ARMOIRE
OVERVIEW
Armoire is a leading company in the fashion rental industry, offering clients personalized clothing selections through a machine-learning algorithm. Recommendations are tailored to each user's preferences. I was part of a team that redesigned their new-user onboarding process. The challenge was to enhance the new-user experience: if users subscribed after one month, they tended to be long-term customers. We ended up not only making their onboarding faster, but gamifying it to accelerate the data new users start with, which will increase their likelihood of becoming long-term customers.
DOUBLE DIAMOND PHASES
KEY CONTRIBUTIONS
-
Applied best UX practices to optimize the new-user onboarding experience.
-
Collaborated with engineering and data-science teams to determine success metrics and design constraints.
-
Conducted usability testing as part of iterative design process.
BACKGROUND
To optimize Armoire's onboarding process, I began by conducting usability tests that revealed a significant issue: new users were often skeptical of subscribing when the initial garment recommendations didn’t align with their style and fit preferences, leading to a negative signup experience.
After collaborating with Armoire's data science team, we discovered that the existing signup process failed to collect sufficient style preference data, resulting in less accurate initial curated offerings. To address this, we aimed to redesign the signup process to better serve both the data science needs and enhance the user experience for potential clients.
Leveraging Human-Centered Design principles and an iterative design process, we focused on improving the signup experience by capturing users' sizing needs, style preferences, and relevant algorithm data points through engaging and delightful interactions. Success was measured by increased user retention during the critical first two months of subscription.
​​
EXISTING METRICS
Armoire's current signup process is 47 screens long. It incorporates users' style preferences, fit, and personal information. Many users felt that the current signup process is too long, does not leave them feeling positive, and ultimately makes them question the value of Armoire's service.
We asked users to rate their satisfaction with each screen on a 1 to 5 scale, as reflected in the following graph which captures their emotional journey through the current signup process:
Figure 1.1: Aggregated emotional map of user journeys through the existing signup process. Each dot represents a screen, highlighting a significant drop in user sentiment just before the critical email entry screen.
To solve this, we decided to split the signup process into two stages: style and fit.
​​
In the first style stage, users complete questions to determine what their style needs and preferences are. We aimed to make this stage fun, and to capture the most pertinent data required to generate a curated closet using Armoire's algorithm. Since Armoire does not carry every item in every size, users are required to enter some sizing information in this section. We condensed this into one simple screen so as to not disrupt user flow in this section.
​
After completing the style stage, users can choose to see their curated closet immediately, thereby shortening the onboarding time but reducing the curation accuracy as there were limited data-points. Or, they could also continue with their signup and fill out their FIT requirements.
​
We were able to reduce the total signup process from 47 to 34 screens split into two stages, including four transition screens between that did not require user input. We were also able to capture highly relevant user style preferences as required by the data science team, which will be discussed in detail later.
Fig1.2: The aggregated user journey map of our redesigned signup process. Users remain satisfied and engaged throughout the process. It culminates in user excitement and anticipation to see their curated closet.
INTERACTION DESIGN - KEY SCREENS
The wireframe prototype process went through four rounds of iterations based on usability tests of five-to-seven participants per round. Key screens from Armoire's current signup process are shown below, followed by relevant, final wireframes.
MATERNITY WEAR
​
Unlike their competitors, Armoire's first screen of their signup process includes a separate user flow for maternity wear. We decided to remove this screen, and incorporate maternity wear within the signup process, as users felt it was either exclusionary, or that they would be missing out on certain offers based upon their initial choice at this screen.
STYLED LOOKS
​
Armoire allows users to select their styled looks preferences according to six categories. However, the given categories are relatively ambiguous as they are text only - 'office appropriate' is not the same for all work places. Hence, we aimed to incorporate a collection of stock images that represent a given look.
EVERYTHING ELSE
​​
This screen was confusing to users, as it combines three unrelated categories: an area of the body they are comfortable showing, a bra preference, and a fabric preference.
​
This screen was expanded into an 'areas I'm proud to show off' screen, a 'fabric preferences' screen, and the strapless bra preference was included into a 'bra size' screen.
CREATING "STYLE UP" TO ASSIST DATA SCIENCE TEAM
We developed a better understanding of Armoire's algorithm after meeting with their data science team. Unfortunately, Armoire's initial customers are not presented with an accurately curated selection. The more users interact with the service - by returning, rating, liking and browsing items - the more accurate their closet curation becomes.
The current sign up process uses a very limited set of data points, and as such many new users are dissatisified with their initial curation. However, once the algorithm 'catches up', usually at the two-month mark, users are likely to subscribe to Armoire indefinitely.
​
We created this simple visual to illustrate the effectiveness of Armoire's data capturing below:
Fig3.1: Armoire's algorithm operates effectively when it is populated with a lot of user feedback on individual items. The current signup process does not capture enough data. We therefore aimed to take collect more user-preference data than the current signup process to create a more accurate initial closet.
The data science team asked us to find a way to gather ranked and weighted data points for specific items. We wanted users to enjoy this process, too, and so we developed 'STYLE UP'. Users are presented with two similar items from Armoire's inventory, and simply select whichever they like more. We also gave users the option to select 'neither' to prevent false positives.
STYLE UP
​
Users enjoyed going through the Style Up process. We designed it to work with automatically generated images from Armoire's existing inventory.
Armoire encourages users to explore styles, and so instead of asking "What styles do you wear?", we deliberately used open language that would encourage users to think outside of their current wardrobe.
STYLE UP | POPUP
​
After a given number of 'rounds', a popup appears that prompts users to either move onto the next question in the signup process, or to keep 'playing'.
​
Style Up was designed to be used outside of the signup process, too. Instead of browsing other apps, users would be able to open the Armoire app and cycle through 'Style Up' to keep interacting with Armoire and continue curating their closet.​
VISUAL DESIGN: KEY SCREENS
Many new users were not aware that Armoire provides free shipping, dry cleaning, and the ability to purchase and keep items. Furthermore, the user community is active and supportive.
We introduced these screens to better inform new users of Armoire's services, and to serve as 'rest points' in a relatively long sign up process.
SUMMARY
This project showcases the efficacy of partnering UX and data science teams. What was a problem for the data science team, becomes an opportunity for the designer, and ends up creating an improved user experience on two fronts, which in turns positively impacts the key driver for user retention (and in a subscription-based app, that's kind of a big metric).