The Showroom Nobody Walks Into Anymore
A Foshan furniture manufacturer I spoke to last spring has a 900-square-metre showroom on the factory floor. Sixty-one dining sets, forty coffee tables, twenty-two storage units. For eleven years buyers walked through twice a year, picked twenty items, and placed container orders.
Those buyers still come. They just order less. So the owner did what most factory owners eventually do. She put forty SKUs onto TikTok Shop, waited for orders, and got four in the first month.
The products were not the problem. A TikTok Shop creator cannot sell “storage units.” They can sell a slim shoe cabinet built for a narrow hallway. She knows the buyer has that hallway. Her catalog did not say so anywhere. It listed model numbers and wood finishes instead.
This is for sellers who own or sit next to the production line — factory-backed brands in Foshan, Dongguan, Zhongshan, Huizhou, Shantou. You carry 50 to 200 active SKUs and sell into the US and increasingly Southeast Asia. You do not lack products. You lack a way to choose which twenty deserve a creator’s attention first.
What This Is and Is Not
This is a selection method, not a content strategy. We will narrow a catalog, match what survives to specific creator types, and run one controlled batch. We will not cover hooks, audio, or editing. None of that rescues a mis-picked catalog.
One constraint up front. This assumes you can fulfil small quantities quickly. If your line needs a 500-unit run before anything ships, read the sample section closely. That is where most factory programs stall.
Why Most Factory Teams Pick the Wrong First Twenty SKUs
Ask a factory team how they chose their first TikTok Shop SKUs and you usually get the trade show answer: whatever sold best to wholesale buyers last year.
That list fails here for one reason it does not fail anywhere else. Wholesale buyers reward breadth. They want to see the whole range so they can build a container. A creator judges one object against one problem in fifteen seconds.
So the two attribute sets barely overlap. Container-friendly means a sharp price point, neutral styling, easy logistics. Content-friendly means specific, slightly unusual, and obvious without narration. A plain white five-drawer chest is a wholesale winner. A chest with a built-in charging station, or one narrow enough for a rental hallway, is the content winner. Same tooling often. Very different results.
There is a second failure mode that hits sellers who own the line, and it costs more. Your production cost is visible to you in a way it is not to a trading company. That creates a pull toward your highest-margin item. That item is usually your most complex one. It is also the hardest to show on camera and the likeliest to arrive damaged.
The Three Filters That Actually Narrow a Catalog
Run your SKUs through three filters in order. Stop after each one. Do not carry everything forward hoping the next filter sorts it out.
The first filter is demonstrability without narration. Can someone grasp the point in three seconds with sound off? If value only appears after you explain the mechanism, the product can still sell. It just will not sell through mid-tier creators who will not rehearse a technical pitch.
The second is a named buyer problem. Not a category. “Small-space storage” is a category. “My hallway is 60cm wide and no shoe cabinet fits” is a problem somebody typed into a search bar. Find that sentence in your own reviews and support tickets. If nobody has complained about it, the demand may not exist.
The third is ship-and-survive. Aim for under 2kg. No glass. Nothing that takes over a minute to assemble on camera. Packaging has to arrive presentable, because creators film the unboxing before they film the product.
| Filter | What it removes | Typical pass rate (factory catalogs) |
|---|---|---|
| Demonstrability without narration | Anything needing a verbal mechanism | 45–60% eliminated |
| Named buyer problem in reviews | Generic SKUs nobody searched for | Another 50–70% |
| Ship-and-survive | Fragile, heavy, assembly-heavy items | Another 30–40% |
Three filters in sequence routinely takes 160 SKUs down to eight to fifteen. That surviving list is your first batch. It is usually nothing like your wholesale bestseller list.
Matching One Catalog Line to One Creator Type
Once you have ten SKUs, do not send all ten to all twenty creators. Give one creator one item. Make the pairing defensible enough to explain in a single sentence.
The logic is problem matching, not category matching. A creator making content about rental apartments is not a “home” creator. They make “small space, cannot drill holes, landlord won’t let me” content. Your narrow hallway cabinet belongs with them. It does not belong with a luxury-interior creator who owns a large house, even though both count as furniture accounts.
So read the last sixty days of a creator’s posts before you shortlist them. Look for what their audience complains about in comments. Comments give you the problem language far better than a bio does.
Where the Match Breaks Down
Two patterns account for most first-batch disappointment. Both show up inside a week.
The first is quality mismatch. Factory-direct pricing tempts you into the premium position you aspire to rather than the one the product has earned. Sell a well-made $40 item as premium and it reads as overpriced. Call it well-built and honest and it converts. Creators notice this faster than you will. Some quietly never post the product.
The second is commercial mismatch. Some audiences buy on price and will grind your margin through discount requests in the comments. If factory cost is your whole edge, that is survivable. If your margin is already thin because you cross-border ship a bulky item, a price-driven audience will sink you.
This is also where a tool layer earns its keep. DAMI’s sample management records which creator received which SKU, so a pattern of quiet non-posting becomes visible in the dashboard instead of living in group chat anecdotes.
The Sample Problem Factories Already Solved
Factory-backed sellers usually assume sampling is their advantage. In one narrow sense they are right: unit cost is low and stock is on hand. The advantage dies at the shipping step.
Sending two units domestically costs almost nothing in your normal operation. Sending them to a creator introduces customs paperwork and a courier bill that can exceed unit cost on anything bulky. Add a delivery window of one to three weeks. A coffee table sample can cost more to ship than to make.
That is why bulky-goods programs have a different shape from apparel programs. You will not sample broadly. You will sample narrowly, to fewer creators, behind a much higher screening bar. You will ask the creator to confirm their filming slot before anything ships.
What Changes When You Ship Overseas
Three things change, and all three argue for fewer, better-targeted samples.
Cost per sample jumps roughly 5 to 15 times versus a domestic parcel, for anything over 2kg. Time-to-content stretches because customs clearance sits between your warehouse and the creator’s door, and it is unpredictable enough that you cannot promise a filming date. Returns stop being practical. You will not ask a creator to ship a bulky sample back, so who receives one is effectively a permanent decision.
None of that means skip sampling. It means treat each sample as a purchase decision, not a fulfilment task.
Pricing and Commission When You Control Cost
Sellers who own production tend to make one of two pricing errors. They pull in opposite directions.
The first is pricing too low, anchored on manufacturing cost plus a familiar factory margin. Cross-border shipping and platform fees then eat everything, and the program dies at the commission step. The second is pricing too high, anchored on what the market leader charges. Without that brand equity, the listing converts at a rate too weak to carry any commission.
The useful number is neither. Work backwards from what comparable listings actually sell for. Then subtract your landed floor and the commission you can defend.
The Margin Floor Nobody Writes Down
Compute four numbers before you set commission. Landed cost per unit, including cross-border freight and duty. Platform fees and payment processing. Realistic returns as a share of revenue, not the aspirational figure. Packaging plus pick-and-pack labour. Whatever remains is fundable.
Then set the commission. The common error is picking a round number that feels normal for the category — 15%, 20% — and finding out later that contribution margin is negative per order. Choose the rate that holds your floor even after returns. Then judge whether that number attracts creators at all. If it does not, the product does not belong in a creator program. Better to know that before you sample than in month three.
One more practical note: track every rate and sample decision against the SKU, not against the month. Teams that do this in one place — DAMI’s targeted plan management keeps rates and sample status per product — spend far less time reconstructing why a batch underperformed.
One boundary worth stating. This arithmetic assumes you fund the creator channel out of product margin. If you are funding it out of cash elsewhere and hoping scale fixes unit economics later, commission tuning will not save you. Volume multiplies the per-order result. It does not invert it.
Running the First Batch: What to Actually Measure
A first batch of eight to fifteen creators is a measurement exercise. Decide in advance which number ends it. Otherwise you will extend it forever, because the results are ambiguous rather than bad.
Measure four things. Note that only the third is revenue.
| Metric | Good signal | The trap |
|---|---|---|
| Post rate (% of samples becoming content) | Above 60% | Strong post rate with weak sales means creator fit, not product fit |
| Views per post, median not average | Consistent across creators | One viral post hides nine flat ones |
| Orders attributed in 14 days | Any, spread across creators | All orders from one creator is a creator bet, not catalog validation |
| Return rate on creator orders | At or below site average | This is where content that oversells shows up |
Read the batch at day 14, not day 3. Creator video often keeps producing orders after the first 72 hours disappoint everyone. This is especially true once a video ranks for a search phrase.
Reading a Batch Before It Finishes
Two early indicators have held up across programs I have reviewed.
If post rate is high and orders sit near zero across many creators, the creators are rarely at fault. The SKU’s problem statement was clear to you and not to a stranger. That is usually a listing fix, not a resampling job.
If post rate is low, either your screening slipped or the product is hard to film. Both are cheaper to fix now than after you ship eighty samples.
From One Batch to a Standing Program
After the first batch, resist expanding the SKU list. The highest-value move here is usually deepening what already showed signal. Put more creators on the same three SKUs. Do not put twenty more SKUs in front of three creators each.
Depth does three things breadth does not. It gives you enough data on one product to tell signal from luck. It lets you refine the listing against real creator feedback. And it lets each creator see others covering the same product, which is the cheapest social proof available.
When Adding SKUs Stops Helping
Add SKUs once the winners have eight to ten active creators each and you start seeing the same ones re-post them. That is your signal the creator pool for those products is saturated.
The limit worth respecting is operational, not strategic. Cross-border bulky-goods programs tend to break around 30 to 40 SKUs with active creator programs. Demand does not run out. Sample logistics, returns handling, and listing variation management all scale superlinearly instead. A factory able to produce 200 SKUs cannot necessarily support creator programs for more than forty at once.
“If your constraint has moved from choosing SKUs to tracking who has which sample,” a Dongguan hardware operator running two US shops told me, “you have already outgrown spreadsheets. That was the month we started losing samples in the warehouse. It was not the month we started winning.” That matches what I have seen elsewhere. The ceiling is usually tracking, not selection.
The DAMI workspace handles that layer — bulk creator invitation, sample management, and AI outreach in Thai, Vietnamese, Indonesian, and Malay — for teams running several markets off one catalog. If you are already past twenty active SKUs, see how it works on the product page. If you are still shortlisting, the method for building a first fifty covers the sourcing side. You can also create your account and run these filters against your own export first.
Frequently Asked Questions
How many SKUs should I put in front of creators to start?
Between eight and fifteen. Fewer than eight and you cannot separate signal from luck across creators. More than fifteen and you are managing sample logistics instead of learning anything about demand.
Should I push my highest-margin product first?
Usually not, because it tends to be your most complex item. Run the three filters first — demonstrability, named buyer problem, ship-and-survive — and take what comes out, even at lower margin.
What if my minimum order quantity makes sampling impossible?
Sample from finished stock rather than running the line. Or find creators willing to work from detailed footage you shoot at the factory. Both are weaker than a physical sample. Worth knowing going in.
How long before I should judge a first batch?
Fourteen days from content posting. Video tied to search phrases often keeps converting well past the first 72 hours, which is exactly when most teams wrongly write a product off.
Do I need different SKUs for Southeast Asia than for the US?
Often yes, even off the same catalog. Smaller living spaces in Malaysia and the Philippines shift which dimensions are demonstrable. Price ceilings differ enough that one SKU can read premium in one market and mass in another.
