Sourcing goods for Ozon: suppliers, China and cargo

On Ozon the profit is decided before the card is ever created: who you buy from, at what price and how long the goods take matters more than the selling price. Everything below is drawn from seller videos we watched — 1688 walkthroughs, supplier-sourcing talks, and first-person accounts from people who imported from China and lost money on it. These are independent seller accounts, not Ozon documentation. The transcripts were collected in the summer of 2026; where a video states its own time frame, it is noted next to the figure. The sources contradict each other in several places; rather than picking a winner, both versions are kept.

Where sellers look for suppliers

In the supplier-sourcing video the first step is not choosing a channel but narrowing the target: which product, in what volume, at what price, to what quality requirement. After that, the channels that recur across the videos are these.

Channel	Who it suits	The limit the videos describe
Local wholesale markets (Sadovod and Yuzhnye Vorota in Moscow)	A beginner testing a niche on a small budget	You can buy 10-20 units, handle the goods, and haggle; but what is there today may be gone tomorrow, and there is no branding or customisation
Alibaba	First step into China	English support, sample orders, customisation, contracts and insurance; prices higher than 1688
1688	A more advanced seller	The lowest prices; the site is in Chinese and browser translation is a must. One video puts the gap at 15-30 percent, another at roughly 20
Manufacturers inside Russia	Anyone who wants their own model made	Free choice of colour, size grid and batch depth; you inspect the sample in person; deferred payment opens up over time. Small batches are taken reluctantly and you queue
Abroad: China, Turkey, Kyrgyzstan	Sellers moving to volume	The widest range and the lowest unit price; a large cash outlay for the first batch and a long wait

Two warnings come from opposite ends. From the sourcing video: if the supplier you are considering sells on Ozon or Wildberries himself, drop him — once you become his competitor he will cut the price to squeeze you out and you are left with stock that does not move. And if you work with a Russian factory, agree at the first meeting that your samples and patterns go to nobody else, or the same product can turn up on the marketplace against you.

There is a method warning too: do not choose among two, three or even ten suppliers; screen at least 20-30, ideally 50 companies, markets and pavilions. The long course video reverses the direction of the choice — you pick from the product, not from the supplier. That is, not going to a market and buying what you like, but finding the product through analytics and only then hunting for it in China.

How sellers vet a supplier on 1688

Both 1688 walkthroughs are aimed mainly at Wildberries sellers; the site mechanics are the same, but do not carry their marketplace figures over to Ozon. Search works two ways: translate from Russian into simplified Chinese and paste the characters into the search box, or run an image search with a screenshot of a card that is already selling. Every video recommends image search; when results are thin, the find-similar button widens the list.

What they read on the supplier card:

• Medals and the top-factory or top-seller badges. According to the video these are hard to earn and hard to keep; a supplier who ships defects can lose them.
• The repeat-purchase rate. This is not a return rate in the marketplace sense — it counts buyers coming back, so higher is better. The thresholds contradict: one video never even considers anything under 20 percent and says 50-70 percent belongs to the largest suppliers; the other says even 10 percent is workable and that 29-40 percent is already the ceiling.
• The sub-scores behind the review rating. One seller ignores the response-speed and communication scores because they are chronically marked down, and reads the quality score instead. Two out of five, for him, means a real problem.
• Years on the site and orders per year. Ten years is strong, a "1" means one year. A supplier with zero reviews on the item and ten orders in total is treated as unreliable; on 1688 an absence of reviews alone is normal, an absence of orders is not. A supplier with 11,000 sales in a year, on the other hand, is described as trustworthy.

The card itself holds two traps. The low price on top often belongs to a different variant in the list; the item you actually want sits further down and may be 25 yuan instead of 10. Whether the price is for one unit, ten or a hundred has to be read separately. The stock figure shown is not always current. And the best way to tell apart dozens of suppliers using identical infographics is the real buyer photos in the reviews.

Samples, the first batch and your own product

The test batch is kept small, but the test itself is slow. The seller who lives in China and sells on Ozon Global says testing one product takes three to four months because of shipping time and the wait for reviews, and that even after choosing the niche the product trials ate six to eight months. The seller who started with tote bags says he ordered "a couple of sacks" without doing any maths, sat on them unsold for three or four months, and finally dumped 60-70 percent wholesale to one buyer from Avito; the rest is still in his warehouse.

A seller working with Russian production describes the sample step more concretely: she looked at Moscow and Petersburg firms that both sew and print, and refused to order remotely — she wanted to walk into the workshop and see the fabric and the print quality herself. She ordered the first trial batch in the standard size and in the colour analytics showed the most demand for.

One step beyond that in China is contract manufacturing. The Ozon Global seller says that on the standard goods everyone buys from 1688 his margin fell to 5-10 percent with Chinese competitors on top of him, so he moved about 60 percent of his turnover to contract production: small customisations such as a colour change are not hard, and the product ends up slightly different from everyone else's. He also found a small factory that does not sell on 1688 at all, which effectively made him the only seller of a product unknown in Russia. After a year of trial and error he puts his net margin at 25 percent, with around 70 items and an average order value of 1,300 roubles.

Cargo, timing and the cost lines

One point is shared: the factory has no logistics of its own. One guide says he has "never seen a single one"; factories hire transport from outside, routes are not worked in, and boxes arrive dented. That is why both 1688 guides recommend working through an intermediary — a buyer or cargo company. The intermediary inspects for defects before shipping, takes over the haggling, and can have labelling done in China where it is far cheaper.

Line	Value quoted in the videos
Shipping arranged by the factory	5-7 dollars per kilogram, "what they consider a very good price"; another video says 5-6 dollars and up
Logistics through an intermediary	2-3 dollars per kilogram with a good buyer; 1.5-2 dollars is called a find
Intermediary commission	5-10 percent of goods value; one spreadsheet assumes 10
Worked example	At 3 dollars per kilogram and 100 roubles to the dollar, logistics for one batch came to 78,000 roubles
China to Moscow transit	20-30 days; "a month" in another video; roughly 45 days in total including the purchase
Supply horizon	1-1.5 months for the test batch; a standing 2-3 month horizon afterwards

Weight is its own line: logistics is charged by the kilogram, so the weight has to be known before ordering — the course video adds the item to the 1688 cart and reads the weight there (150 grams in the example). Cargo companies may each apply their own dollar and yuan rate. The other lines counted into unit economics: box repacking around 5 percent, a 2 percent defect allowance, acquiring at 1.5 percent, 23 roubles of packaging per unit, advertising at roughly 20 percent of turnover, and on the tax side 6 percent plus 1 percent above 300,000 roubles of turnover, so 7 percent goes into the model. On certification, one seller says about 10 percent of the items in his shop need declarations, the rest are covered by refusal letters, and the whole thing cost him 12,000 roubles. In the same course video the worked example closes in the red: a 37,000 rouble outlay, 17,000 roubles of loss. The guide's line is blunt — a ten-times gap between purchase and retail is not profit.

The cost of money is a line too. One seller explains that in a period when credit ran at 34-36 percent a year, 3 million roubles meant 90,000 roubles of interest a month, which is why they switched to negotiating deferred payment with suppliers instead: four suppliers out of five grant it, and once trust is built goods can even be taken on a sale-or-return basis. In one of the channel's own challenges a participant took a million roubles of goods that way at Sadovod, on first contact.

On the ordering step two guides contradict each other outright. One says never order through the site — you talk to the seller and arrange shipping separately. The other says goods can now be put in the cart and ordered straight from the site, that this got far easier than a year ago, and that single units are sometimes available. One more difference: cargo companies always bring goods into Moscow, whereas working directly makes other cities possible — though if the stock is going to Moscow warehouses anyway, that advantage is debatable.

Customs comes up under two headings. First, the work of official import: finding a customs representative or broker, learning which certificates and permits apply, and getting an invoice and packing list out of a Chinese seller who often does not know the terms. Second, classification: if you are unsure of the commodity code you can request an advance ruling from customs — you file a product description, photos, samples and technical documentation, it takes from a few days to a couple of weeks, and it is recommended for high-value shipments and for new or unique products. The videos also discuss "grey" import routes; we do not describe them here. The legal risk sits with the seller, and the same videos report that enforcement is tightening and documentation demands are growing.

What goes wrong

• Quality expectations do not match. The seller who imported wooden chandeliers says the two sides' idea of quality is diametrically opposed: a Chinese buyer hangs a lamp with small defects and forgets it, a Russian buyer paying 5,000 roubles does not. The same seller says the boxes arrived without instructions even in Chinese, so they wrote the text and drew the diagrams themselves.
• The factory does it its own way. The seller living in China says missed deadlines and quietly reworking the product are normal at the overwhelming majority of factories; one factory director would nod on a call and then say "I know better what the customer needs" and do as he pleased. His advice is one word: patience. The chandelier importer landed in the same place — if the factory cannot organise assembly or source decent materials, it is easier to change factory.
• Defective goods do not go back. Reverse logistics is expensive, so defects are thrown away or disposed of; one guide notes that by the time goods arrive after thirty days the warranty window is closed. The seller buying at Russian markets says the opposite side of it: there he can return a defective item, whereas everyone he knows repairs Chinese appliances themselves.
• Transit time eats the season. A video filmed at the end of June uses a water pistol as the example: the season is over before the goods land. During the autumn 2025 jam out of China, stock ordered in September arrived months later; one agency says it stopped charging its clients during that period while shops sat frozen.
• Bulk is punished. The seller who imported paddle boards says they filled half his warehouse, that moving one from Moscow to Petersburg cost 400-500 roubles, and that on a 10,000 rouble item this is heavy — after which he stopped importing large, oversized goods.
• Wrong goods and fraud. One guide says scammers do turn up on 1688, but that the repeat-purchase filter lowers the risk a lot. On the Ozon Global side, cards copied by machine vision mean the buyer can receive an entirely different item, and if your customised product has a lookalike on 1688 a cheap copy may be what arrives. The seller telling this registered his own brand in Russia, after which clone cards were taken down easily.

Short summary

• Pick the product first and the supplier second. Going to a market and buying what you like is the beginning the videos criticise most.
• A market is fast and offers credit but is not continuous; China is cheap but slow and paid upfront; Russian production gives you your own product but takes small batches unwillingly.
• Screen 1688 suppliers on medals, repeat-purchase rate, the quality sub-score, years and order counts — the thresholds differ from source to source.
• Cost logistics by the kilogram, with the intermediary commission, before you order; a ten-times gap between purchase and retail is not profit.
• Keep the test batch small and the supply horizon at 2-3 months; on seasonal goods the real risk is time.
• Defective goods do not travel back to China — put the defect allowance into the model from the start.

Commbot carries purchase cost and logistics lines per card into its pricing model, which is useful for seeing where unit economics moves when you change sourcing route.