Building a Manufacturer Sourcing Platform from Nothing
Pomu • 2024
No brand, no research, no product. Two founders with a thesis about AI matching. I turned it into a platform fashion designers trust.
Jump to solutionDesigned for
Pomu (AI-powered manufacturer sourcing platform, web)
Team
2 Designers, 2 Engineers, 2 Founders
Timeline
Apr–Oct 2024 (~6 months, 2 months discovery)
My role
Product Design, User Research, Visual Branding
THE CLIENT
Pomu
Pomu connects fashion designers with clothing manufacturers worldwide. Emerging designers spend months chasing leads through Instagram DMs, trade shows, and Alibaba searches, gambling time and money on partners they can't vet. Pomu's bet: AI-powered image search and verified manufacturer profiles can replace that guesswork.
I joined when Pomu was a pitch deck and a domain name. The founders had a hypothesis about using AI to match designers with manufacturers by uploading reference images, but no brand, no user research, and no product. Two months of the timeline went to discovery because the founders were still refining the business model. I used that time to run research and build the brand foundation so we could move fast once alignment landed.
THE CHALLENGE
Fashion designers distrust sourcing platforms. 85% of the designers we interviewed wouldn't use a platform without factory photos or certifications. The good manufacturers don't advertise because they already have full order books. The platforms that do exist (Alibaba, Maker's Row, Sewport) optimize for volume, not trust. We needed to build credibility from day one with zero track record.
PROJECT GOALS
- 1Define the brand identity, visual language, and design system from scratch
- 2Validate the core value proposition through user research before building
- 3Design the end-to-end platform experience: search, manufacturer profiles, onboarding
- 4Ship a launched product within 6 months
THE PROCESS
Discovery & Research
Apr–May 2024Interviewed 15 designers and 5 manufacturers. Ran competitive audits of 6 sourcing platforms. Conducted card sorting with 8 designers to prioritize platform attributes. Synthesized insights using Otter.ai and Miro.
Define & Brand
Jun 2024Created the Pomu brand identity: logo, type system, color palette, iconography. Defined the primary persona, key user tasks, and journey map. Framed the core HMW: simplify ethical manufacturer sourcing through transparency.
Ideation & Prototyping
Jul–Aug 2024Prioritized features into high-impact (AI image search, trust dashboard) and deferred (AR factory tours). Built low-fi wireframes for 4 core flows: landing, onboarding, dashboard, manufacturer page. Tested clickable prototypes with 6 designers.
Design & Launch
Sep–Oct 2024Designed the full hi-fi UI, component library, and marketing pages. Iterated through 3 rounds of usability testing. Handed off specs to engineers and supported QA through launch.
Discovery & Research
Apr–May 2024Interviewed 15 designers and 5 manufacturers. Ran competitive audits of 6 sourcing platforms. Conducted card sorting with 8 designers to prioritize platform attributes. Synthesized insights using Otter.ai and Miro.
Define & Brand
Jun 2024Created the Pomu brand identity: logo, type system, color palette, iconography. Defined the primary persona, key user tasks, and journey map. Framed the core HMW: simplify ethical manufacturer sourcing through transparency.
Ideation & Prototyping
Jul–Aug 2024Prioritized features into high-impact (AI image search, trust dashboard) and deferred (AR factory tours). Built low-fi wireframes for 4 core flows: landing, onboarding, dashboard, manufacturer page. Tested clickable prototypes with 6 designers.
Design & Launch
Sep–Oct 2024Designed the full hi-fi UI, component library, and marketing pages. Iterated through 3 rounds of usability testing. Handed off specs to engineers and supported QA through launch.
MEET MAYA
She wanted to design clothes. Instead she's on her laptop all day.
Maya is a composite persona built from 15 designer interviews. Her story showed up in almost every conversation: talented designers burning their creative energy on sourcing logistics instead of the work they trained for.
Meet
Maya, an emerging fashion designer in Brooklyn
“I just want to design. I spent three weeks emailing factories, scrolling Alibaba, and DMing brands for referrals. I found one manufacturer. They ghosted me after I sent my tech pack.”
Key tasks:
THE PROBLEM
Fashion designers gamble on every new manufacturer.
Emerging designers find manufacturers through Google searches, Instagram DMs, trade shows, and Alibaba. Each channel has the same problem: no way to verify quality, ethics, or reliability before committing money. A designer ordering 200 units from an unknown factory in another country is placing a bet with no information. Most lose that bet at least once.
User journey: Google → Alibaba → Instagram DMs → trade shows → email chains → ghosted
Discovery takes weeks of manual outreach
Designers described spending 2-3 weeks per manufacturer search. The process: Google for factories, scroll through Alibaba, DM brands on Instagram asking for referrals, attend trade shows, email factories one by one. Most inquiries go unanswered. Established brands guard their manufacturing sources.
Competitor landscape: Alibaba (volume), Maker's Row (basic certs), Sewport (rigid search)
Existing platforms optimize for volume over trust
Alibaba lists thousands of manufacturers with price-based filters and a verification process that means little. Maker's Row offers basic certifications but relies on text search. Sewport has client reviews, but keyword search makes discovery rigid. None of these platforms answer the question designers actually ask: can I trust this factory with my designs?
Interview data: trust signals ranked by importance (factory photos, certifications, reviews, price)
85% won't use a platform without factory photos or certifications
The sharpest finding from our interviews. Designers need visual and documentary proof before they'll engage. Stock photos and self-reported descriptions trigger instant distrust. Platforms that show real factory floors, ISO certifications, and production samples convert. Platforms that don't get closed in a tab.
THE RESEARCH
20 interviews, 6 platforms audited, 8 designers card-sorting.
We interviewed 15 designers and 5 manufacturers across experience levels, from independent founders producing 50-unit runs to established brands managing 5+ factory relationships. We audited 6 competing platforms. We ran card sorting with 8 designers to rank 20+ platform attributes by importance. Three themes dominated every data source: transparency, speed, and proof.
Interview quotes: visual search preference across 12 of 15 designer participants
Designers want to search the way they think: visually
12 of 15 designers described their ideal search as 'show them a picture of what I want and find someone who can make it.' Text-based search forced designers to translate visual ideas into keywords (e.g., 'ponte knit drop-shoulder oversized jacket') that rarely matched manufacturer vocabulary. Image-based search matched how designers already communicate with each other.
Card sorting results: certifications and reviews ranked 2x higher than price
Certifications beat price in card sorting
We gave 8 designers 20+ cards representing platform attributes (price transparency, ISO certifications, client reviews, MOQ flexibility, location, factory photos, response time). Certifications and reviews ranked at the top, twice as important as price. Designers will pay a premium for a manufacturer they can verify. The cheapest option with no proof gets ignored.
Manufacturer interview insight: quality factories don't need platforms, platforms need them
Manufacturers don't advertise when they're good
The 5 manufacturer interviews revealed the supply-side problem. Reputable factories run at capacity and rely on referrals. They don't list on Alibaba because they don't need to. The factories that do advertise tend to be newer, less established, or desperate for orders. Pomu needed to attract quality manufacturers by reducing their onboarding friction, not by offering them more visibility.
Certifications and client reviews ranked twice as important as price in our card sorting exercise. Designers will pay more for a manufacturer they can verify. The trust problem isn't about cost. It's about proof.
COMPETITIVE AUDIT
Six platforms, two gaps nobody filled.
We audited Alibaba, Maker's Row, Sewport, Kompass, ThomasNet, and Fashionphile's supplier network. Every platform had some version of search and filtering. None offered real-time collaboration tools between designers and manufacturers. None surfaced sustainability or ethical sourcing standards in a meaningful way. Those two gaps became Pomu's positioning.
Maker's Row
Maker's Row: text search, basic certification badges
US-focused manufacturer directory with basic certifications. Search relies on text filters, which limits discovery for designers who think visually.
Sewport
Sewport: keyword search, limited client reviews
Connects designers with manufacturers and allows client reviews. Keyword search makes browsing rigid, and the review system has limited adoption.
Alibaba
Alibaba: massive volume, weak trust signals
The largest manufacturer marketplace. Massive selection and price-based filters, but no meaningful trust signals. Verification badges are pay-to-play, not quality indicators.
THE BRAND
A dark, premium identity that signals credibility.
Designers associated light, minimal interfaces with low-quality sourcing platforms (Alibaba's white-and-orange, Maker's Row's generic SaaS look). We went the opposite direction: a dark palette with purple accents that felt closer to a fashion brand than a B2B tool. The visual identity needed to say 'we take this seriously' before a user read a single word.
Pomu logo system: wordmark, app icon, favicon, dark and light applications
Logo and mark
The Pomu wordmark uses a custom letterform with a crossbar 'K' symbol. The mark works at app-icon scale (iOS home screen) and at billboard scale (marketing collateral). We tested it against 3 alternatives with the founding team and 4 external designers. The final mark scored highest on 'professional' and 'trustworthy' in blind preference testing.
Color palette: dark base (#1a1a2e), purple accent (#A855F7), usage examples
Dark palette with purple accents
The color system centers on near-black (#1a1a2e) with electric purple (#A855F7) as the primary accent. White text on dark backgrounds. The purple carries interactive elements: buttons, links, hover states, active filters. We tested the palette against a light alternative. Designers preferred the dark version 3-to-1 in a quick preference study, citing 'more professional' and 'feels like a real product.'
Custom icon set: 12 icons on dark rounded cards, 2px stroke, consistent grid
Iconography system
We designed a custom icon set for core platform actions: search, filter, location, certifications, messaging, and reviews. Each icon uses a consistent 2px stroke weight on a rounded dark card. The icons read at small sizes in the dashboard sidebar and at large sizes on the landing page feature grid.
Brand touchpoints: landing page, business cards, iOS home screen, social templates
Marketing collateral and app presence
We designed the landing page, pricing page, social media templates, and App Store listing as a unified system. The business card uses a debossed logo on dark stock. The iOS app icon tested well for recognizability against 20 competitor icons in a simulated home screen layout.
THE PRODUCT
AI image search, trust dashboard, and manufacturer profiles.
Three features emerged as high-impact from our research: AI-powered image search (designers upload a reference photo and get matched manufacturers), a trust dashboard (certifications, reviews, and communication history in one view), and detailed manufacturer profiles (factory photos, production capacity, specializations). We designed each to solve a specific research finding.
AI image search flow: upload reference → AI matching → ranked manufacturer results
AI image search
Designers upload a reference image — a sketch, a fabric swatch, a competitor product — and Pomu's AI returns manufacturers who produce similar items. The results page shows match confidence, pricing ranges, and MOQ. In usability testing, designers found relevant manufacturers in under 3 minutes. The same search took 2-3 weeks through manual channels.
Manufacturer profile: factory photos, certifications, reviews, capacity, MOQ
Manufacturer profiles with trust signals
Each manufacturer profile surfaces factory photos, ISO/sustainability certifications, production capacity, minimum order quantities, client reviews, and response time. We prioritized certifications and reviews at the top because card sorting ranked them highest. A designer scanning a profile gets the trust-critical information before scrolling.
Designer dashboard: saved manufacturers, active conversations, order tracking
Dashboard with saved searches and relationship tracking
The designer dashboard centralizes active conversations, saved manufacturers, order status, and sample tracking. Designers managing 3-5 manufacturer relationships at once told us they used spreadsheets and email threads. The dashboard replaced that fragmented workflow with a single view of every active relationship.
Onboarding flow: production needs → material preferences → timeline → first search results
Onboarding that sets expectations
The onboarding flow collects production needs (garment type, quantity, timeline, material preferences) before showing any search results. This serves two purposes: it pre-filters the AI matching to return relevant results, and it signals to designers that Pomu understands their workflow. Usability testing showed 92% completion rate on the onboarding flow.
FINAL DESIGNS
From landing page to manufacturer profile, one system.
The shipped product covers 4 core surfaces: the landing page (acquisition), designer onboarding (activation), the search dashboard (core loop), and manufacturer profiles (trust and conversion). The dark visual identity carries across all surfaces. Every screen connects research findings to design decisions.
Landing page: hero with AI search demo, value props, social proof section
Landing page with AI search demo and trust messaging
Pricing page: Free, Pro, Enterprise tiers with feature comparison
Pricing page with three tiers
AI image search: reference image uploaded, manufacturer results ranked by match
AI image search results with match confidence scores
Manufacturer profile: factory photos, certifications, reviews, messaging CTA
Manufacturer profile with full trust signal system
Designer dashboard: saved searches, active conversations, order tracking
Designer dashboard with relationship management
Designer onboarding: production needs, material preferences, first AI match
Onboarding flow collecting production context
Brand system: logo, icons, business cards, App Store listing, social templates
Complete brand identity system
THE IMPACT
85%
Usability task completion
Participants completed core tasks (find a manufacturer, evaluate their profile, send an inquiry) in final usability testing.
2.4 min
Avg. time to first inquiry
Designers found a matching manufacturer and sent their first message in under 3 minutes during testing. Participants reported spending 2-3 weeks on the same task using existing methods.
4.6 / 5
Trust perception score
Average trust rating for manufacturer profiles in post-test surveys. The trust dashboard and certification badges drove the highest confidence.
Launched
Product status
Pomu shipped and is live. The platform onboarded its first cohort of designers and manufacturers within the first month.
We shipped a platform that designers described as 'the first sourcing tool I'd actually use.' Usability testing showed 85% task completion and a 4.6/5 trust score on manufacturer profiles. The AI image search cut discovery time from weeks of DM-chasing to under 3 minutes.
LOOKING BACK
Slow stakeholders gave us better research
The two-month discovery phase felt frustrating at the time. The founders were still debating the business model, so we couldn't start designing screens. Instead, I ran all 20 interviews, the competitive audit, and the card sorting exercise. By the time the founders aligned, we had a research foundation that made every design decision faster. The delay produced a better product.
Brand and product can't be designed separately
I built the Pomu brand and the product UI in parallel, not sequentially. The dark, premium visual identity came from the research: designers associated light, airy interfaces with low-quality Alibaba knockoffs. The aesthetic itself became a trust signal. If I'd handed off brand guidelines to someone else and designed the product separately, that connection would have broken.
AI features need manual fallbacks
The AI image search was Pomu's marquee feature, but it didn't always return relevant matches for niche fabrics and unconventional silhouettes. We added customizable filters (material, MOQ, location, certifications) as a manual backup. In testing, 40% of users combined both: uploaded an image, then narrowed results with filters. The hybrid approach outperformed either method alone.
WHAT I'D DO DIFFERENTLY
I would have pushed for manufacturer interviews earlier. We talked to 5, but their perspective came late in the process. The manufacturer onboarding flow was the weakest part of the launch because we designed it with less research depth than the designer-facing side.
LOOKING AHEAD
Niche-specific filters for specialized production needs (vegan materials, custom certifications), a messaging system with order tracking built in, and a manufacturer review system with verified purchase badges.