Rewritten 12 August 2026 — rebuilt on two named factories' own published terms, a carrier service document, a State Council decree and the holiday notice under it, a port dwell report, a port tariff and a schedule-reliability release. The old version priced three kinds of sample and timed a seven-step calendar without a source behind a single figure, and its finish date was earlier than the published ocean crossing on its own. Those numbers are withdrawn below rather than quietly adjusted.
A note on who wrote this. Backper sells custom backpacks and this page previously offered our own sample box at "$84 DDP, credited back on the first bulk order". We are not restating that as though it were a market price, and the sentence had a hole in it that this article now argues is the most important sentence in any sampling quotation. It did not say what the credit excludes. The only factory we could find publishing its refund policy in writing states that in the refundable model, mould and material fees charged by third parties are excluded from the refund — and the only other factory we could find publishing its sampling terms puts a tooling range of up to $800 on a single custom part.
So "credited back on the first bulk order" can be entirely true and still leave you paying for the sample. That correction applies to our own offer as much as anyone's, and the fix is the same in both directions: the exclusion goes on the quotation, in writing, before the sample is cut.
Every guide to sampling in this category, including the one this page used to be, is written as an inspection checklist. Spray the fabric, count the stitches, work the zips, hold the thread against a Pantone chip. All of it is sensible and none of it is the point. A pre-production sample's real function is not to show you what the bag looks like — you can see that in a photograph, and by the time you are holding one you have usually decided. Its function is to become the reference against which a disagreement about the bulk shipment gets settled. That reframing is ours and not a sourced finding — no standard, trade body or manufacturer we could find states it, which is the subject of section 1. The bag is the easy part of that. The paperwork around the approval is the whole of it.
That reframing changes what matters. If the sample is a preview, the interesting question is what it costs and how long it takes. If the sample is evidence, the interesting question is what exists in writing at the moment you approve it, because that — not the bag on your desk — is what you can hold a factory to three months later when 2,000 units arrive in a shade you did not agree. On the sourced numbers below, the approval decision is worth several times the sample fee and the fee is the thing every article talks about.
We should say at the outset that this is a poorly documented corner of the trade, and this article is shorter on citations than it should be. Two named manufacturers publish their own sampling terms in quotable detail. Beyond those two pages we found nothing: no trade association, no industry body, no standards document, no second factory willing to print a number. Where the old version filled that silence with figures, this version leaves it empty and says so.
1. Almost nothing about sampling is documented, and that is the first finding
We went looking for the ordinary things an article like this should be able to cite. How long a pre-production sample takes. What one costs. How many rounds a typical programme runs. What tolerance a factory works to on a logo placement. What happens to the approved sample afterwards, and who keeps it. Whether there is any recognised convention for how a bulk shipment is compared back to it.
Two pages answered any of it, and both are factories publishing their own commercial policy: Quanzhou Osgoodway Co., Ltd, in an article dated 13 May 2026, and Quanzhou Huafang Bags Co., Ltd, in one dated 25 September 2023. A factory's own published terms are a primary source for the fact that this factory charges this and takes this long. They are not an industry benchmark, and two of them are not a sample of anything.
| Question | What we found | Source |
|---|---|---|
| How long a sampling round takes | Around 5–10 days, varying by style and requirement | Huafang, 25 September 2023 |
| What a first pre-production sample costs | $120–$300 | Osgoodway, 13 May 2026 |
| What a revision costs | Minor adjustments usually free; major design changes $80–$150 | Osgoodway, 13 May 2026 |
| What tooling costs | $200–$800 one-time for custom hardware or an EVA mould, with a worked zipper-pull example at $650 | Osgoodway, 13 May 2026 |
| Whether a sample fee is refundable | Two named models, with the exclusions stated | Huafang, 25 September 2023 |
| When the factory's own lead-time clock starts | At locked tech pack (OEM) or confirmed style selection (ODM) | Osgoodway, 13 May 2026 |
| How many rounds a programme typically runs | Nothing found | — |
| What placement or dimensional tolerance is normal | Nothing found | — |
| What happens to the approved sample afterwards | Nothing found | — |
| Any trade-body documentation of the artwork-to-sample sequence | Nothing found, from any industry association | — |
| Published laboratory testing turnaround, if your programme has a test gate | Nothing found with a dated working-day count | — |
The five empty rows are the honest state of the evidence, not a gap in the searching. They are repeated in the unverified list at the end of this page so they cannot be read past.
That inventory is why the reframing at the top of this page is offered as an argument rather than a finding. We cannot cite anyone for the claim that a sample's primary function is evidentiary. No standard says it, no association has written it down, and neither of the two factories that publish anything at all addresses what the approved sample is for after it has been approved. What we can show is the shape of the numbers around it, and the shape is consistent: the money and the calendar both sit somewhere other than where the checklist articles point.
What this means for you. Ask any supplier — including us — to name the document behind a sampling number before you plan against it. If the answer is a range with no source, treat it as a negotiating position rather than a fact, and put your own terms in writing instead of accepting theirs.
2. The calendar this page published finished before the ocean crossing did
The old version of this page gave a seven-step sample-to-shelf timeline: brief and deposit on day 0, first sample by day 7 to 10, feedback by day 12 to 14, revised sample by day 18 to 21, approval and production start by day 21 to 23, bulk complete with QC photographs by day 35 to 42, and delivery by day 42 to 56. It is a tidy artefact, every number in it was ours, and nothing sat behind any of them. Set it against the only figures a factory or a carrier has actually published for the same journey.
| Route | Factory leg, to FOB | Ocean leg, port to port | Total | Source |
|---|---|---|---|---|
| What this page used to publish | Brief to delivered, all steps | 56 | None. Our own claim, withdrawn | |
| ODM, fastest published case | 45 days | 40 days, Shanghai–Rotterdam | 85 | Osgoodway 13 May 2026; CMA CGM FAL 8, April 2026 |
| ODM, slowest published case | 60 days | 58 days, Yantian–Felixstowe | 118 | as above |
| OEM, fastest published case | 90 days | 40 days, Shanghai–Rotterdam | 130 | as above |
| OEM, slowest published case | 120 days | 58 days, Yantian–Felixstowe | 178 | as above |
Three things in this table are ours rather than the sources'. The totals are our addition of two published figures. Osgoodway measures to FOB shipment, so we have assumed the sample rounds sit inside its window rather than before it — the conservative reading, and an assumption the source does not make. And Osgoodway quotes FOB Guangzhou while CMA CGM's document lists Yantian sailings; pairing the two is our inference about which port a factory quoting FOB Guangzhou ships from. Osgoodway is a Quanzhou company, so its own location is not the evidence for that pairing; the FOB term it publishes is. Every ocean figure is port-to-port and therefore excludes origin cut-off, destination clearance and inland delivery, so each total is a floor and not a door date.
The comparison that ends the argument is the last row against the first. Yantian to Felixstowe is 58 days on CMA CGM's own FAL 8 service sheet, marked April 2026, routed via the Cape of Good Hope. The old page's entire calendar, brief to delivery, was 56 days. The whole programme finished two days before the ship would have docked. That is not a stale estimate that has drifted; it is a number that was never checked against anything.
The 5–10 day sampling figure Huafang publishes is the only duration in this area we can cite, and note what it is: 5–10 days per round, for the sampling itself. It says nothing about the artwork iteration before it, the courier time to get the sample to you, the days it sits on your desk while three people look at it, or the revision cycle after. The old page's "Day 7–10: first PPS arrives" quietly folded all of those into the factory's number.
Our own figure of 7 to 21 days for a pre-production sample is withdrawn for the same reason: it has no source, and it sits alongside the one published duration in a way that implies agreement where none was established. If a supplier tells you fourteen days, ask whether that is fourteen days of sampling or fourteen days from your last email, and whether the courier is inside it.
What this means for you. Plan the sampling phase in rounds, not in a total, and give each round its own dated gates: artwork frozen, sample cut, sample shipped, sample received, decision returned. The published 5–10 days covers one of those five. A calendar that reaches a delivery date without naming a carrier service is telling you it has not been built from anything.
3. The largest published cost at the sampling stage is the one the old page never mentioned
Three price ladders appeared on this page before today: a pre-production sample at $80 to $300 and above, a stock or catalogue-match sample at $40 to $120, and a multi-option sample box at $50 to $150. None carried a source. We went looking for anything published that could confirm or contradict them, and found one factory's page.
| Line | Published range | Status |
|---|---|---|
| Custom hardware or EVA mould, one-time | $200–$800, with a worked zipper-pull example at $650 | Sourced — Osgoodway, 13 May 2026 |
| First pre-production sample | $120–$300 | Sourced — Osgoodway, 13 May 2026 |
| Major-revision re-sample fee | $80–$150; minor adjustments usually free | Sourced — Osgoodway, 13 May 2026 |
| Pre-production sample, as we published it | $80–$300 and above | Withdrawn. No source. Its floor sits $40 below the only published range |
| Sample box, as we published it | $50–$150 | Withdrawn. No published counterpart found anywhere |
| Stock or catalogue-match sample, as we published it | $40–$120 | Withdrawn. No published counterpart found anywhere |
All figures are US dollars as published by the source. No currency conversion is applied anywhere on this page and none is implied; where a sterling figure appears later it is never added to a dollar one. One factory publishing its own commercial policy is primary for its own terms and is not a market benchmark: n = 1, and the range it publishes is what one company charges.
Read the chart top to bottom rather than as a price list. The tooling range tops out at $800, against $300 for a first sample and $150 for a major revision. At the top of each published range, the tooling costs more than the sample and the revision together, and Osgoodway's own worked example of a $650 zipper pull costs more than two first samples at the top of the range. Yet the old version of this page discussed sample fees in three separate places and mentioned moulds nowhere.
That mis-framing survives the withdrawal of the numbers. The sample fee is a small, visible, negotiable figure that arrives early, so it dominates the conversation. The tooling charge arrives in the same email, is larger, and is described as a one-time cost — which makes the biggest line on the page sound like an administrative detail.
Our own withdrawn pre-production range is worth a sentence because of how it failed rather than that it failed. It ran $80 to $300 against a published $120 to $300 — the same ceiling, a floor forty dollars lower. A number invented to be attractive tends to land just under the real one, which is exactly the pattern that makes an unsourced range hard to spot. The two ladders with no published counterpart at all could not even fail that test.
What this means for you. Ask for the tooling line before the sample line and get it quoted separately, with the part it belongs to named. If the bag needs a custom pull, a moulded buckle or a bespoke logo plate, the published range says that decision is worth more than the entire sample budget — and unlike the sample fee, it is a design choice you can still reverse at this stage.
4. "Free sample" and "credited against your bulk order" are different offers, and the difference is written down
Huafang's page is the only document we found in which a factory sets out how sample money actually works. It describes two models, and the distinction between them is the most commercially useful thing we could source about sampling.
| Model | What the factory states | What is excluded |
|---|---|---|
| Free-sampling model (refundable deposit) | "The sampling deposit collected before sampling will be fully refunded to the customer when they place a bulk order", or applied as payment credit | Mould and material fees from third parties are excluded from the refund |
| Paid-sampling model | "The sampling fee will generally not be refunded to the customer after the sampling is completed", with costs disclosed up front | Not applicable — nothing is refunded |
Quanzhou Huafang Bags Co., Ltd, published 25 September 2023. Both quotations are the factory's own words in its own English rendering. This is one company's stated policy and we are not presenting it as a trade norm; what makes it useful is that it is written down at all, which lets you ask a different supplier the same question in the same terms.
Put the exclusion next to the tooling range from the previous section and the arithmetic is unavoidable. The refundable model refunds the deposit and not the mould. On the only published tooling figures, that leaves $200 to $800 outside the refund, with a worked example at $650 — against a first sample of $120 to $300 that is inside it. A supplier offering a free sample on a bag that needs a custom zipper pull is offering to absorb the smaller half of the number.
This is not an accusation of bad faith and the Huafang page is not hiding anything: it publishes the exclusion plainly, which is more than most. The problem is that the offer is usually communicated in four words in an email — "samples are free, obviously" — and the four words are compatible with three quite different commercial arrangements.
The three arrangements a buyer should be able to tell apart. First, a genuinely free sample, where the factory absorbs the cost as a cost of sale. Second, a refundable deposit, where you pay now and it comes back as credit when the bulk order lands — and, on the one published policy we could find, third-party mould and material charges do not come back at all. Third, a paid sample, where the fee is disclosed up front and is simply not refunded. The three feel similar in a conversation and differ by several hundred dollars on the published ranges. The question that separates them is not "is the sample free" but "which charges are outside the credit, and what happens to them if I do not place the bulk order".
There is a second question the sources cannot answer at all, and it belongs to this article's argument. None of the material we found says who owns the approved sample, whether the factory retains a counterpart, or how a bulk shipment is formally compared back to it. The fee model tells you who paid. It does not tell you what you have bought. If the sample is going to be the reference in a disagreement, the retention arrangement matters more than the refund, and it is the term nobody publishes — which means it is a term you have to write yourself.
What this means for you. Get the fee model named in the quotation, with the exclusions listed by line rather than by category, and add a retention clause of your own: a sealed, signed and dated counterpart sample held at the factory, and one held by you, both referenced by the same identifier. Nobody will offer this. It costs the factory a bag.
5. The lead time you were quoted starts at your approval, not at the factory's
Osgoodway publishes two lead times and the wording around them is worth more than the numbers.
| Route | Published duration | Measured from | Published MOQ |
|---|---|---|---|
| OEM — your specification, made to order | 90–120 days to FOB shipment | "From locked tech pack" | 500–1,000 pieces |
| ODM — the factory's style, branded for you | 45–60 days to FOB shipment | "From confirmed style selection" | 200–300 pieces |
Quanzhou Osgoodway Co., Ltd, published 13 May 2026, modified 15 May 2026. The same page gives a worked unit cost of $18–$26 FOB Guangzhou for a 30-litre daypack at 500 to 1,000 units. One factory's published terms; not an industry average, and not a quotation for your programme.
Both clocks start at an event that is yours. A tech pack is locked when you stop changing it. A style is confirmed when you confirm it. Everything before that moment — artwork iteration, material selection, colour approval, the sample rounds that produce the lock — sits outside the ninety days, and it is the part of the schedule the buyer controls and the part no supplier will put a number on, because it depends on how fast you answer.
So when a factory says ninety days, the honest reading is ninety days from a moment that has not happened yet, and the pre-lock phase is the only phase whose length is decided in your office. A programme that takes three weeks to agree a lining colour has spent three weeks of a calendar that had not started, and the quoted lead time is still ninety days, still accurate, and now three weeks later.
The old version of this page said most projects approve at round two and that one round of clear feedback saves three rounds of back-and-forth. Both are withdrawn. We could find no source for either, no study either could have come from, and no factory publishing a round count. The underlying advice — write specific feedback with measurements and codes rather than impressions — survives on its own merits, and we are no longer dressing it in statistics we invented.
The largest calendar lever at this stage is a specification decision, not a negotiation. On the published figures the gap between the two routes is 45 to 60 days against 90 to 120, with an MOQ of 200 to 300 pieces against 500 to 1,000. If the programme can run from a catalogue style with your branding on it, the factory's own numbers say that halves the production calendar and drops the minimum by more than half. That decision is made during sampling, by what you ask the sample to prove: a branding sample against an existing style is a different object from a first article against a new pattern, and the two should not be called by the same name in your own planning.
What this means for you. Define the lock in writing before sampling starts — which documents, which approvals, whose signature — and name one approver with final authority. Then count your own calendar from first enquiry and the factory's from the lock, and never add the two as though they were one number. If your dates are tight, ask what an ODM route would cost you in specification before you ask what an OEM route costs you in weeks.
6. The one published sample round is 5 to 10 days. One Chinese holiday block is nine.
The sampling phase is the part of a programme most exposed to the Chinese holiday calendar, and it is the part least likely to be planned against it. Bulk production gets booked around Spring Festival because everyone knows to. A revision round gets sent whenever the feedback happens to be ready.
| Holiday | Announced block | Days | Make-up working days |
|---|---|---|---|
| New Year | 1–3 January | 3 | 4 January |
| Spring Festival | 15–23 February | 9 | Saturdays 14 and 28 February |
| Qingming | 4–6 April | 3 | — |
| Labour Day | 1–5 May | 5 | 9 May |
| Dragon Boat | 19–21 June | 3 | — |
| Mid-Autumn | 25–27 September | 3 | — |
| Golden Week | 1–7 October | 7 | 20 September and 10 October |
| Total | — | 33 | — |
General Office of the State Council notice Guo Ban Fa Ming Dian [2025] No. 7, issued 4 November 2025, in the English rendering published by Beijing Haidian District Government. The blocks are longer than the statutory cores in State Council Decree No. 795 — Spring Festival four statutory days, Labour Day two, National Day three, the rest one each — because the make-up working weekend mechanism bridges them. The 33-day total is our addition of the seven published blocks. Note that Mid-Autumn and Golden Week are four days apart, which is two stoppages rather than one.
Now put the one published sampling duration against that calendar at the same scale. A round is 5 to 10 days. Spring Festival 2026 was a nine-day announced block. Golden Week was seven. Labour Day was five. A revision round posted into the wrong fortnight does not slow down; it stops, and it resumes at the back of whatever queue the factory returns to.
The stoppage is also wider than the block. Maersk's Chinese New Year 2026 guide, published 4 November 2025, states that factories "typically reduce output 2–3 weeks before the holiday and may not resume full capacity until mid-March", and describes the total disruption window as "up to six weeks" against a 17 February start. That is a carrier's operational guidance rather than a Chinese official statistic — we found no provincial-government or port-authority return-to-work figure for 2026 from any Chinese source, and the shaded band on the chart is our rendering of Maersk's words as dates, not a published range.
Sampling sits at the front of the programme, which is where a lost fortnight is cheapest to absorb and easiest to ignore. A round that lands inside the reduced-output window is not a round that takes ten days instead of seven. It is a round that has not started.
There is no 2027 version of this chart, and anyone showing you one is inventing it. As of 12 August 2026 the State Council has not published the 2027 arrangement. The 2026 notice was issued on 4 November 2025 and the one before it in November 2024, so the 2027 notice should appear around November 2026. What is citable today is the astronomical date from the Hong Kong Observatory's official conversion tables — Lunar New Year 2027 falls on Saturday 6 February and 2028 on Wednesday 26 January — and the four statutory days from Decree 795. Applying the four-day rule to those dates gives a statutory core of 5–8 February 2027 and 25–28 January 2028, which is our inference from two primary sources and not a published schedule. The block around it is not decided yet. If a supplier gives you precise 2027 shutdown dates today, ask which document they are reading.
What this means for you. Put the announced holiday blocks in your sampling plan, not just your production plan, and treat each block as consuming a whole round. Working backwards from the published 2026 calendar, a programme that needs two rounds before Spring Festival needs its artwork frozen well before the reduced-output window opens, not before the holiday itself. And plan the 2027 sampling calendar in November 2026, when there is finally a document to plan against.
7. The variance you inherit at approval is larger than the fee you argued about
Approval is the moment the programme stops being reversible. Everything after it is a shipment moving through a chain whose short steps have far wider spreads than its long ones, and those numbers are public.
| Dwell bucket | Containers | Share |
|---|---|---|
| 0 to 4 days | 29,659 | 75.1% |
| 5 to 8 days | 4,345 | 11.0% |
| 9 days or more | 5,484 | 13.9% |
| Total on terminal | 39,488 | 100% |
Port Optimizer Control Tower "Import Containers by Dwell Time" report, hosted by the Port of Los Angeles. Percentages are the report's own; counts are shown so the shares can be checked. There is no mean in the source and we have not calculated one — a distribution with three quarters of its mass in the first bucket and a seventh beyond nine days has no representative middle. This is one US port on one day and it is not a backpack-specific statistic; it is here because it is the only distribution, rather than an average, that we could source anywhere in the chain.
| Item | Published figure | Source and period |
|---|---|---|
| First pre-production sample | $120–$300 | Osgoodway, 13 May 2026 |
| Major revision re-sample | $80–$150 | Osgoodway, 13 May 2026 |
| Custom mould or hardware | $200–$800 | Osgoodway, 13 May 2026 |
| Ocean schedule reliability | 62.6% on time; late vessels late by 5.31 days; best carrier 77.1%, worst 35.6%, a 41.5-point spread | Sea-Intelligence, June 2026, published 27 July 2026 |
| Customs control rate | From under 1% to over 60% of import declarations, by EU member state of entry | European Court of Auditors, Special Report 04/2021, on 2019 data |
| Terminal dwell tail | 13.9% of boxes at 9 or more days while 75.1% clear within 4 | Port of Los Angeles, 10 August 2026 |
| Inland leg | Rail-destined dwell 5.43 days against truck-destined 2.89 days | Pacific Merchant Shipping Association, June 2026 |
| Berth-time predictability | Standard deviation 33 hours at Los Angeles, 8 hours at the Port of Virginia | US DOT BTS, Port Performance Freight Statistics: 2026 Annual Report, data January–October 2025 |
| Port storage, once free time is gone | Five free days, then £50.05 per day for a 40ft box, then £135.28 per day from day 13 | Port of Felixstowe Rates and Charges, effective 1 April 2026 |
The rows are not commensurable — a price range, a probability, a percentage-point spread, a distribution tail, two durations and a tariff — which is why no honest single figure for "the cost of getting the approval wrong" exists. The dollar and sterling lines are never added together: no exchange rate is sourced or applied anywhere on this page. The 1.9 times ratio between the rail and truck dwell figures is our division of two published numbers.
Read the reliability row and the dwell row together, because they describe the same failure mode from two ends. In June 2026 roughly 37% of vessel arrivals were not on time — our complement of the published 62.6% — and the late ones were late by an average of 5.31 days, a figure that barely moved from February's 5.49 while the on-time rate moved 3.6 points. Meanwhile one import box in seven at Los Angeles sat on the terminal for nine days or more. Neither of those is an average you can add to a plan. They are tails, and a fixed launch date is a bet against a tail.
Now the connection to sampling, which is this article's argument and is ours rather than a finding. Approval is the last cheap decision in the programme. Before it, a change costs $80 to $150 as a major revision and 5 to 10 days as a round. After it, the same disagreement is resolved against a shipment that has already crossed an ocean, cleared a customs authority whose control rate you did not choose, and started accruing storage at a published £50.05 a day. The sample fee is the smallest number anywhere in that sequence and it is the one that gets negotiated.
Two figures worth carrying out of this section. A fortnight of storage at Felixstowe on the published tariff — seven days at £50.05 plus seven at £135.28 — is £1,297.31 per 40ft container, our arithmetic on the port's own rates, and quay rent only, before anything a carrier charges in demurrage on its own box. UK public holidays are excluded from the free-time count, which quietly lengthens free time over Christmas and Easter. Against that, the entire published sampling and tooling budget for a programme with one custom part runs to roughly $850 to $1,100 on the Osgoodway ranges. The two figures are in different currencies and we are not converting them, but they are plainly of the same order — and only one of them buys you a bag.
What this means for you. Spend the sampling stage buying certainty rather than saving fees. Pay for the extra round, pay for the second sample so you and the factory each hold one, and put the resulting reference into the purchase order by identifier. Then choose the port of entry deliberately and ask your forwarder for the carrier's last published monthly reliability rank before the container is stuffed, because those are the two levers left once approval is given.
8. The sample is where the outer-face material stops being a line in a document
One specification decision made during sampling has a duty consequence months later, and it is decided by which fabric you approve.
| Commodity code | Third country duty | DCTS standard preference | Other measures |
|---|---|---|---|
| 4202 92 91 00 | 2.0%, ERGA OMNES | 0%, India excluded, from 19 June 2023 | CITES import control from 1 February 2024; CITES export control from 1 February 2025 |
| 4202 92 11 00 | 8.0%, ERGA OMNES | 3.3%, India excluded, from 19 June 2023 | CITES import control from 1 February 2024 |
Pulled from the Department for Business and Trade's open data API, dataset uk-tariff-2021-01-01, table measures-as-defined, v4.0.1576 published 3 August 2026. The legal base recorded in the data is S.I. 2020/1430, the Customs Tariff (Establishment) (EU Exit) Regulations 2020; both measures are in force from 1 January 2021. Preferences are under the Developing Countries Trading Scheme, S.I. 2023/561; the CITES measures are UK types 710 and 715 under S.I. 2021/54. The equivalent EU rates are not printed here because we could not verify them — see the unverified list.
Both headings carry the same words in the tariff. The split between them is the outer surface material, and it is worth six percentage points of duty on a shipment into Great Britain. That is a materials decision, usually taken on cost per metre and hand-feel, months before anyone opens a customs file.
The sampling stage is the point at which that decision becomes a physical object rather than a line in a specification. A tech pack can say "coated 600D polyester" and leave a classifier guessing about the outer face; a sample can be put in front of a broker. Our reading, flagged as ours: the approved sample is the most useful artefact a buyer has for getting a classification opinion early, and asking for it at that moment costs nothing, whereas asking at entry costs whatever the difference turns out to be.
The CITES rows deserve a line of their own because they are a schedule risk as well as a compliance one. Both codes carry import controls and one carries an export control. They bite only where a trim comes from a listed species — an exotic-skin patch rather than ordinary bovine leather — but a licensed clearance is a different animal from a routine one, and the trim gets chosen during sampling, often as a late aesthetic flourish.
What this means for you. Put the intended commodity code in the tech pack next to the outer fabric specification, and have your broker look at the physical sample rather than the written description before you approve it. If your two candidate materials sit either side of the 2% and 8% split, run that arithmetic while the material is still a choice. Classification remains your broker's call and your liability, not your supplier's.
A worked example, with the arithmetic shown
The point of this is not to give you a number for your programme. It is to show which lines can be sourced and which cannot, so that when a supplier hands you one total you know which half of it was measured and which half was decided. The programme: 1,000 units of a 30-litre custom daypack, OEM against a locked tech pack, textile outer face, one custom zipper-pull mould, two sample rounds including one major revision, FOB Guangzhou, one 40ft container into Felixstowe on CMA CGM's FAL 8.
| Step | Days | Source, or assumption |
|---|---|---|
| Brief, artwork iteration, material selection | not sourced | Yours, and no published figure exists. We are not estimating one |
| Two sampling rounds at 5–10 days each | 10 to 20 | Huafang's 5–10 days per round, published 25 September 2023, multiplied by two. Our multiplication, and it excludes courier time each way and the days the sample sits with you |
| Locked tech pack to FOB shipment, OEM | 90 to 120 | Osgoodway. Assumption: that it holds at 1,000 units, the top of the 500–1,000 band it is published against |
| Yantian to Felixstowe, port to port | 58 | CMA CGM FAL 8 service document, April 2026, via the Cape of Good Hope |
| Origin inland and cut-off | not sourced | Excluded from the carrier's figure by definition |
| UK clearance and delivery | not sourced | No UK control rate was obtained. The EU range of under 1% to over 60% is not a UK figure and is not borrowed here |
| Sourced subtotal | 158 to 198 | Our addition: 10 + 90 + 58 and 20 + 120 + 58. 22.6 to 28.3 weeks, and incomplete at both ends |
| If the vessel is late | +5.31 | Sea-Intelligence June 2026 conditional mean, on roughly 37% of arrivals — our complement of the published 62.6% |
| If the box lands in the dwell tail | +9 or more | 13.9% of Los Angeles import boxes on 10 August 2026. Inference: a Los Angeles distribution used as a shape, not a Felixstowe number |
A note on where the sample rounds sit, because we have treated them two ways on this page. The chart in section 2 folds them inside Osgoodway's 90–120 days, which is the conservative reading. Here they are shown separately, because rounds that produce the lock cannot logically sit inside a window measured from the lock. Both readings are ours; the source states neither, and the difference between them is the 10 to 20 days on the second row. Our lead-times guide takes the first reading — it states the same assumption, that the sample rounds sit inside the 90–120 days — so it adds the two published legs alone and reaches 148 to 178 days for a larger order. Neither figure is a door date, and neither should be quoted as one.
| Line | Low | High | Source, or assumption |
|---|---|---|---|
| 1,000 units at $18–$26 FOB | $18,000 | $26,000 | Osgoodway's published 30L daypack figure, at the top of its published 500–1,000 band |
| First pre-production sample | $120 | $300 | Osgoodway, 13 May 2026 |
| One major revision | $80 | $150 | Osgoodway. Assumption: one major revision, not zero and not three. No published round count exists |
| Custom zipper-pull mould | $650 | Osgoodway's own worked example inside its published $200–$800 range. Not refundable under Huafang's deposit model | |
| Sampling and tooling subtotal | $850 | $1,100 | Our addition of the three lines above |
| Goods and tooling subtotal | $18,850 | $27,100 | Our addition of all four lines |
| UK duty at 2.0%, textile outer face | $377 | $542 | Our arithmetic on the subtotal at the 4202 92 91 00 rate |
| UK duty at 8.0%, plastic-sheeting outer face | $1,508 | $2,168 | Same arithmetic at the 4202 92 11 00 rate. The material decision is worth $1,131 to $1,626 on this one container |
| Felixstowe storage, 14 days beyond free time | £1,297.31 | Our arithmetic, (7 × £50.05) + (7 × £135.28), tariff effective 1 April 2026. Quay rent only | |
Two assumptions inside the duty lines, pulling in opposite directions. We have not sourced the UK customs valuation rule, and customs value is not normally the FOB figure, so the base is understated. Against that, we have left tooling inside the base, which probably overstates it. Neither is corrected because correcting either would require a rule we did not source. No exchange rate is applied and none is implied, so the sterling line is never summed with the dollar ones and there is no grand total. Ocean freight is absent: this page prints no freight rate figure at all, because a published index composite is not a quotation for your box.
Three things fall out of the two tables. The sourced schedule subtotal of 22.6 to 28.3 weeks is 2.8 to 3.5 times the eight weeks this page used to publish, while still missing four steps for which no published figure exists. The entire sampling and tooling budget, $850 to $1,100, is between 4.2% and 4.7% of the goods value — our division — which is to say the decision that consumes most of the buyer's attention costs about one twenty-third of the order. And the outer-face material moves $1,131 to $1,626 of duty, more than the whole sampling budget, on a choice made while looking at the sample.
The fourth thing is the one this article exists to say. A fortnight of quay rent at Felixstowe is £1,297.31 on published rates. We are not converting that and we are not adding it to anything. But it is plainly of the same order as the entire sampling budget above it, and it is what an unresolved disagreement about a shipment looks like while it is being resolved.
What to demand in writing before you approve a sample
This is the part to screenshot. Each row is a term to have in the quotation or the approval email before you say yes. The list is built on this page's own argument — ours, not a sourced finding — that approval is the last moment any of it is cheap and that the sample's job afterwards is evidentiary. Where no published convention exists — and for six of these rows there is none we could find — the row says so, because that is precisely where you have to write the term yourself instead of relying on custom.
| Demand | Why it matters | A non-answer sounds like |
|---|---|---|
| 1. The fee model named: refundable deposit or paid sample | The two are different offers with the same four-word description | "Samples are free" |
| 2. The exclusions from any credit, listed line by line | The one published policy excludes third-party mould and material fees from the refund | Silence, until the invoice |
| 3. Tooling quoted separately, per part, with the trigger for the charge | Published tooling runs $200–$800 against a $120–$300 sample. It is the larger number | Tooling folded into "sample cost" |
| 4. Who owns the mould, and what happens to it if you do not reorder | No published convention found. Write your own or inherit theirs | "It stays here for your next order" |
| 5. A written definition of tech-pack lock, and one named approver | The factory's whole quoted lead time is measured from this event | A lead time with no starting point attached |
| 6. Which object this is: first article against a new pattern, or branding sample against an existing style | The published routes differ by 45–60 days against 90–120, and by MOQ | Both called "the sample" |
| 7. How many rounds sit inside the quoted lead time | No published round count exists for anyone to appeal to | "As many as it takes" |
| 8. What a major revision costs and adds, against a minor one | $80–$150 published for major changes, minor usually free — but the boundary is undefined | "We will look after you" |
| 9. A sealed, signed and dated counterpart sample retained at the factory | Without one there are two recollections and no reference | "We keep photos" |
| 10. The same identifier on both counterparts, and that identifier in the purchase order | A reference nobody can name is not a reference | "The one we sent in March" |
| 11. Dated photographs of the approved sample from stated angles, held by both parties | The cheapest part of the evidence and the first thing skipped | Undated phone pictures in a chat thread |
| 12. Materials identified on the approval record by supplier name and reference, not description | "600D polyester" identifies a category, not a fabric | A generic description repeated back to you |
| 13. Colour recorded against a physical chip retained with the sample | A screen reference is not a colour standard | A code with no chip behind it |
| 14. Which measurements are controlling, and the tolerance on each | No published tolerance convention found. If you do not state one, there is not one | A drawing with no tolerances |
| 15. What happens if bulk deviates from the reference: who inspects, against what, and what remedy | No published convention found for comparing a shipment back to an approved sample | "That has never happened" |
| 16. The holiday blocks the sampling calendar is planned against | A round is 5–10 days; a block can be 9. For 2027 the correct answer names a document that does not exist yet | Precise 2027 shutdown dates, quoted with confidence |
Rows 4, 7, 9, 10, 14 and 15 are the ones with no published practice behind them anywhere we could look — mould ownership, round count, counterpart retention and its identifier, tolerance, and the deviation remedy. That is not a reason to skip them. It is the reason they are the six most valuable rows on the list: a term with no custom behind it is decided entirely by whoever writes it down first.
Row 16 is the cheapest diagnostic on the table, for the same reason it is in our lead-times guide. There is no 2027 arrangement published, so a confident answer tells you how the rest of the numbers in that quotation were produced.
Sampling with Backper, on this page's own terms
The article argues that the fee is the least interesting thing about a sample and the approval record is the most. Applied to our own previously published page, it produces a list we would rather not print.
| Test | Where we stand |
|---|---|
| Our three sample price ladders | Withdrawn. $80–$300 for a pre-production sample, $40–$120 for a stock sample, $50–$150 for a sample box, all with no source. Only the first has any published counterpart, and our floor sat $40 below it |
| The day 0 to day 56 sample-to-shelf calendar | Withdrawn in full. Its finish date was two days earlier than the published Yantian to Felixstowe ocean leg on its own |
| "A 3D mockup tells you 30%, a physical sample the other 70%" | Withdrawn. Invented. There is no measurement behind either figure and there was never a study to cite |
| "Most projects approve at round 2"; "one round of clear feedback saves three rounds" | Both withdrawn. No source, and no published round count exists for anyone |
| The inspection protocol: 30–60 minutes, 10kg for 15 minutes, 20+ zip cycles, 5mm placement, three days of use | The numbers are withdrawn; the practice is not. None of them came from a test standard. Loading a strap and working a zip before you approve is sound. Presenting our own habits as though they were specified thresholds was not |
| Our sample box at "$84 DDP, credited back on the first bulk order" | Self-reported, unaudited, and not restated as a market price. The sentence also failed to say what the credit excludes, which this page now argues is the important half of any such offer |
| Tooling | Quoted separately and named as non-refundable, per part. Self-reported, like every supplier's terms on this subject |
| Counterpart sample retention | We will seal, sign, date and retain a counterpart against the same identifier as yours if you ask. We do not offer it unprompted, and nor does anyone else we could find |
| Where our clock starts | At locked tech pack, defined in writing. The time before that is yours, and it is where most late programmes go late |
| Tolerances on the approval record | We state controlling measurements when asked. There is no published convention we can point you at, ours or anyone's |
| Laboratory testing gates | We cannot give you a turnaround. No dated working-day count for softlines testing was obtainable from any laboratory we could find |
| Commodity code | We state the code we expect and the material driving it. Classification is your broker's call and your liability |
| 2027 holiday dates | We will not give you any. The statutory core is 5–8 February 2027 on our reading of Decree 795 against the Observatory's tables; the block is announced around November 2026 |
Eight of those thirteen rows record something withdrawn, something we cannot do, or something nobody in the trade appears to publish. That is the accurate proportion for this subject, and a supplier page on sampling that reads cleaner than this one is not better informed.
If you are scoping a programme, four answers set most of it: whether the bag can run from a catalogue style or needs a locked tech pack, whether any part needs tooling, which market and port of entry it is landing in, and the outer-face material. Current silhouettes are in the Custom and Branded collection, stock-priced items in bulk deals, and you can ask for a sampling quotation here. Put rows 2, 4 and 15 of the checklist to us when you do — the exclusions, the mould ownership and the deviation remedy. If a supplier cannot answer those three before the sample is cut, the sample will not settle anything later, whatever it looks like when it arrives.
Related reading. What varies in the schedule after approval, and by how much: lead times for wholesale custom backpack production. What the shipment is checked against once it exists: quality control and AQL inspection. What sits inside a quoted unit price: how to read a wholesale backpack quote. The thresholds that decide OEM or ODM: the MOQ guide. The decoration methods a branding sample has to prove: logo printing methods. And upstream of all of it, the supplier checklist.
What we could not verify
- The central argument of this page is not sourced and is presented as an argument. No standard, trade body or manufacturer we could find states that a pre-production sample's primary function is evidentiary, or describes how a bulk shipment is formally compared back to an approved one. The reframing is our reasoning. The figures around it are sourced; the frame is not.
- Every sampling money figure on this page comes from one factory, and every sampling duration from one other. Quanzhou Osgoodway's page of 13 May 2026 and Quanzhou Huafang's of 25 September 2023 are primary for their own terms and are not industry benchmarks. Two published pages are not a sample of a market, and we found no third.
- The Huafang page is nearly three years old at the date of this rewrite. Its two fee models are quoted as published in September 2023. We have no way to confirm the policy is current, and neither does anyone else quoting it.
- No trade-body documentation of the artwork-to-pre-production-sample sequence with per-step durations was found — nothing from any industry association, in any market.
- No published round count exists. Nobody publishes how many sample rounds a typical programme runs, which is why the old page's "most projects approve at round 2" is deleted rather than corrected.
- No published tolerance convention exists for logo placement, dimensions or colour on a bag of this kind. The old page's "within 5mm" was ours and is withdrawn. If your approval record does not state a tolerance, there is no default to fall back on.
- No published convention for sample retention or ownership was found — not for the counterpart sample, not for the mould. Checklist rows 4, 9, 10 and 15 exist because the practice does, but the documentation does not.
- No published laboratory testing turnaround times were obtained. Support-portal pages describing turnaround and rush policies exist, but no dated working-day count for softlines or textile testing. If your programme has a test gate, that step is unquantified here.
- The State Council holiday arrangements for 2027 and 2028 do not exist yet, expected around November 2026 and November 2027. Only the astronomical dates from the Hong Kong Observatory and the statutory four-day core from Decree 795 are citable today, which is why no 2027 block length appears above and why the statutory cores for 2027 and 2028 are labelled as our inference.
- No Chinese official return-to-work statistic was found. The "2–3 weeks before", "mid-March" and "up to six weeks" figures are Maersk's operational guidance, labelled as such wherever they appear. No Guangdong or Zhejiang provincial-government or chamber-of-commerce resumption-of-production figure for 2026 was located.
- Container dwell at Chinese export ports is not published usably. The Ministry of Transport publishes monthly throughput, which is not dwell; a referenced series on vessel time in port and at berth was located but both URLs returned 404. So the dwell distribution on this page is a US one, and it is used as a shape rather than as a figure for your lane.
- The EU third-country duty rates for CN 4202 92 91 and 4202 92 11 are not on this page. The instrument is cited — Regulation (EU) 2025/1926 of 22 September 2025, published in the Official Journal on 31 October 2025 and applying from 1 January 2026, with Chapter 42 of Annex I at about page 344 — but ten retrieval routes failed, including the TARIC consultation service on both hosts, four EUR-Lex renderings, Access2Markets, Poland's ISZTAR4, Finland's Fintaric and three commercial mirrors, none of which printed a percentage either. Subheadings 4202 92 15, 19 and 98 are unverified for the same reason. Read the rate off TARIC before it goes into a landed-cost model.
- No current EU-wide customs inspection rate was obtained, and no UK control rate at all. The only sourced figure is the European Court of Auditors' range of under 1% to over 60% by member state, on 2019 data published in 2021 — five years older than this article. The worked example therefore leaves UK clearance unquantified rather than borrowing the EU range.
- We did not source the UK customs valuation rule, so the duty lines in the worked example are calculated on the FOB goods and tooling figure and flagged in the table as understating the base in one direction and overstating it in another.
- No Felixstowe dwell statistic exists that we could find. The port publishes a priced storage tariff, which is what we used. World Bank Container Port Performance Index 2025 per-port ranks could not be retrieved, the Port of Rotterdam dashboard served placeholder values on 11 August 2026, and per-port monthly berth-time tables from US DOT BTS were not obtained — only the 33-hour Los Angeles against 8-hour Virginia comparison quoted in the report's own text.
- No absolute published port-to-port transit time for Maersk or Gemini Transpacific services was found, and COSCO's could not be retrieved, so this page contains no Transpacific transit figure. The only ocean figures here are CMA CGM's FAL 8 numbers for Asia to North Europe.
- No exchange rate is sourced anywhere on this page, which is why the sterling and dollar figures are never summed. Carrier demurrage and detention are also absent: the Felixstowe figures are quay rent only.
- Our own terms are self-reported. Everything in the Backper table is what we know about ourselves today, and every line can be evidenced on request.
Sources
Manufacturers' published terms. Quanzhou Osgoodway Co., Ltd, OEM vs ODM Backpack Manufacturing Explained, published 13 May 2026, modified 15 May 2026 — first pre-production sample $120–$300, major-revision re-sample $80–$150, custom hardware or EVA mould $200–$800 with a $650 zipper-pull example, OEM 90–120 days from locked tech pack, ODM 45–60 days from confirmed style selection, unit cost $18–$26 FOB Guangzhou at 500–1,000 units. Quanzhou Huafang Bags Co., Ltd, article on sample costs, published 25 September 2023 — sampling 5–10 days, and the refundable-deposit and paid-sample models with the third-party mould and material exclusion.
Chinese holiday calendar. State Council Decree No. 795, published 12 November 2024, in force 1 January 2025. General Office notice Guo Ban Fa Ming Dian [2025] No. 7 of 4 November 2025, in the English rendering published by Beijing Haidian District Government. Hong Kong Observatory Gregorian–Lunar Calendar Conversion Tables, 2027 and 2028, also released on DATA.GOV.HK. Maersk, Chinese New Year 2026 Supply Chain Prep Guide, 4 November 2025.
Carrier services and schedule reliability. CMA CGM, FAL 8 service document, April 2026. Ocean Network Express, FE1 and FE3 service update via the Cape of Good Hope, 28 May 2026. Sea-Intelligence, Global Liner Performance, press releases 379 (February 2026) and 400 (June 2026, published 27 July 2026).
Ports, dwell and terminal charges. Port Optimizer Control Tower import container dwell report, Port of Los Angeles, snapshot 10 August 2026. Pacific Merchant Shipping Association, San Pedro Bay dwell times for June 2026, 21 July 2026. Port of Felixstowe Rates and Charges, effective 1 April 2026. US DOT Bureau of Transportation Statistics, Port Performance Freight Statistics: 2026 Annual Report, January 2026, data January–October 2025.
Customs and tariff. UK Department for Business and Trade Data API, dataset uk-tariff-2021-01-01, table measures-as-defined, v4.0.1576 published 3 August 2026. S.I. 2020/1430, S.I. 2023/561, S.I. 2021/54. GOV.UK reference document for the Customs Tariff (Establishment) (EU Exit) Regulations 2020, v1.33 in force 10 July 2026. European Court of Auditors, Special Report 04/2021, on 2019 data. Regulation (EU) 2025/1926, OJ 31 October 2025, and the TARIC consultation service, for the EU rates this page does not print.
Rewritten 12 August 2026. This page previously priced three kinds of sample and timed a seven-step calendar with no source behind a single figure, ended that calendar two days before the published ocean crossing would have finished, stated that a mockup tells you 30% and a sample the other 70%, asserted a typical round count and a feedback saving that no published data supports, and presented a set of inspection thresholds as though they came from a standard. It also offered our own sample credit without saying what the credit excludes. Corrections have been made in the open rather than quietly deleted. The 2027 Chinese holiday arrangement is expected around November 2026, the EU duty line still has to be read off TARIC, and both factory pages quoted here are one company's terms rather than a market — re-check anything on this page before it goes into a contract, an approval record or a landed-cost model.
Tags:
Previous
Best Backpack Gift Ideas for Travelers and Hikers: The Return-Risk Data Nobody Else Checks
Next
Custom Corporate Gift Backpacks with Embroidery: What You May Actually Spend, and Who Pays Tax on the Rest
