Why finding a good supplier doesn't end the risk
You spent three weeks vetting a factory in Guangzhou, ordered a $1,800 test batch, got clean units back, and locked in a standing order. Six months later the same SKU ships with visibly thinner fabric and the factory says nothing changed. Nothing changed on paper. The raw material supplier they use switched, or got squeezed on cost, and nobody told you because nobody had to.
This is the part most 1688 sourcing guides skip. They walk you through picking a supplier: check the shop age, check the transaction count, check the reviews, sample the goods. All useful. All a snapshot. A factory that scored well in March can look completely different in September, and not because you did the vetting wrong.
A few ways this plays out in practice:
- A factory that hit every deadline for two quarters suddenly starts slipping ship dates by 5 to 7 days once Double 11 season hits, because your $2,000 order sits behind a client placing $40,000 orders.
- The person you have been messaging on 1688 is a trading agent, not the factory. When the actual workshop has a problem (material shortage, worker turnover, a burst pipe), the agent finds out late and passes it to you later still, sometimes after your goods have already missed the ship window.
- Quality drifts slowly. One lot has a 2% defect rate, the next has 6%, nobody flags it because each individual shipment still "passed."
None of this shows up in a one-time supplier scorecard filled out at onboarding. It shows up in the operating relationship over months, which means you need a system that runs continuously, not a checklist you complete once and file away. That is the gap this framework fills.
The 3-layer framework: selection, monitoring, backup
Supplier risk on 1688 breaks into three distinct problems, and each needs its own mechanism. Trying to solve all three with a single "vet the supplier well" step is why operators keep getting surprised.
Layer 1, selection, reduces risk before you commit real money. This is where you filter out factories that were never going to be reliable, regardless of how the relationship unfolds later.
Layer 2, ongoing monitoring, catches degradation while it is still small. A supplier that was solid at selection can slide over time, and the only way to catch that early is to track a few numbers on every order, not just react when a shipment arrives damaged.
Layer 3, backup sourcing, limits the damage when something outside your control hits your primary supplier: a factory closes for a week during a provincial inspection, a bigger client buys up production capacity, a material cost spike triggers a 15% price jump with two days' notice.
Run these three layers as a repeating quarterly cycle: select and test new suppliers, monitor active ones on a schedule, and revisit your backup bench every quarter. Treat this as an operating rhythm, not a project you finish. Shops that skip the repeat cycle are the ones who get blindsided by a supplier problem they had six months of warning signs for and never looked.
Layer 1: vet suppliers before you commit to long-term orders
Start by figuring out whether you are talking to the actual factory or a trading agent (a "maoyi" account reselling from multiple workshops). Neither is automatically bad. Trading agents can be useful if you need small MOQs across several SKUs. But a trading agent means an extra layer between you and the source of any quality or capacity problem, and that layer adds lag to information you need fast. If a factory floods, you want to know in hours, not after a trading agent hears about it from someone else's order.
Check factory identity signals on the 1688 listing: does the shop show a real manufacturing address with photos that match, or just a showroom? Does the "supplier type" tag say manufacturer versus distributor? A quick way to test is to ask for photos of your specific SKU mid-production, not stock images. A real factory will send you a phone photo of your order on the line. A trading agent usually stalls or sends the same catalog shot twice.
Beyond that, look at the numbers 1688 actually gives you: shop age (2+ years reduces the odds of a fly-by-night operation), transaction volume in the last 90 days, and the review score, but weight review score less than volume. A shop with 4.9 stars and 40 transactions tells you less than a shop with 4.6 stars and 3,000 transactions.
Then test the MOQ claim. A listing advertising "MOQ 50 units" often means 50 units per color per size, which turns into 300+ units once you account for your actual assortment. Ask directly: "if I order 50 units in one color, one size, can you produce that as a standalone order." Get the answer in writing before you commit budget.
Whatever the answer, place a small test order first, even with suppliers who look excellent on paper. $200 to $500 is enough to check real production quality, real packaging, and real communication under a live order rather than a chat window. If you are ordering at a new volume threshold for the first time, this is also where a lot of operators get tripped up on things like sample fees and shipping terms; the mistakes are common enough that we wrote up the specific ones to avoid when ordering at a new volume threshold for the first time.
Layer 2: monitor factory quality on a regular schedule
Selection tells you a supplier was good on day one. Monitoring tells you whether they still are on day 180. Set your monitoring frequency by order volume: if you order weekly, check every lot. If you order monthly, do a monthly review. If you order quarterly, monitoring can run quarterly too, but no looser than that, because a full quarter of silence is enough time for a factory to change materials twice without you noticing.
Track three numbers on every order, even when nothing looks wrong:
- Defect rate per lot (units with visible flaws divided by units inspected). A jump from 2% to 6% between two consecutive lots is worth a message to the factory before your third order, not after.
- Ship date slippage (actual dispatch date versus committed date). One late shipment by 2 days is noise. Three consecutive lots late by 5+ days is a pattern, usually tied to the factory prioritizing bigger clients during a busy period.
- Response time on flagged issues. If you message about a defect and it takes the factory more than 48 hours to reply with a resolution, that is a signal worth logging, separate from the defect itself.
The value here is comparing lot to lot, not judging each shipment in isolation. A single bad lot happens to every factory occasionally. A downward trend across three or four lots is the thing you are actually trying to catch, and you only catch it if you are writing the numbers down somewhere consistent.
This is also where pre-payment quality inspection earns its keep, not as a one-off gate but as a recurring data point feeding your monitoring log. Every inspection report you already run before releasing final payment (assuming you are running one) gives you a defect rate and a photo record for free. If you are not yet inspecting goods before final payment, that process is the first piece to put in place, and we cover exactly how to structure it in our guide to checking goods quality before you release payment.
Score your 1688 suppliers on a regular cycle with a simple scorecard
Turn the monitoring data into a number you can act on, not just a folder of notes. A simple scorecard across five categories works better than an elaborate one you will not maintain:
| Category | Weight | What you're scoring |
|---|---|---|
| Price | 20% | Stability versus your last 3 orders, not absolute cheapness |
| Quality | 30% | Defect rate trend over the last 3 lots |
| Delivery speed | 20% | Actual ship date versus committed date |
| Communication | 15% | Response time and clarity when issues come up |
| MOQ stability | 15% | Whether stated MOQ terms hold across repeat orders |
Score each category 1 to 5, multiply by weight, and track the total over time. A supplier sitting at 4.2 in January and 3.1 in April has told you something concrete, even if no single order looked disastrous.
Run this review quarterly, timed to your actual restock cycle rather than an arbitrary calendar date. If you reorder core SKUs every 8 to 10 weeks, review right before you place the next big order, when the decision to keep, reduce, or drop a supplier still matters.
Set an action threshold before you need one, not after. A reasonable rule: a total score drop of 1.0 point or more in a single quarter triggers a conversation with the factory and a review of your backup option. A score below 2.5 in quality specifically, regardless of the total, is grounds to shift new orders to a backup supplier while keeping the primary on a shorter leash. Below 2.0 total for two consecutive quarters, stop placing new orders.
Assign ownership clearly. In a two-person shop, the person placing orders should own the scorecard, since they are the one seeing lot-level detail firsthand. Once a shop has a dedicated sourcing or ops person, that role owns it, but the founder should still see the quarterly summary. A scorecard nobody reviews is just a spreadsheet.
Layer 3: build backup so you're not dependent on one supplier
Single-sourcing a SKU feels efficient until the factory closes for a surprise inspection, gets bought out by a bigger buyer's exclusivity deal, or raises prices 15% on 48 hours' notice because a raw material cost spiked. None of these are rare. They are the normal cost of depending on one workshop for a product line that drives real revenue.
The fix is not "find a second supplier and forget about it." A backup supplier you have not touched in six months is untested, and the first time you actually need them is the worst time to discover their real MOQ or their real lead time. Identify a second factory for the same SKU during your normal Layer 1 vetting process (you are likely already sourcing quotes from 3 to 4 shops anyway), run the same small test order, and keep them active with a small recurring slice of volume rather than a one-time trial.
A working allocation for most shops: 70% of volume to your primary supplier, 30% to backup, adjusted by SKU risk. Higher-margin, higher-volume SKUs deserve a more even split, even 60/40, because the cost of a primary supplier failure is larger. Lower-volume SKUs can stay closer to 90/10, just enough to keep the backup relationship alive and their pricing current.
This is really a supply chain design question, not just a supplier relationship question, and it connects directly to the broader question of structuring your sourcing so no single factory can hold your revenue hostage. We go deeper on the allocation math and factory relationship structure in building a 1688 supply chain that isn't dependent on one supplier.
Put the framework into your daily sourcing workflow
The three layers only work if they attach to steps you already do every week, not a separate process that competes for time.
At the ordering step, pull up the supplier's current scorecard before confirming a new PO. Takes 30 seconds and stops you from quietly increasing volume with a supplier who is trending down. At the negotiation step, use your monitoring data as leverage: a supplier with a slipping delivery record has less standing to hold firm on price. If you are pushing back on cost, our breakdown of negotiating price with 1688 suppliers covers how to use order history as a bargaining position, not just asking for a discount cold.
At the packing and outbound shipping step, log the actual ship date against the committed date, since this is the easiest data point to lose track of once goods are moving. At the point goods land and clear import, that is when your defect count and inspection notes need to get logged against the same supplier record, closing the loop back to the scorecard.
For a shop running 5 to 10 SKUs, a shared spreadsheet with one row per lot (supplier, date, defect rate, ship slippage, notes) covers this fully. It does not need to be complicated to be useful. As order volume grows past what one person can track by memory, that is the point to move into a dedicated sourcing tool, which is part of what we are building with Ordinex Scout.
Set your escalation threshold in advance so it is not a judgment call made under pressure: two consecutive quality flags, a scorecard drop of 1.0 or more, or a single ship delay past 10 days should each independently trigger a move from "monitor" to "act," meaning either a hard conversation with the supplier or shifting volume to backup. None of this happens in a vacuum either. Supplier reliability affects your landed cost directly, since late shipments mean rushed air freight instead of budgeted sea freight, and quality issues mean returns and rework you did not price in. If you have not mapped your full landed cost structure recently, our guide to calculating 1688 import fees is a useful companion to this framework, since supplier risk and cost control are really the same spreadsheet viewed from two angles.
FAQ
How many suppliers should a small shop realistically manage per SKU? Two is the minimum that actually functions: one primary, one active backup getting real volume. A third becomes worth adding once a SKU crosses roughly $10,000 to $15,000 in monthly order value, where the cost of concentration risk outweighs the extra coordination overhead.
What if my supplier refuses to send WIP photos or share more transparency? Treat refusal as data, not paranoia on your part. A factory confident in its own process usually has no issue sharing a mid-production photo. Repeated refusal, especially paired with vague answers about MOQ or lead time, is a reason to keep orders small regardless of how good the finished samples look.
Is this framework overkill for a shop doing under $5,000 a month in 1688 orders? Scale it down, don't skip it. At that volume, a lightweight version works: track defect rate and ship date on every order in a basic spreadsheet, and do a 15-minute quarterly review instead of a formal scorecard. The habit matters more than the sophistication.
How do I know if a quality drop is the factory's fault versus a one-off shipping or QC issue? Look at pattern versus incident. One bad lot with a clear external cause (a shipping delay, a specific batch defect the factory already flagged) is an incident. A defect rate that climbs across three consecutive lots with no external explanation is the factory's process changing, and that is exactly the signal Layer 2 monitoring is built to catch.
If you are managing more than a handful of 1688 suppliers across active SKUs, this is the kind of tracking that gets hard to hold in a spreadsheet past a certain point. We are building Ordinex Scout to handle supplier scorecards and order-level monitoring automatically, and Ordinex Orders to connect that data straight into your purchasing workflow. Both are in private beta right now. If you want early access, reach out through ordinex.cc.