Role & Responsibility
Lead UX Designer
2015 - 2016
Area of Impact
User Experience, User Interface Design
When I joined Coupang in 2015, it was a small startup with company-wide culture of rigorously running A/B tests and conducting user tests for almost every newly released features; Over the years, Coupang differentiated itself and disrupted the market its innovative next-day delivery system and now Coupang is known as "Amazon of Asia," the largest unicorn online marketplace in South Korea, with $21 billion revenue.
As I joined its Search Team as Lead UX designer, team's focus was to optimize for the UX of product search through A/B tests and user testings (UT.) Thus, I began with discovering most urgent user problems and ideating for new design solutions.
User Problem Discovery
through UTs and interviews
Inaccurate and Redundant Results
When a user types in a keyword to search for product in a shopping site, such as Coupang, s/he expects to find accurate items at the top of the result page immediately. This is so commonly anticipated that, not having this experience will create frustration very quickly and can result in a drop of user activation and engagement.
Coupang's search team was no stranger to this problems; In order to tackle this, I turned to the UX Lab and asked to conduct user testings to gather key insights.
Users were experiencing difficulty finding products because search result page showed inaccurate and redundant results. Naturally, users took longer time to scroll and sift through lists, hence their frustration escalated after even just a few minutes.
Their frustration indicated something clear; Nobody has time to scroll through search result page all day. This is costing us user engagement.
So what was causing this problem? I began to dig deeper.
Voice of customers gathered from UTs was a pool of qualitative data mixed with users' wants, needs, and problems. Here are some snippets of what I found :
I pulled few recurring problematic points from the data:
Product Cluster Problem
Products from same brand but with different attributes, (i.e., color, size, scent) were being listed as separate items. This was causing search result page to be redundant and long.
Better UX would be
to cluster same products and list them as one item with options.
Too Many Attributes Problem
Due to the previously defined problem, it became more difficult for users to find product with the right attribute because result page showed endless duplicates with slightly different options. It needed a way to cut down searching time drastically.
Better UX would be
to offer easier, more intuitive way to filter out the result (a.k.a don't make users think.)
Hard to Scan Problem
Without user-centric sorting logic, searched products seemed to be listed in a random order, furthering the chaotic messiness.
Better UX would be
to list items in the order of users' priority and importance.
To design UX solutions, I worked with engineers to understand what was causing aforementioned problems, and quickly learned that the way product data was stored and classified was the root cause. The company was scaling too fast and the initial back-end data architecture could no longer keep up with current influx of new product information.
Therefore, a sustainable UX design improvement needed a two-fold change:
1. Data structure redesign, which takes a sizable amount of time and effort of back-end engineers (Root cause fix takes long time.)
2. Fast remedy to improve users' sifting process in the mean time (Low-hanging fruit approach.)
Root cause fix and fast 'remedy'
1. Restructure of Product Data
As mentioned earlier, to fix the root cause was to redesign data structure, which required back-end engineers' time and resource. Luckily, we were working as agile team, which meant both engineer and designer worked towards the same goal.
I worked with back-end and data engineers to provide specific product attributes and classification that are be useful for users. After few rounds of reviews with product owner and data engineer lead, the team finalized the new direction for product data restructure, and the back-end data team began working on it.
2. Quick Filter
With data restructuring under way, I began ideating quick 'remedy' UX solutions that could be launched and applied right away; The idea was to provide users with a fast way to filter out search result lists, in a format that required less input, such as tapping, rather than typing.
After researching most frequently searched adjectives and attribute words, I designed keyword filter as buttons, right below the search bar, in color code. This is to ensure the visceral correlation between initial searched keyword and product attributes and to provide a fast and intuitive way to filter out result list.
3. New Sorting Logic
Another way to provide more accuracy and relevancy within the list was to sort the list according to users' priority. The pre-existing sorting order logic, according to user testings, was not catering to users' needs. I reorganized the order of list according to key insights harvested from user interviews.
Users want to see the fastest delivery on the top.
→ Next-day (Rocket) Delivery Items
If next-day delivery is not applicable, free delivery is important.
→ Free Delivery Items.
Users prefer and shop certain brands more over others.
→ Top Brand's Home Access
All Other Items
Through A/B tests and user testings, the team was able to understand how newly designed features were affecting users.
This included 46% increase in revenue through search, 5.2% increase in conversion (from search result page to cart,) and 5 minute decrease in total time spent on search result page before finding an item after landing on the search result page.
Each feature was tested one by one with A/B testing tool developed by the company's internal data analytic team.