Sourcing: finding suppliers, testing samples, negotiating right

• Look at the range: a real manufacturer has a coherent range (200 variants of ONE product type). A trader sells spice grinders next to phone cases.
• Use the filters: “Verified Supplier”, years active, reviews. Verified profiles include factory videos and audit reports — watch them, don't just tick the box.
• Ask questions only manufacturers can answer: monthly capacity, material options, tooling costs for customizations. Vague answers = trader.
• A good trader isn't forbidden, by the way — they cost margin but can be more flexible at small volumes. You just need to KNOW who you're talking to.

• Who you are: two sentences — Amazon seller from Germany/Europe, building a brand (no need to fake size; professionalism counts more).
• Exactly what you want: the product with specification (dimensions, material, color) and your customizations (logo, packaging, the 2–3 differentiation levers from lesson 6).
• Your questions: price tiers for e.g. 300/500/1,000 units, MOQ, production time, sample cost and time, existing certificates/test reports (EU requirements!).

Write to 10–15 suppliers — the answers (speed, precision, English, flexibility) are your first quality test. After a week you'll have 3–5 serious candidates.

• Order samples from your best 2–3 candidates — in parallel, not sequentially. Sample costs (often €20–80 incl. express shipping) are research budget, not waste.
• Check against your written specification: measure, compare materials, weigh. Anything you didn't specify is left to chance in mass production.
• Real-life test: use the product for two weeks. Dishwasher, drop from the table, sustained load — your customers test harder.
• Order the best competing product from Amazon and compare directly. Your product must win that comparison — otherwise back to lesson 6.

• Tone: polite, long-term, specific. Manufacturers think in relationships — those who only squeeze the lowest price get lowest-price quality.
• Payment standard: 30/70 — 30 % deposit at order, 70 % after production (ideally: after passed quality inspection, BEFORE shipping). Never 100 % upfront.
• Pay through Alibaba Trade Assurance: the contract incl. specification then sits with Alibaba, with a dispute process behind it. Suppliers pushing around it (“bank transfer to private account, better price”) are a red flag.
• Everything in writing: price, quantity, specification, production time, late-delivery clause — inside the Trade Assurance contract.

One page that settles every argument: dimensions with tolerances, material (specific: “304 stainless steel”, not “stainless steel”), colors as Pantone codes, weight, logo placement with print file, packaging (size, material, print), contents item by item — and the mandatory markings on product/packaging: your address as responsible person (GPSR, lesson 2), warnings, CE if applicable, EAN print. This sheet goes to the supplier, to the inspection service and later to every new supplier.

• Minimum (small first order): photos and videos of the finished goods BEFORE the 70 % payment — product, packaging, carton print, random inside shots.
• Standard (from ~500 units or critical products): a pre-shipment inspection by an independent service (around €250–350): AQL-based sampling, report with photos. Those €300 are cheaper than a single pallet of scrap in Amazon's warehouse.

• Price far below every other quote (quality or existence of the product questionable).
• 100 % upfront payment or pushing you off the platform.
• No video call, no factory photos on request, evasive answers on capacity.
• Certificates/test reports “coming later” or visibly patched together.
• Great sample, but a written specification is called “not necessary”.

The pre-shipment inspection from this lesson almost always follows the AQL method (ISO 2859-1): a sample is inspected, its size derived from the lot size via a standard table — for 281–500 units at the usual severity (Level II) that is 50 units. Defects found are sorted into three classes, each with its own acceptance limit:

Two expensive misunderstandings follow from this:

• “Passed” does not mean defect-free. A passed report with 3 major defects in the 50-unit sample means a 6 % defect rate in what was inspected — the 500-unit lot can contain 20–30 defective sets and still be fully within the standard. For the AURELO set that is 20–30 return and 1-star candidates, each costing you ~€25 in refunds plus fees and review damage. AQL 2.5 is also no promise of at most 2.5 % defects in the lot — it is the process quality the method will just barely keep accepting over time; individual worse lots slip through, because 50 units are statistically a coarse sieve.
• The knee-jerk response of “then demand AQL 0” does not work. Zero accepted defects across all classes would effectively mean 100 % inspection — unaffordable, and no serious service will promise it in a sampling procedure. The effective lever is classification: YOU define in the spec sheet which defect is critical, major or minor for your product. A jamming grinder mechanism sits on generic defect lists as major — for your business it is a critical case, so you upgrade it and set its acceptance to zero.

And the method's limit: the inspection is a snapshot of a sample, not a guarantee and no substitute for warranty claims — it shifts no responsibility. Its value lies in the timing: it finds problems while the open 70 % payment still puts the leverage in your hand.