Build the trade desk before the marketplace.
The first business is not an “AI Alibaba.” It is a narrow trust-transfer service: Ef qualifies Turkish suppliers for Japanese buyers, Kuanysh turns every conversation into a reusable merchant graph, RFQ, quote comparison, and buyer-facing decision pack.
Non-consensus view.
Most people would start with a directory. That is exactly backwards.
Why you two.
The advantage is asymmetric because it combines lived corridor access with system-building.
| Asset | Ef | Kuanysh |
|---|---|---|
| Language / culture | Turkish + Japanese communication, trust cues, seller instincts. | English/Canada-facing packaging, product thinking, structured briefs. |
| Market role | Supplier qualification, negotiation, old trade memory, category instincts. | Merchant graph, RFQ workflow, AI extraction, scoring, buyer packs. |
| Moat formation | Human trust transfer. | System memory: every conversation becomes reusable data. |
First wedge.
Start where Ef has corridor experience, but choose categories that do not drown the first sprint in regulation.
Non-electrical home décor
Visual, sampleable, low-regulation, strong curation/story value.
Scarves / simple accessories
High value-to-weight, easy samples, manageable labeling discipline.
Home textiles
Strong Turkish supply and repeatability; requires fiber/care label rigor.
Decorative ceramics
Strong design story; packaging and breakage must be solved.
Food / olive oil
High opportunity but only with Japan importer/compliance path.
Children’s clothing
Only if Ef has real reachable supplier/buyer contacts; avoid safety categories.
Operating model.
The unit of work is not a listing. It is an RFQ loop.
Extract Ef memory.
Past products, suppliers, buyer types, order sizes, failures, still-reachable contacts.
Build private merchant graph.
Suppliers, buyers, products, trust checks, interactions, RFQs, quotes, deals.
Run a narrow buyer conversation.
Do not ask “are Turkish products interesting?” Ask “what product would you source if supplier risk were handled?”
Send RFQs to suppliers.
Normalize the replies into price, MOQ, sample, lead time, incoterms, documents, risk.
Sell the buyer pack.
Supplier shortlist + quote comparison + risk notes + next action. Charge before marketplace features.
Two-week validation sprint.
The goal is evidence, not completeness.
| Days | Action | Exit criterion |
|---|---|---|
| 1–2 | Ef memory dump + choose first two categories + prepare supplier/buyer/RFQ templates. | 10 names worth contacting. |
| 3–5 | Enter 20 suppliers and 20 Japan buyer/importer leads into CSV/database. | At least 10 supplier outreach messages sent. |
| 6–9 | Buyer outreach with specific asks around décor/textiles/accessories. | 3 real buyer conversations or explicit product requests. |
| 10–14 | Send RFQs, collect catalogs/quotes, generate first buyer pack. | One paid pilot, sample request, or explicit buying mandate. |
What not to do.
Constraints preserve speed and credibility.
No public marketplace
Without density, it is a trustless catalog and a distraction.
No DDP promises
Importer-of-record, taxes, broker liability, and compliance must be solved first.
No compliance guarantees
Food, cosmetics, electronics, child safety, and medical categories need partners.