A second-hand marketplace for the Carnegie Mellon community, where students buy from and sell to each other in one place: browse by category, contact the seller directly, and list an item in three short steps. I joined it as its product designer, to make it easier to use from end to end.
Studio
BTG Product Studio
Role
Product designer
Type
Product design · Web · Mobile
Year
2026 — present
Original team
Bryan Huang, Annabelle Chow, Amy Cha, Andrea Guo, Yutong Zhen
Why CMUsed?
Students trade in group chats and marketplaces built for strangers. Listings get buried, and you can’t tell who’s selling.
CMUsed is one campus marketplace, sorted by what students sell, where every account is a verified CMU email.
What was the problem?
Solutions?
Low activation.
A shorter path to a first listing.
From browsing and searching to contacting a seller or selling an item.
Low seller response rate.
Real-time updates through notifications.
Email notifications now, then notifications on the phone once the app is out.
Thin supply.
Seasonal sales and promotions.
Promoting move-in/move-out, Halloween and Christmas sales.
Low retention.
A mobile CMUsed app to simplify the process.
From listing to messaging to the deal, on the phone. In progress.
How I read the data
Where does this two-sided marketplace actually break? I analysed nine snapshots of the admin dashboard, from Sept 24 to Oct 20, 2026, across acquisition, activation, engagement, retention and transactions.
In short.
Acquisition isn’t the bottleneck; marketplace liquidity is. Sign-ups keep coming, but few users activate as sellers, supply depends on a handful of power sellers, seller response is the biggest drop-off in the funnel, and engagement is falling faster than MAU shows. The data points to the supply side first.
Acquisition is healthy; activation isn’t.
In the 30 days to Oct 20, CMUsed gained 100 sign-ups but only 7 new listings, a 14:1 ratio, up from 6:1 when I joined. Across all 1,822 accounts, 9.5% have ever listed an item and 4.6% have converted to a sale.
Sign-ups, last 30 days100
New listings, last 30 days7
Accounts1,822
Activated: listed an item173 · 9.5%
Converted: sold an item83 · 4.6%
Engagement is falling faster than MAU shows.
Weekly active users dropped 69% (116 to 36) while monthly actives dropped 23% (288 to 222). Stickiness, the share of monthly users active each week, fell from 40% to 16%, so MAU will likely keep falling. Only 4 items sold in those 26 days, and unsold inventory is priced 22% above what sells. Most of this is likely the start-of-semester spike wearing off, which another semester of data would confirm.
Sept 24116
Oct 284
Oct 591
Oct 688
Oct 786
Oct 1158
Oct 1447
Oct 1936
Weekly active users at every snapshot.
Supply depends on a few power sellers.
Seven accounts, 4% of all sellers, posted 32% of every listing and drove 37% of all sales. Their sell-through is also higher, 55% against 43% for everyone else. Losing even one of them would visibly thin the marketplace.
Listings from the top 7 sellers279 of 876 · 32%
Sales from the top 7 sellers152 of 407 · 37%
Seller response is the biggest drop-off.
Of 355 conversations buyers started, only 11.5% got a response, after a median of 28 hours, and 3.7% became a real back-and-forth. 87% of sellers who were contacted never responded. No other step in the funnel loses this many users.
Buyer messaged a seller355
Seller responded41 · 11.5%
Buyer replied back13 · 3.7%
Most transactions happen off-platform.
83 sellers have made a sale, but only 14 ever responded in the app’s chat, so most deals are closed somewhere else. That means the metric that matters most, completed transactions, isn’t being tracked at all.
What I’d prioritise.
Track the north-star metric. Completed transactions per week, captured with a one-tap “Sold through CMUsed?” when an item is marked sold, plus listing drop-off, which isn’t logged today.
Lift the seller response rate. Message notifications and a 24-hour reminder, aimed at the biggest drop-off in the funnel.
Grow supply at move-out. Recruit sellers when inventory peaks, rather than spending on more sign-ups.
Cut listing friction once the data shows it. If users start listings and abandon them, that becomes the next fix.
Source: CMUsed admin dashboard, nine snapshots from Sept 24 to Oct 20, 2026. Seller figures are totals only; individual accounts are not shown.
From browsing to a deal.
The site as I found it, built by the original team. It follows the moments a buyer goes through, and each screen carries only what that moment needs.
01Browse.
Two large ways in, clothing and furniture, then the most recent listings.
02Search and filter.
Results beside filters for price, quality and category, with sorting on the right.
03Decide.
The listing expands the card: more photos, price and whether it is negotiable, condition, when it was posted, who is selling, and one clear Contact Seller button.
04Message.
Contact Seller opens a chat with the listing already attached and a first message written, so reaching out takes one send.
The card at a glance
Most of the time on CMUsed is spent scrolling through listings, so the card only carries what it takes to decide whether to look closer: a big photo, the item's condition, its category, the name and price, and the two things you can do next.
condition tag & share
primary product image
category
name & price
primary actions (contact & save)
From mobile web to an app
On phones, CMUsed began as a mobile web prototype: the same site in the phone's browser, with the pieces stacked into a single column. We are now turning it into a mobile app, so browsing, messaging and selling work the way people expect on a phone.
Mobile web prototype · Home
Mobile web prototype · Listing
Listing an item.
Selling is its own short flow, one tap away from Sell Now at the top of every page.
