When we started building Ordinex, we had to answer one question before anything else: build what, for whom, starting where. The answer looks obvious in retrospect. At the time it was not.
The starting point: sellers importing from China to Vietnam
The market we understood best was Vietnamese sellers operating on TikTok Shop, Shopee, and Lazada, sourcing mainly from China through 1688. This group is large, growing, and carrying real operational problems every day.
They source from a large B2B platform and sell across at least two or three e-commerce platforms. They manage everything themselves: product selection, landed-cost calculation, purchase order tracking, inventory, per-SKU profit and loss. Most do this by hand or in a spreadsheet. That is a problem large enough to build for, and specific enough that we could actually understand it.
Why 1688 and not Taobao or Yiwu
Attaching 1688 to our starting point was not accidental. Three specific reasons drove it.
1688 is where most Vietnamese sellers actually place wholesale orders. Not Taobao (retail pricing, higher per-unit cost), not Yiwu (requires in-person visits or a dedicated agent), not Alibaba (higher base prices, larger MOQ, built for export orders). When we talked to real sellers, 1688 came up overwhelmingly. If we wanted a tool that was useful immediately, we had to start where users were actually working.
1688's data has enough structure to build on. Price, sold count, reviews, product images, factory details: these are present and consistent enough for an analytical tool to rely on. Not every B2B platform has that. Other wholesale sources we looked at had no usable API or data too fragmented to process automatically.
Vietnamese sellers hit a specific wall right at 1688. They know how to open the site and search. But going from a list of twenty near-identical results to a confident decision on what to order, how much, and at what actual landed cost is the step that takes the most time and produces the most mistakes. That is exactly where a tool can make an immediate difference.
Why not broader from the start
There were conversations about supporting other sourcing channels: Taobao, Pinduoduo, some domestic wholesale platforms. In theory, more sources means serving more people.
But each sourcing channel has its own data structure, its own pricing logic, its own usage patterns. Doing one well is much harder than doing five passably. If we split our attention early, the result would be a tool that is not good enough for anyone.
The decision to stay narrow on 1688 was not about lack of ambition. It was about getting one group of users to genuinely feel "this was built for me" before expanding. If the first core users did not find the tool truly useful, adding more source channels would not fix that.
Why Vietnamese sellers and not another market
The market we understood most deeply, through people we knew and conversations we had directly, was Vietnamese sellers. Understanding a market is not just reading the numbers. It is understanding how people describe their problems, how they make decisions, which pain is real and which is generic frustration.
Sellers in other regional markets may have similar problems, but the specifics differ: different platforms, different payment rails, different import paths, different on-platform language. Building a tool that is genuinely good for a market requires knowing those specifics in detail. We could do that for Vietnam. Doing it simultaneously for three or four markets from day one was not realistic.
This is not a permanent boundary. If we eventually understand another market well enough to serve it well, we will go there. But expanding geography before the first group is well-served is dilution, not growth.
The cost of staying focused
Staying narrow meant turning things down. Some potential users asked about Taobao, about domestic sourcing, about other markets. The honest answer was: we do not cover that yet, not well enough to commit.
There were also moments when someone compared Ordinex to a broader tool and asked why we did not do the same. The answer was the same: broad coverage and doing it well are not the same thing. An aggregator that pulls prices from five sources but does not calculate real landed cost, does not understand the cost structure of importing into Vietnam, and does not help with actual decisions is a different product. We wanted to build the second kind.
Scout, the first tool we built, filters 1688 products by real margin after adding freight, platform fees, and shrinkage. That step requires understanding how 1688 goods move to Vietnam, how Vietnamese platforms charge fees, and how Vietnamese sellers actually think about profit. It is not something you can copy from a generic template built for another market.
What we did not know at the start
Looking back, the decision to focus on 1688 and Vietnamese sellers was correct. But we did not know that with certainty at the beginning.
Some things we guessed right: real demand was there, users would try a new tool if it genuinely solved their problem, and depth mattered more than breadth at the early stage.
Some things we had to learn in the process: how sellers actually used the tool (not how we imagined they would), which steps in their workflow were hardest (not the ones we assumed), and how quickly they were willing to change existing habits.
Margin Engine v0.2 shipped in February 2026, batch import in March, Orders beta in April. Each of those came from watching how real users used the previous version, not from the original assumptions.
Bottom line
We chose 1688 as the starting point not because it is the only market worth building for, but because it was the one we could do well from the start: we understood the users, the data, and the problem in enough detail to actually solve it. Focus is not a constraint on ambition. It is the only way to avoid building everything at once and doing none of it well.