Features
conversational search
customizable fyp
daily lookbook updates
time-sensitive offers
QR payment system
Popular actions
thumbnail hierarchy
rebox ai is focal
standardized listing
communicative inbox
Problem
I interviewed five individuals from my network to hear from their experiences with purchasing secondhand furniture. Participants have prior purchase experience in the last 12 months, are tech-savvy, and have used platforms like Facebook Marketplace and eBay. One young professional — a tech investor and an avid thrifter — sums up the overall sentiment:
“Facebook marketplace is just an absolute junkyard war type situation … just make it a more positive experience to shop and spend less time shopping.”
Clear trends also emerged:
5 out of 5 users said finding a suitable item is a time-intensive process that requires significant effort
3 out of 5 users indicated price is a priority, and 2 users noted the absence of secure payment options
4 out of 5 users mentioned a cluttered layout and limited or missing product details as major pain points
5 out of 5 users reported poor seller response rates as a source of frustration leading to missed purchases
opportunity
Product discovery dominated user conversations, signaling a massive opportunity. Today, LLMs are increasingly able to better understand context and predict what we want, which is transformative for discovery. We see this already in top search engines. Its potential in eCommerce is largely untapped, with the exception of Amazon’s beta launch of Rufus in 2024, which reaffirms what the future of shopping could look like.
LLM-based chatbots could provide a solution to one of the most time-consuming and frustrating parts of a shopper’s journey: browsing.
This is even more useful in the case of P2P marketplaces for pre-owned goods, because unlike traditional retail, secondhand shopping often involves browsing through vast, constantly changing listings which makes filters less effective. I applied LLM search to two core buyer journeys: Browsing and Buying.
optimized browsing flow
Users continuously refine search using conversational language instead of conventional trial and error
optimized buying flow
Communication features incentivize response, resolving for major drop-offs
solution
Conversational search
Message Rebox AI to instantly pull up listings based on your preferences, no more endless scrolling
Customizable For You PAge
Save search queries as “lookbooks”. Return anytime to find your search exactly as you left it
Daily Lookbook Updates
“For You” page lookbooks refresh every 24 hours with new finds matching saved criteria
I integrated an offer system that emulates a free market to drive increased transaction volume for Rebox while making both parties happy (buyer gets a deal, seller makes a sale).
time-sensitive offers create a real-time negotiation flow
Expanded offer header with key actions ensures negotiations stay clear, while each action is logged in-line
When users are picking up and paying for items in person, often with cash, there's no guarantee that the agreed-upon price will be honored. The key lies in QR codes as a way to lock in offers without introducing risk to the buyer, legitimizing the offer system. Inspired by AliPay and WeChat Pay, I used this technology for Rebox’s payment system.
Seller receives a unique QR code for the transaction which the buyer scans at pickup, releasing payment
This incentivizes use of the offer system and ensures:
No mismatched payment
No ghosting or selling to another buyer
No need for cash or change
Seamless payments integrated with platform
I reduced browsing friction with intentional designs.
popular actions are quickly accessible
Commonly used filters and location settings are pinned to the top
hierarchical framework for thumbnail details
Photos, Price, and Distance are the big three when it comes to browsing decisions
rebox ai is the focal point
The search feature was repositioned from the traditional top navigation to the main bar
standardized product listing
Related details are grouped into cards, reducing cognitive load
Messages from potential buyers and messages to sellers live separately, keeping the inbox neat. I designed a nudge button inspired by Facebook’s “poke” as another means to encourage proactive response. After 36 hours of no communication, either party can use the button to send a gentle reminder to check messages and take action.
inbox Architecture enhances communication
Nudge feature and offer expiration are accessible at the inbox. Inactive conversations are automatically archived
next steps
After conducting two usability tests from 10 randomly selected participants from my network, all testers were easily able to complete the browse and buy journeys. One tester mentioned,
“It’s really easy, easier than any other used furniture app I’ve probably used.”
That said, there was some initial confusion. As another tester noted, “I think it was just that offer part that was a little confusing if you haven’t used it before, but after you explained it, it was really easy and straightforward.”
What stood out most was that users appreciated the ability to interact naturally with the chatbot instead of constantly refining search filters. They shared that Rebox AI and the communication enhancements would save them significant time.
With further opportunity, I plan to address these user concerns, adding tooltips or popovers to guide new users through the offer process. Moreover, building out the “Events” page for users to host local sales and list individual items instead of in bulk. Overall, designing Rebox was a lesson in marrying product strategy and business goals with user-centered design. It reinforced the power of intuitive, well-crafted experiences in driving both user engagement and business success.