Home › Field notes › Sourcing Modular Bags from China: A Buyer Qualification Guide

Modular Bag Pro · 2026-10-10 · 2,863 words · 13 min read

Sourcing a modular range from China is a documentation exercise long before it is a purchasing one, and buyers who treat it as a shopping trip pay for the difference later. The sequence that works is fixed: close the brief, freeze the specification lines that move price, shortlist against weighted criteria, then use sampling as the instrument that decides. MOQ 500 per reference sets the commercial floor, and AQL 2.5 under ISO 2859-1 sets the release rule. The single document that determines whether a range is still extensible in three years is the interface drawing, built on 25 mm webbing at 38 mm vertically with a 50 mm lateral pitch. This guide is written for a buyer placing a first order rather than one managing an established programme, so it favours checks that close in a week over audits that take a quarter. Every criterion is something a serious partner can answer on paper.

What sourcing from China involves in practice

Sourcing is the work of converting an idea into a set of documents that someone else can execute repeatedly. The physical production is the smaller half of that work; the larger half is the specification, the sample approval, the material decisions and the inspection agreement. Buyers who understand this spend their effort where it pays.

The geography matters for one reason rather than several. China concentrates the overwhelming majority of bag production capacity, along with the material supply, hardware supply and finishing services that sit around it. A production partner in China can typically source fabric, webbing, hardware and trim within a short radius, which is what makes a 35-50 day build window achievable at all. That cluster effect is the practical argument for the geography, and it is an argument about lead time and material availability rather than about anything else.

What does not follow from the geography is any assumption about quality. Quality is a property of a specific production partner's system: its incoming control, its in-process checks, its inspection discipline and its documentation. Those are evaluated the same way anywhere, and they are evaluated from paper before they are evaluated from a sample.

The commercial frame deserves stating once, because it shapes everything downstream. MOQ 500 per reference, a sampling stage of 6-10 working days reaching 12-15 on complex builds, bulk output over 35-50 days, release at AQL 2.5, and terms at T/T 30/70 against FOB Xiamen. Buyers who know that frame can tell quickly whether a counterparty is describing a standard arrangement or inventing one.

Selection rule: Selection rule: Write the brief before making contact, and include duty profile, interface geometry, target landed cost and the acceptance limit at AQL 2.5, because a production partner asked to guess a 500 unit programme will quote a guess.

Closing the brief before making contact

The most expensive hour in a sourcing programme is the hour spent emailing before the brief is closed. A quotation requested against an incomplete specification is priced against assumptions, and those assumptions surface as change requests after the deposit has been paid, which is the worst possible time to discover a disagreement about scope.

A closed brief answers five questions. What is the chassis, in dimensions with tolerances? What is the interface, with a geometry, a tolerance band and a version? What materials, named by class and by brand where hardware is concerned? What does the product have to survive, expressed as a test method per claim? And what is the destination market, which sets the compliance file?

One field deserves emphasis because it is routinely left blank. The duty cycle, stated as trips per year, load in kilograms and the cleaning the product will experience, is what turns a material discussion into an engineering one. Without it, fabric selection becomes a matter of taste, and taste is a poor basis for a specification that has to survive three years.

Bottom line: Bottom line: Close the brief internally first, and treat any specification still changing after the quotation is requested as the main cause of both the price movement and the extra 6-10 working day sampling round that follows.

Shortlisting against weighted criteria

Shortlisting is a scoring exercise, and scoring works only if the criteria are written down before the first conversation. Otherwise the loudest presentation wins, which is a reliable way to select the wrong partner.

Two criteria are diagnostic out of proportion to their apparent weight. Interface gauging capability tells a buyer whether a modular range will actually work, because spacing drift is invisible without a gauge and fatal with one. And communication responsiveness, measured on technical questions rather than on commercial ones, predicts how the relationship will behave the first time something goes wrong.

Our vetted partner facilities are assessed against a list very close to this one, which is worth stating plainly: the criteria are not a test of cleverness, they are a checklist that any serious operation passes without strain. A counterparty that struggles with three or more of them will struggle later with something that matters more.

Sampling as the primary qualification instrument

Sampling is where claims become objects, and it is the only stage at which the counterparty has to do engineering rather than sales. Everything learned here is worth more than everything in the brochure, and the cost is a sampling fee of USD 50-150 per reference, refunded against the order.

Timing sets the first filter. A conventional reference takes 6-10 working days; a build involving a moulded panel, a bespoke pull or a welded pouch takes 12-15. A sample delivered materially inside that span deserves a dimension-by-dimension comparison with the specification, since a fast sample is often a stock unit rather than a build.

The review itself should be a measurement exercise rather than an opinion. Check the dimension callouts against tolerances. Check the interface repeat with a gauge across the full panel width, not at one convenient point. Check the bar-tack coverage at each anchor against the stated area. Check the closure by cycling it and noting resistance. Record each result against the specification line it belongs to. That record is what makes a second sample round productive instead of circular.

Corrections should be documented. Every iteration should carry a revision note identifying what changed and why, and the specification document should be updated to match the physical unit. Without that habit, the approved sample and the approved specification drift apart, and the production run follows whichever one the production floor happens to have.

Two sample rounds is normal; four suggests the brief was not closed. When a third round is needed, the useful move is to stop and re-read the brief, because the cause is almost always an ambiguous specification line rather than an execution failure.

Verdict: Verdict: Treat the sample as the primary qualification instrument, and measure the sample against the drawing rather than against the photograph, because a sample that looks right but measures 25 mm webbing at the wrong pitch will reproduce the error at scale.

The interface specification, one document that decides compatibility

For a modular range, one document matters more than all the others: the interface drawing. It fixes the geometry, the tolerance band, the reinforcement behind the mounting face, and the version. Everything else in the range can change without stranding the catalogue; this cannot.

The grid published on this site runs on 25 mm tape with 38 mm row spacing and a 50 mm pitch. Adopting it gives instant fit with pouches many buyers already own, which removes the empty-catalogue problem that every new modular range faces: on day one the chassis is useful, since buyers may already hold pouches that will fit it.

Tolerance is what makes the drawing enforceable. A repeat stated as a nominal figure cannot be inspected, because there is no band to measure against. A repeat stated as a band can be gauged, and gauging across the full width of a panel catches the drift that a single reading hides. Drift is rarely uniform; it accumulates toward the edges of a panel where the fabric is under the most tension during sewing.

Versioning closes the system. each pouch names the revision it was cut for, each chassis states the revisions it will still accept, and the register lives in the brand's specification set rather than only with the production side. That single discipline prevents the costliest modular failure imaginable: a silent change to the grid nobody announced, which voids an installed accessory catalogue months after the change ships.

Commercial terms and where the risk sits

Standard terms in this category are T/T 30/70 against FOB Xiamen. Thirty per cent moves on order and releases material and capacity. Seventy per cent is paid against release after final inspection passes. The structure is ordinary, and its value is in what it protects: the buyer holds quality leverage until the goods are finished.

Incoterm choice determines who carries what. Under FOB Xiamen the seller's obligation ends at the ship's rail at the named port, and the buyer carries main freight, insurance, import duty and onward delivery. That suits a buyer with a forwarder and a known lane. A buyer without either should say so early, because the alternative arrangements are priced differently and asking at quotation stage is far cheaper than discovering the gap at arrival.

Pricing should always carry its basis. An indicative figure written as FOB Xiamen at a 500-unit minimum is comparable between counterparties. A bare unit price is not, because it may or may not include setup, inspection, packaging or freight. Setup itself sits in a band of USD 300-2,500 depending on how much is new, and it should be itemised rather than folded into a unit figure.

Release discipline is the buyer's remaining leverage and should be written into the order rather than assumed. AQL 2.5 under ISO 2859-1, defect classes defined in the brief, and interface fit carried as its own class because a visual check passes a panel whose pouches jam. With that in place, the seventy per cent balance is contingent on a measurement rather than on an assurance.

Takeaway: Takeaway: Fix the delivery term, the inspection trigger and the contract currency before the deposit, and keep the balance under T/T 30/70 tied to a written AQL 2.5 result rather than to a shipping date.

Compliance file for the destination market

Compliance depends on the market and belongs fixed before artwork is finalised, because it drives labelling content, and relabelling at destination is expensive and slow. The file is short but each item has to name the actual product.

For Europe, REACH (EC 1907/2006) substance restrictions apply, and a declaration should be supported by test evidence naming the material batch. Nickel release to EN 1811 applies to metal components in prolonged skin contact. Where a brand asserts textile safety, OEKO-TEX Standard 100 proof must list the actual article rather than a generic fabric reference, because a certificate covering a different construction supports nothing.

For the United States, CPSIA applies where a product is positioned for children, and California Prop 65 considerations arise where a listed substance appears above the stated threshold. Fibre composition, origin marking and care symbols are routine in both markets and are the fields a compliance reviewer checks first.

Performance claims need method references, not adjectives. Abrasion to ISO 12947 with a cycle count and a failure criterion. Water resistance to AATCC 127 as a hydrostatic head. Grab strength to ASTM D5034 on the assembled unit. Corrosion on finishes to ASTM B117. Each converts a marketing word into something a laboratory can settle, which protects the brand as much as the buyer, because an unenforceable claim is a liability rather than an asset.

Running the first order as an instrumented trial

A first order of 500 units is not a purchase, it is an experiment designed to remove uncertainty from the second order. That framing changes what a buyer asks for and what they measure, and it is the difference between a programme that improves and one that repeats.

Before the order ships, decide what will be measured. Attach rate, meaning the share of chassis buyers who later buy a module. Return reasons, coded by failure mode rather than by customer sentiment. And interface complaints logged separately from cosmetic ones. Those three numbers say whether the grid is correct, whether the assembly holds, and whether the finish survives.

During the build, ask for the artefacts that make the trial legible: the first-piece approval record, the in-process inspection results, and the final inspection report at AQL 2.5 with defect counts by class. Those documents cost nothing and convert an opinion about quality into a measurement of it.

After arrival, hold a structured review against the brief before planning the second buy. Which specification lines held, which drifted, and which were never actually measured? The answer usually identifies two or three lines that need tightening, and tightening them before a 1,500-unit reorder is far cheaper than tightening them after.

The second buy is where the modular model pays. Concentrate volume in the colourway and module that the data supports, keep the interface unchanged so the installed base stays compatible, and use the scale to amortise the setup already paid. That is a materially better position than a second full-range buy, and it is available only to ranges that ran the first one as a trial.

Questions worth asking, and the shape of a good answer

Qualification correspondence is short, and four questions carry most of the information. Each has a good answer, and the shape of that answer matters more than its content.

Ask what happens when a lot fails inspection. A good answer names containment, a rework or re-inspection path, and who pays. A vague answer about working together is a warning, because the situation is not hypothetical; at AQL 2.5 some lots fail, and the response to that is the part of the relationship that matters.

Ask which materials are already booked. A production partner holding fabric for a comparable programme can name the mill and the delivery week. That answer is what makes the short end of a 35-50 day build window credible, and its absence is what makes a quoted completion date an aspiration.

Ask how the interface is measured. The good answer describes a gauge, a stated band, and a reading taken across the full width of the panel on the first-off unit. An answer describing a visual check tells a buyer precisely how spacing drift will be handled, which is to say not at all.

Ask what the counterparty would decline to do. This is the most revealing question in the set, because a production partner that names a tolerance it will not hold, or a finish it cannot qualify, is supplying information that costs nothing and prevents a failure. One that accepts everything has either not understood the specification or has not considered whether it can meet it, and both states produce the same outcome later.

Two references belong in the brief folder: the grid drawing and the reference products the new range is being compared against.

Spec rule: Spec rule: Run the first order as an instrumented trial at 500 units, record defect classes by cause, and hold the second order until corrective actions from the first have been verified on a follow-up inspection.