
Carousell Design Challenge: UX Personalization
Carousell is a mobile and web C2C platform for the buying and selling of new and secondhand goods. I was approached by the design team to perform a design challenge around May 2016, and I thought I would share my process for this design challenge with everyone!
My Role
I spent eight days going through my design process to come up with a proposed redesign of their iOS app to solve the problem provided.

The Challenge

Problem: The company noted that a majority of buyers were using the Search bar and not browsing on the homescreen
Assumption: Noted behavior is a result of content on the homescreen not being personalized nor relevant enough to incentivize browsing
Explore the Proble
Before diving into the design process, I studied the information provided to me to extrapolate goals and additional assumptions. I moved forward with a couple of new assumptions as I believed the user behavior observed may have been influenced by factors other than the assumption given to me.
Extracted Goals
- Better the user experience by providing a more relevant and personalized browsing experience
- Make it easier for users to discover and find items of interes
New Assumptions
- A significant portion of users who reach the homepage have the intention of buying specific items
- Users with specific items in mind to be purchased will usually opt to use the search bar
Existing Research
I researched and found existing studies that would validate (or invalidate) the assumptions, as the validity would influence the design decisions made to address the needs of the user:
Assumption #1
A significant portion of users who reach the homescreen have the intention of buying specific items
Research findings:
Researched data suggests that millennials make up a majority of purchase intenders, and roughly half of millenials do not browse online for entertainment-shopping. Assuming this is true of Carousell users, then it is likely that a significant portion of users landing on the homescreen have intent to purchase specific item(s).
Research data:
- Millennials (ages 21–34) make up 49%-59% of global purchase intenders who browse online and 52%-63% of those who buy online
- Almost 50% of millennials regularly browse through items online for entertainment; the remainder do not
- In 2015, 79% of APAC millennials reported purchasing goods online within the last 12 months, which is significantly more than the general population
Assumption #2
Users with specific items in mind to be purchased will usually opt to use the search bar
Research findings:
In a research performed by Stanford University and Pinterest personnel around almost 3 million Pinterest users and their browsing/ purchasing behavior, it was found that users with built-up purchase intent exhibit more searching and click-through behavior at the expense of browsing and content saving
Usability Tests + User Interviews
Without access to the Carousell user-base, I opted to further validate my assumptions by usability testing and interviewing a comparative demographic: millenial, online-shoppers who are users of C2C platforms such as Letgo, Craigslist, and Mercari.
Assumption #1
A significant portion of users who reach the homescreen have the intention of buying specific items
Findings:
Answers collected support the assumption that there is a significant portion of users who shop online only when they have purchase-intent for specific item(s):

Assumption #2
Users with specific items in mind to be purchased will opt to use the search bar
Findings:

Users’ behavior in response to the above scenarios further validate assumption #2, as users with purchase-intent used “search” significantly more than “browse”
Assumption #3
Content on the Homescreen lacks personalization and relevance
Findings:
I ran users through scenarios that encourage browsing during testing and followed up with questions around relevance of content presented on the homescreen:

Research Summary
- Most buyers may be using the search bar more often than exploring because they have purchase-intent for specific items
- Homescreen content can be more personalized for most users tested
Additional Findings

Empathize
USMO to Job Stories
After research, I had a better understanding of shopper behavior; I further used the USMO and job story framework to get into the mindset of the users and see what the ultimate desired outcome would be for a shopper on Carousell.

User Journey Map
Leveraging my understanding of user behaviors from research and empathy, I created a journey map
depicting common behavior exhibited by e-commerce shoppers. Doing so helped me visualize which areas were in need of improvement so that I could map potential solutions to them.

Ideate
Hi-Fi Mockups
Research validated that users were probably not browsing the homescreen because they had specific-item purchase intent. As such, in order to achieve the ultimate goal of providing a personalized and relevant user experience, my proposed solutions branched out from simply personalizing the homescreen.
The app has a wealth of demographic and behavioral data that was not being utilized effectively. I tried to leverage such data to personalize the user experience in my designed solutions:
Home Page


* — Behavioral data may not directly suggest that a user would enjoy items related to gardening, but if other Carousell users who exhibit similar behavior also typically look at gardening items, then there is a higher likelihood that this specific user would also be interested in this content. The same extrapolations could be made at the listing-level to curate a “For You” section that pushes relevant and personalized items to users.


Listing Details Page

Likes Page


Validation Plan
Prioritization
I used a 2x2 matrix and made some assumptions around cost of implementation and impact of feature on KPIs in order to prioritize the features:
1. Sub Navbar — (in place of carousel)
2. Customized Homescreen Layout
3. “Similar Listings” — (on Listing Details page)
4. “You May Also Like…” — (on My Likes page)
5. “For You” — (recommendation tab on new sub navbar)

Feature Flags
The impact of my designs on the user experience would have to tested and measured through controlled AB testing in order to obtain statistically significant results
I would elect to AB test and roll out the features via feature flags, which are often used by e-commerce apps to display customized layouts and content to users based on user data. It further allows for the ability to roll out and test in a controlled environment while reserving the capability of turning off the features completely at any time.
AB Test Plan
I wrote out a high-level test plan for my design complete with hypotheses, KPIs considered, expectations, and risks for each of the features designed:




Fin
By completing this design challenge, I learned a lot about an industry that I didn’t know much about. Although challenges can be a lot of work, especially when you try to go through the entire design process to come up with tangible suggestions, I’ve found that they are one of the best ways to hone and practice your skills.