The agent guide settles the fee question: 3-6% boutique, $50-100 minimums. This ledger settles the rest. Agent economics over a year is decided by five hidden lines (FX spread, wire fees, storage days, placeholder QC, missed rebates) plus one structural choice: per-order vs contracted. The crossover sits near $25-30k/year of goods — below it, buy per-order and stay free; above it, a 2-3% contracted agent with rebate passthrough nets cheaper than any 5% per-order observed. And the $15k go-direct line is not about the fee at all — it is where a single PO's QC risk earns direct factory accountability.
The Fee Is Not the Ledger
Buyers compare agents on one number — the percentage — and then wonder why the "4% agent" year cost more than the "5% agent" quoted elsewhere. Reason: the percentage prices a service slice; the ledger prices your whole year through this agent. A real example shape, scale $40k goods/year:
| Ledger line | Per-order agent (5%, no retainer) | Contracted agent (2.5% + retainer) |
|---|---|---|
| Service fee on $40k goods | $2,000 | $1,000 |
| Retainer / account minimum | $0 | $600-1,200/yr |
| FX spread on funding | 0.8-1.5% card/PayPal funding: $320-600 | wire-in at batch rates: $0-100 |
| Wire fees, 12 orders/yr | $300-600 (order-by-order) | $600 (single annual funding, 6 wires) or $0 batched |
| Storage beyond free days | $200-500 (per-order pacing) | $50-150 (consolidation rhythm) |
| Rebates passed through | rare — walk-in pricing | 2-4% on volume accounts: $800-1,600 back |
| Net ledger | ≈ $2,820-3,700 | ≈ $2,250-2,750 |
The percentage race is inverted by the ledger: the 2.5% contracted agent, priced at half the headline fee, actually nets cheaper only above the crossover — below it the retainer and minimum funding requirements swamp the rebate upside. Same machinery as the $4,200 order teardown: quoted anchor lines tell less than the sum of quiet lines.
Five Places Agent Cost Hides
- FX spread on funding. Per-order agents charge your card/PayPal in your currency and convert at retail rates: 0.5-1.5% of goods value evaporates invisibly. Contracted agents take USD wires and convert at interbank-adjacent batch rates, or you pay RMB directly. On $40k/year this line alone is $300-600.
- Wire fees per transfer. $25-50 per wire × order count. Ordering weekly from three suppliers = 150+ wires across a year in the worst shape; contracted agents batch-fund monthly (12 wires) and absorb domestic RMB legs inside their account.
- Storage days beyond free window. Every agent warehouse gives free storage days (typically 7-30), then bills $0.30-0.80 per carton per day. Per-order buyers who stack small shipments wait and bleed; contracted agents consolidate on a rhythm and stay inside free days.
- Placeholder QC. The agent's included "QC photo per order" is a doorway check, free and worthless in depth. The cost shows up later as shippable defects — the economics of this line is a whole section below.
- Missed rebates. 1688 sellers run account-tier rebates (2-4% at volume) and seasonal coupon stacks. Walk-in orders routed through a per-order agent collect none; contracted agents negotiate at account level and (the honest ones) pass them through line-itemed. Ask for last quarter's rebate statement in writing — refusal to show it is a markup tell.
The Rebate Line Nobody Sees
1688's invisible pricing layer matters more than the agent's visible fee, and it explains the fee paradox: a 2.5% agent with rebate passthrough can land goods cheaper in total than a "free" buying app. The mechanics:
- Account-tier pricing. Volume accounts see lower listed unit prices on many 1688 sellers — 1-3% below the retail-browsing price. A per-order agent ordering as a walk-in sees the web price; a contracted agent ordering as a volume account sees tier price. Same SKU, different shelf.
- Coupon stacks. Platform coupons, seller coupons, category subsidies stack on Chinese-calendar sales days (618, Double-11, CNY sales) — up to 5-2% additional on qualifying volume, claimable only in RMB accounts. Per-order USD-funded purchases opt out of this layer entirely.
- The honest test. Ask any candidate agent for one reconciliation: same SKU, their unit cost vs your own browsing cost, screen-shared, this week. Agents who pass the test twice in a row keep it; agents who deflect with "our rate is still good" price the markup into the delta you cannot see.
Per-Order vs Contracted: The Real Crossover
Segment the buyer population by annual goods value and the answer is arithmetic, not opinion. The working crossover, with the logic visible:
| Annual goods through agent | Recommendation | Why the line sits here |
|---|---|---|
| < $8-10k | Per-order, platform model acceptable | A $600-1,200 retainer is 6-12% of volume; any rebate upside is smaller. Freedom to walk matters more than basis points. |
| $10-25k | Per-order with a boutique agent; negotiate minimums waived at this volume | Rebate upside exists but unproven until you repeat-order the same SKUs; lock nothing yet. Track the ledger quarterly. |
| $25-30k+ | Contracted dedicated agent at 2-3% + retainer | This is where FX, wire batching, storage rhythm and 2-4% rebate passthrough sum past $1.5-2k/yr — the retainer pays for itself with the volume argument, not the fee one. |
| Single PO > $15k | Skip the agent layer for that PO: go direct | The guide's threshold holds — but the economics behind it are QC risk and accountability, not % points. |
One honesty note: contracted agents love the phrase "priority production slots." Real contracted agents win you payment convenience, rebate layers and QC continuity; if the production-slot claim arrives, demand it as a written SLA with a dated penalty, or treat it as sales talk.
QC Economics Inside the Agent Relationship
The fee comparison treats QC as if it were a feature checkbox. It is a cost line with two very different shapes:
- Doorway QC (included, free): one photo set of carton exteriors. Catches wrong-color bulk errors, nothing else. Retained because it costs the agent minutes.
- Sampling QC ($3-8 per SKU): open-carton count, unit weight, spec sheet check, serial photo. This is the cheapest tier that catches the mixed-batch problem — the defect that eats entire e-comm reviews — and the regional price is standard.
- Full inspection ($80-150 per man-day): AQL-sampled, defect-classified, same rigor as a PSI. Pay this only on goods where one defect class can sink the shipment (regulated, fragile, high-ASP).
The ledger connection: doorway-QC agents are 0.5-1% "cheaper" headlined by often ship 3-6% defect-laden goods that a sampling tier would have caught and remade. The cheapest QC tier on this ladder pays for itself at roughly 10-15% of shipment value when defect risk is mixed-batch-prone — same defect-hump logic as the new-factory ramp. Buying "no QC" is buying deferred rework at full price.
What Switching Agents Actually Costs
Switching agents is priced mostly in time on the accounts that matter — the interaction with the factories, not the pallet transfers:
- Portable in a day: your product links, spec sheets, last orders' rebate records, warehouse stock (forwarder transfer $0.5-1.5/carton). The warehouses are interchangeable; loyalty has no technical lock-in.
- Half-reattachable: tier-price accounts and coupon eligibility are seller-relationship-bound. A new agent's account re-earns volume tier over 2-3 order cycles: budget 2-4% effective cost drift for a quarter during migration.
- Not portable at all: accumulated QC history (which batch was remade, which seller cheats on serials) and the negotiated exceptions (a seller who accepts odd packing for your SKU). This knowledge survives only in conversation logs; exporters with it discount heavily into freight rate comparisons without knowing what they're giving up. Export it deliberately: 30-minute handover doc before you exit.
When switching is justified: any two consecutive quarters where the ledger (build it quarterly — the table above is the template) shows the hidden lines eating a quarter of the headline fee, or any rebate-usability test of the honest kind failing twice.
Frequently Asked Questions
Is a per-order agent or a contracted agent cheaper?
Below roughly $8-10k of goods per year, per-order wins (zero retainer beats even a small one). Above $25-30k per year, a dedicated contracted agent at 2-3% with rebate passthrough usually nets cheaper than a 5% per-order arrangement once FX batching, wire batching, storage rhythm and rebates are counted. Between those bands, run per-order with a boutique agent and track the quarterly ledger.
Where do agent costs hide besides the percentage?
Five places: the FX spread when funding per-order in your card's currency (0.5-1.5%), wire fees per transfer ($25-50 each, unbatched), storage beyond the free window ($0.30-0.80 per carton per day), placeholder-only QC that misses mixed batches, and rebate lines unavailable to walk-in orders (2-4% at volume tier). Summed, they frequently exceed the headline fee delta that made the choice.
What are account-tier rebates on 1688 and who collects them?
Volume accounts get lower listed unit prices (1-3%) and stackable seller/platform coupons on RMB rails. Contracted agents ordering at account tier collect them; per-order agents ordering as walk-ins don't. Test any agent with a same-SKU cost reconciliation, screen-shared, twice — pass both times or price the delta into their real fee.
Does a bigger agent fee include QC, and is that QC any good?
Included QC is usually doorway-only (exterior carton photos): right color, right count of cartons, nothing deeper. It catches bulk errors only. For defect risk that matters — mixed batches, spec drift — sampling QC at $3-8 per SKU is the tier that pays for itself, roughly 10-15% of shipment value when mixed-batch risk is real.
How painful is switching agents if I find a better ledger?
Links, specs and stock move in a day (stock transfer $0.5-1.5/carton). Tier-price accounts need 2-3 order cycles to re-earn (budget 2-4% drift for a quarter). The irreplaceable asset is QC/negotiation history — export it as a 30-minute handover doc before leaving. Switch when two consecutive quarterly ledgers show hidden lines eating a quarter of headline fee, or the rebate test fails twice.
At what order size should I just go direct to the factory?
Per the agent guide's threshold: above ~$15k per PO, agent economics stop justifying the layer — not because of percentage points but because QC accountability, custom-line handling and payment leverage deserve a direct factory relationship (see the verification routine first). Repeat high-volume stock replenishment is exactly where the agent layer still wins.
Want your agent ledger audited?
Tell us your annual 1688 spend and current agent arrangement. We'll reply within 24-48h with where your hidden lines are flowing and whether your situation has crossed the contract crossover.
Request Ledger Review →