E-commerce
Manufacturing
Material Exchange
I led design with a small team to create and launch beta for Material Marketplace, allowing manufacturers to buy and sell excess materials; reducing waste by finding buyers that want to acquire ingredients below minimum-order quantities. Redesigned in 2026 from beta learnings.
Timeline
4 months, Beta in Oct, 2024
Team
Me, 1 PM, 1 engineer
Role
Design Lead, Research, Strategy
Disposing of unused materials can cost up to $250 in disposal charges and an additional $150 in logistics.
Problem
Manufacturers over-buy raw materials constantly.
If a projects or a contract was terminated, the excess takes up space in a warehouse only to be disposed of at a loss. But… Other manufacturers may need those exact materials.
But why would competitors ever trade with each other?
These companies are direct competitors, and what they use is a trade secret.
“
— representative partner concern
Our design partners formed to cut losses, not to profit off each other, so fairness mattered as much as secrecy.
The core tension: these competitors were willing to trade, but only if nobody could tell who was buying or selling what.
We're a small company with paying clients, so we couldn't burn months on a maybe. Each iteration was a lean way to test whether the next stage was worth building. We trusted what partners did, never what they said.
The groundwork: regular focus groups with the 20 partners, semi-structured interviews, contextual inquiry, and joining real transactions end-to-end — down to the warehouse and inventory managers who actually know what's on the shelf. From that, I mapped everything a deal requires: how materials are verified, how deals, shipments, and payments are documented (certificates of analysis, purchase orders, shipping labels, container types and statuses, expiration dates, unit and product types), and every person involved along the way.
Stage 01: Validate
Running the exchange over email
What's the cheapest possible version of an anonymous exchange? Email, with us in the middle.
A survey only proves what people say they'd do; a real deal proves everything. So we became the exchange: someone asks for a material, it goes out to the group as "Participant A needs X," and we relay everything in between. I ran the research behind it, turning what partners said — and more importantly, did — into requirements.
The demand was real: people negotiated prices, produced the paperwork, and trusted us to hold money until delivery. But ~95% of inquiries hit sellers who couldn't help, and the more noise they got, the less they replied.
Anonymous trades happen — that was the green light to build real software, with a warning about the inquiry model noted.
STAGE 02: Build
The inquiry platform
We knew the marketplace was the destination. So why didn't I build it first?
I pushed for the quick version we knew we'd replace. The team was still running deals by hand, and automating the model we'd already proven would free them up while the beta collected the data to justify the bigger build.
I mapped the user journeys and process flows for inquiries, transactions, and payments — happy paths and the edge cases for messy situations — which set the standards, design, and information architecture of the site. The design was iterated, tested, and engineering approved before documented for handoff.
The hardest design problem was anonymity itself. You can't hide everything, since a shipping label needs a real address. So anonymity works as a sequence. Identity stays hidden throughout negotiation, and documentation moves through the platform rather than company to company. Disclosure only comes at purchase order and shipping label creation, once the price is locked and knowing who's on the other side no longer matters, so partners never pushed back.
STAGE 03: learn
6 months of beta data
Within 6 months, users each saved an average of $4k in potentially wasted supplies, preventing ~2 tons of product from entering landfills.
The topline looked good: 20+ customers, 110+ transactions, $80k total saved, $4k saved per customer. People came back and traded again. But… why wasn't engagement growing?
The flaw was structural. Ask 20 people for a material and most say nothing. Once buyers stop trusting they'll get an answer, the whole thing loses its value. No UI polish fixes that.
We got anonymity right — identities stayed protected, and deals still closed. The biggest limiting factor was that buyers had no reference to what inventory was actually available. Every inquiry was a question in the dark, lingering until sellers answered whether or not they had the requested materials.
“
— design partner, beta focus group
Deals were slow for a related reason: buyers and sellers couldn't just call each other, and we hadn't built chat to replace that.
It wasn't easy to accept that our most successful build had this flaw. But now we knew exactly why it failed and what to build instead — and the problems were still sitting there, unresolved.
STAGE 04: iterate
Redesign as a marketplace
This stage is solo work: I designed and built a functional, industry-agnostic prototype, with AI handling the engineering.
The core change flips who does the work. Instead of posting inquiries and waiting, buyers search real inventory that sellers have listed anonymously. Available materials would now be visible while the company behind it is protected, which is exactly what the beta was missing. Sellers only do the work once: they list their excess and let a single account manager handle any requests that are likely to result in transactions.
Three problems drove the design:
Making materials findable
Search by SKU, standard name, or INCI, narrowed by industry, category, and function — a buyer who only knows what a material does can still find it.
Handling orders and partial sales
Orders carry through inquiry, processing, negotiation, shipping, and completion in one flow, with real-time chat replacing the phone call anonymity took away.
To handle partial requests, an active order reserves only its quantity, the rest stays on the market as its own listing.
Structuring organization accounts
Role-based permissions split the work — executives oversee operations, account managers run deals from a sales dashboard, and warehouse and inventory staff bulk upload and draft listings, with admin approval determining what goes live.
The partners set the constraint on growth themselves:
“
— design partner
Vendor registration upholds these values: verification, a user agreement, and vetting before anyone is invited to list.
The company never shipped this — priorities shifted, and I carried the vision forward on my own. The prototype isn't validated yet; the next step is usability testing with real clients.
Outcome
The email POC proved anonymous deals happen with nothing built: demand confirmed, prices negotiated, money trusted to escrow.
The inquiry platform turned that vision into 100+ transactions, ~$80K traded, ~$4K saved per user in 6 months. But it has its flaw: every deal depended on sellers answering inquiries for materials they mostly didn't have, while buyers could only post and wait.
100+
100+ real B2B transactions completed through hundreds of inquiries.
~$4K / user
On average, each user saved at least $4K through selling excess materials.
~2 tons
Excess materials were supposed to enter landfill, now they are being utilized and not wasted.
Both systems directly shaped the marketplace's design: the parts that worked carried over, and the failures dictated what had to change. If it works, responses go up, waits go down, and sellers stay engaged because listing once is all it asks of them. Validating that with real clients is the next step.
Reflection
The call I'd redo is treating "inquiry vs. marketplace" like an either-or. A middle option was sitting right there — sellers add a few inventory keywords, get pinged only on matches. That could have fixed engagement months earlier for way less.
What I'm proud of is that we never bet more than what we had validated.
The next steps follow the same rule. Run usability tests on the prototype with real clients. The metrics change with the model: how much inventory gets uploaded, how many transactions close, and what testers say about the material discovery process. The hard part will be building out the details of vendor registration, since the original design partners had strict requirements for including additional users.