The question "how much should I import" sounds simple, but this is where new operators consistently split in the wrong direction: either too little to get any usable data, or too much and stuck holding inventory when the product does not move. This is about sizing a test batch so it is small enough to limit your downside and large enough to actually tell you whether the product sells.
A test batch buys information, not profit
When you have no sales history on a product, every decision is a hypothesis. The point of a test batch is not to turn a profit on the first run. It is to buy data: does this product sell, at what price, and at what pace. Those answers determine whether you import again and how much.
That means the size of the batch needs to be anchored to something concrete: how many orders do you need before the signal means anything. There is no universal number, but there is a framework for working it out per case.
Set the minimum order count before you calculate anything else
A trustworthy sales signal is not "sold out". It is a conversion rate on the listing, a weekly velocity, and enough customer feedback to know what is working. To measure those three things, you need enough impressions and enough transactions.
In practice, somewhere between 30 and 50 real orders within two to four weeks is the floor for a meaningful read. Under 20 orders, you cannot tell much: the listing may not have gotten seen, the photos may be off, the price may not be right. Above 50, the picture is clear enough to make a next decision.
For low-ticket items under roughly 100,000 VND, sales tend to come faster and you get data sooner. For mid-range products in the 200,000 to 500,000 VND range, expect a slower buy cycle and budget more time.
Work backward from orders to units
Once you have a target order count, you can work backward to how many units to import.
Units to import = Target orders / Expected delivery rate (after cancellations and returns)
Cancellation and return rates vary by category. For everyday consumer goods on TikTok Shop, cancellations typically run between 10 and 25 percent depending on listing quality and how the product is sold. Higher-value goods or items that need detailed descriptions tend to see more returns.
Example: if you want 40 completed orders and you expect roughly 20 percent cancellations, you need to import at least 50 units. Add 10 to 15 percent for defects and transit damage, and you land around 55 to 58 units as a practical starting count.
For most categories, a test batch of 50 to 100 units is a reasonable range. Where exactly you land inside that range depends on landed cost and how much risk you are comfortable taking.
Set a capital ceiling separate from your operating budget
Alongside unit count, put a hard ceiling on test-batch capital before you calculate anything. This ceiling should be treated as money you are willing to write off entirely if the test fails.
For someone starting out, that ceiling often falls somewhere between 3 million and 10 million VND per test. Not because those are magic numbers, but because losing within that range still leaves you capital to run another test on a different product.
To calculate: multiply your import quantity by the real landed cost per unit. Landed cost includes the 1688 product price, domestic China shipping to the consolidation warehouse, freight to Vietnam (by actual weight or volumetric weight, whichever is higher), order-agent service fees if you use one, and an allowance for damaged units. Convert from yuan to dong at the current rate (around 3,600 VND per yuan at the time of writing, though exchange rates shift daily so check when you calculate).
If the quantity you need to get a useful signal times your landed cost exceeds your capital ceiling, you have two options: find a product with a lower per-unit cost so each test is cheaper, or accept that the signal you get will be weaker and factor that uncertainty into your next decision.
Test pricing must match your intended long-term price
This is a point many operators miss. A test batch has a different goal from a clearance batch, and the pricing needs to reflect that.
If you plan to sell a product long-term at 250,000 VND, test it at 250,000 VND. Selling at a discount to move units fast and concluding "it sold" tells you nothing useful. You cannot hold a discount price indefinitely once you are running a larger import, and the economics at the real price may look completely different.
The same logic applies to advertising. If your channel depends on paid ads, running the test with no ads gives you data about organic visibility, not about how the product actually performs in your real setup. Run ads at whatever level you plan to sustain, even if it is modest. A few hundred thousand to a few million VND in ad spend is usually enough to give the listing enough reach to produce real data.
How long to run a test
The right window depends on your product's buy cycle. For fast-moving consumer goods, two to three weeks is usually enough to hit 30 to 50 orders if the listing is set up properly. For products where the buyer takes longer to decide, four to six weeks gives a more reliable read.
Do not read the results after one week. The first week is often distorted: the platform algorithm has not pushed the listing yet, there are no reviews for buyers to reference, and conversion rates are artificially low. Data from week two and week three reflects actual buying behavior much better.
Three questions to answer when the test ends
At the end of a test batch, you need to answer three things:
What was the sell-through rate? Divide units sold by units imported and by the number of days the batch ran. If you moved two to three orders per day and you want to scale, that gives you a rough monthly demand figure to build the next order around.
What was the actual margin? Take real revenue, subtract every cost: product, freight, platform fees, ad spend, and returns. Compare it to what you projected before importing. If they are close, your cost model is solid. If there is a large gap, find the cause. The most common culprits are a higher return rate than expected or ad costs running above the estimate.
Were there any quality issues? Customer complaints and returns linked to product defects are among the most valuable signals from a test batch. If a meaningful share of returns cite quality problems, the issue is with the supplier or the product, not with the market. Learning this from 50 units costs far less than learning it from 500.
Moving from test batch to full import
The three answers above should be clear enough to support one of three decisions:
Import more, larger quantity: sell-through rate is solid, margin matches the model, no quality flags. Time to scale.
Import again but adjust: sold reasonably well but one variable needs fixing before scaling. Common adjustments are pricing, photography, listing copy, or switching to a different supplier for the same product.
Do not import again: sell-through was weak even after giving the listing time and attention, or the margin after ads does not justify the effort.
The third outcome is not a failure of the test. It is the test working correctly. Spending a few million VND to learn that a product is not right for your shop is far cheaper than spending ten times that amount on a large import to reach the same conclusion.
Bottom line
A test batch is a decision-making tool, not a formality. Size it to produce roughly 30 to 50 real orders, within a capital amount you can afford to lose entirely. Price and advertise it the same way you would a permanent product line. When the batch ends, answer the three questions on sell-through rate, actual margin, and quality. Those answers, not intuition, should drive the next import decision.