The story
We built this marketplace once already.
Then we audited it line by line, wrote down every way it could fail, and started over — so those failures are impossible by construction. Beta: September 2026.
01
The origin
SultanaExpress did not start as a plan. It started as a product: a production iOS marketplace — a buyer app and a supplier app — that we built, shipped and ran, with real users, connecting Turkish manufacturers with buyers.
Running it taught us where the real problem lives. Sourcing does not fail at the handshake — it fails after: the tech pack, the sample approval, the capacity slot, the quality inspection, the customs paperwork, the payment, the invoice, the certificate. And it taught us something harder: the system we had built could not carry that chain.
02
The audit
In June 2026 we stopped and did what teams rarely do to their own work: a full line-by-line audit. Before writing one line of new code, we mapped 106 database tables, 341 stored procedures and 687 row-level security policies, and catalogued 228 features one by one. Three findings decided everything that followed.
The database was the application.
687 security rules about who may see what lived inside the database itself — the place where they are hardest to test, review or change. The apps were a thin skin over a database making the real decisions. Picture a warehouse where the shelving decides who may take what off it.
Money was computed in floats.
Price math ran on the customer's phone, in floating-point arithmetic — the kind that turns 0.1 + 0.2 into 0.30000000000000004. Acceptable for graphics. Not acceptable for invoices.
The payment screens moved no money.
The payments layer was an interface with nothing behind it. Screens existed. No money moved.
We wrote all of it down. The new platform's shape comes directly from that document: every defect we found became a design constraint.
03
The decision
We could have patched. A patch fixes the bug in front of you and leaves the class of bug alive. We chose the other path: a clean rebuild in which every failure mode we found is impossible by construction — not caught in review, not fixed after the fact. Impossible.
The database stores rows; it does not decide who may read them, does not move money, and does not encode rules.
What that looks like in practice: money cannot be a float — the build fails. An order below a supplier's minimum gets no warning — the server refuses it. Bulk production cannot start before you approve the pre-production sample — the platform returns a refusal, not a reminder. And the 687 database policies became one auditable authorization layer inside one service — the only thing trusted.
The rebuild runs on a 649-step plan, roughly half executed as of 2026-08-01, every step behind automated checks. The whole record is public — see Building in the Open.
04
The four of us
Four founders, one team of four today. Here is who does what.
Louis Loubes
CEO
Company direction, partnerships and investor relations.
Abdullah Kahvecioglu
CTO
The platform: architecture, the build, and the quality system that walls it.
Leonard Loubes
COO
Operations: the supplier side, process, and the day-to-day running of the company.
Tatiana Loubes
CFO
Finance: accounting, compliance and the numbers.
05
What we hold ourselves to
Precision is respect.
We give you dates, counts and statuses, not adjectives. "Beta — September 2026" is a commitment you can check. A number on our site is counted, sourced and stamped — or it is not on our site.
Impossible beats unlikely.
A rule that lives in a code review can be skipped. A rule the build enforces cannot. Wherever a failure would cost you money or trust, we make it structurally impossible — and write the test that keeps it that way.
We work in the open.
Progress is public, dated and honest — including what is not built yet. Every capability we show carries its real status. Pre-launch trust cannot be bought with claims; it has to be earned with the record.
A refusal is a feature.
The server refuses a sub-minimum order. Production cannot start without your sample approval. Payments stay off until outside experts sign off. Systems that say no at the right moment are the ones you can rely on when it matters.
06
Where this goes
Türkiye first, as the test market — beta opens September 2026. Then Europe. The Türkiye→EU corridor is the roadmap, and we will call a step live only when it is live.
See the record
The plan, the counters and the build log are public.