Rufus. For example if a customer entered in a query like “What do I need to play pickleball” Rufus is able to recognize it as an natural language information seeking query, and surface a response over the search results page. If they wanted to ask a follow up question like “What’s the difference between graphite and composite paddles?”, they would have entered that into the same search field at the top. However, one of the biggest pieces of feedback from research was that participants didn’t understand that you would use that same search field at the top to also ask a follow up question. They thought they were limited to the suggested follow up questions found at the bottom of the response to be able to continue the conversation.

Creation

This created the need to rapidly generate ideas for a second text input field within the conversational layer as we were only two months away from the planned launch. As the only junior designer on the team, I was a core contributor by generating multiple high fidelity concepts for this second field. I worked with the team to make behavioral and mechanical recommendations for the field to senior leaders, and organized decks to present to design leadership for feedback. In preparation for a CEO demo in late December, we worked with engineers within a week to build a separate version of the beta app that included this second text field so leadership could interact with a live build. This effort required me to engage in daily cross-team mechanisms with effective communication to ensure a working build within the shortened timeline.

What do you think you did well?

01 I created, defined, and delivered core design components within a large high visibility project, while driving cross-team alignment as an owner. Being able to see an idea I had for a leadership presentation deck turn into the one core components (category cards) for a project like this was an awesome experience.

02 I delivered design solutions within complex and ambiguous areas of ownership, effectively communicate with leaders and partners, while maintaining advocacy for customers. Comparisons was a challenging vertical to work on within Rufus for a variety of reasons. It being an inherently cognitively heavy experience to design for, especially with limited screen space on mobile. Also because there were already teams working on it within the base page, I had to learn the soft skills of having to align with them to great a CX both would be happy with.

03 I highlighted my ability to work within a small team to rapidly generate solutions for critical problems, make recommendations to senior leaders, and aid in delivering results within an extremely expedited timeline. Working on the second input field design “swat” team showed me just how much you can accomplish as a team within a couple of days. When you and your partners are faced with an sudden hard deadline, you find ways to improve your communication and overall efficiency in the way you work.

Where do you think you can grow?

01 Being more vocal, trusting my own intuition, and advocating for customers for critical design decisions, even when senior leaders may have a different opinion than me. Whenever I was in reviews there was always a lot more executive leadership included that I had experienced before. I was too hesitant to speak up at times debating certain CX decisions, even when I had some kind of customer data to support myself. I feel like I have a strong design intuition, and I didn’t voice it enough out of fear of disagreeing with a senior or executive leader.

02 Advocating for change when I have strong indication there could potential be a problem down the road, and not waiting for my managers, senior leaders, directors to call it out first. I felt like I often hesitated to flag potential issues or roadblocks ahead, and instead waited until more senior leaders to identify them as something that needed to be resolved. Being the most junior designer on the team and lacking some of the career experience my teammates had, I didn’t want to overstep my boundaries. Looking back I could have been more proactive in identifying these issues.