Home / Journal / Inspection booking
Pre-Shipment Inspection Booking: A Pre-Arrival Checklist
Booking a pre-shipment inspection is a scheduling exercise with a quality decision attached: you are promising a third party that on a specific morning there will be a defined lot, finished, packed, and ready to be sampled — and if that promise is wrong, the inspection does not fail the goods, it fails the booking.
Most inspection problems that get blamed on the factory are actually produced in the fourteen days before the inspector arrives. A PO that never defined the lot. A booking made against a completion date that was a hope. Twelve cartons of the previous colourway still on the floor because the line changed over early. A packing list typed from the order rather than counted from the pallet. None of these are quality defects. All of them end in a report that says the goods could not be inspected, or in a failed result that costs a week to unwind.
This article is about the booking half of the job — the part that happens before anyone measures anything. It is written for buyers who commission inspections and for factories that host them, because the two sides consistently prepare for different events.
1. Three calendars, one date · 2. Define the lot before you book · 3. The booking arithmetic · 4. Who books and who pays twice · 5. Put the sampling plan in the PO · 6. The document pack · 7. The 72-hour self-check · 8. Findings that were visible a week earlier · 9. Reading the report · 10. What we do before we invite anyone · 11. Questions
1. Three calendars, one date
An inspection date is the intersection of three independent calendars, and buyers usually see only one of them.
| Calendar | Controlled by | What usually slips |
|---|---|---|
| Production completion — goods finished, packed, palletised, carton marks applied | The factory | Decoration, hand-finishing and packing are the last steps and the first to compress; "99% finished" is not a lot |
| Inspector availability — a booked man-day, plus travel to the site | The inspection company | Peak-season slots disappear two weeks out; a factory in an industrial park outside the city centre costs half a day in travel |
| Shipment — cut-off, container stuffing, port congestion | The forwarder and the carrier | Cut-offs move earlier, never later |
The date that matters is not "when can someone come". It is the latest date at which an inspection can still produce a decision you can act on. If the report arrives after the container is stuffed, you have bought a document, not a control. Work backwards from the cut-off, not forwards from the completion date.
There is a fourth calendar that is easy to forget and expensive to miss: the rework calendar. If the goods fail, someone has to open cartons, repair or replace, re-pack, and re-present the lot. Whatever buffer you keep between inspection and cut-off is really the size of the rework you are willing to fund.
2. Define the lot before you book
A sampling plan is only valid against a defined lot. This is the single most common structural mistake in inspection bookings, and it is made in the purchase order rather than on the factory floor.
A lot is not "the order". A lot is a quantity of one product, produced under essentially uniform conditions, presented for inspection as one unit. If a booking covers four colourways of the same doll, three sizes of the same basket and a mixed carton of samples, the inspector is not sampling one lot — and any statistical statement in the report becomes decorative.
Three definitions have to be written down before the booking, because they cannot be agreed on the morning of the inspection without becoming a negotiation:
- What counts as one unit. For a crochet doll it is the finished piece with its attached parts, its hangtag and its inner bag — or it is the bare piece, if that is what the buyer inspects. Both are defensible; only one can be in the plan.
- Whether partial lots are acceptable. Inspectors are usually asked to work on finished and packed goods. A lot that is 90% packed is not 90% inspectable, because the sample has to be drawn from the population, not from the part that is ready.
- Where the tail sits. The last cartons of a run are statistically the most interesting and operationally the least convenient. If they are excluded from the lot, say so; if they are included, they must be finished before the inspector arrives.
The tail carton deserves its own sentence because it is where most "the lot was fine but the shipment was not" stories begin. A run of 1,000 pieces where the first 950 are perfect and the last 50 were made on Friday afternoon is not a 95% good lot. It is a lot with a different production history at the end, and random sampling is exactly the instrument designed to notice that — provided the tail is in the population being sampled.
3. The booking arithmetic
The booking date is not a preference. It is a sum, and every term in it is knowable in advance:
| Term | What it covers | Typical shape |
|---|---|---|
| Completion | Last piece finished, packed, marks applied, palletised and counted | The date the factory will actually hit, not the date in the plan |
| Presentation buffer | Fixing what the self-check finds: re-packing cartons, reprinting marks, pulling mixed stock | Two to three working days is realistic for a first order with a new factory |
| Booking lead time | The inspection company's scheduling window and inspector travel | Booked slots are usually required several working days ahead; the requirement is set by the company's current rule, not by the factory |
| Report turnaround | Report issued after the visit | Often one to two working days for a standard inspection |
| Decision buffer | Time to choose between release, rework and re-inspection before the cut-off | This is the term that gets set to zero and should never be |
Add them and compare to the cut-off. If the sum does not fit, the correct response is to move the booking earlier and hold the lot, not to compress the buffer — because a compressed buffer does not remove the risk, it converts a quality problem into a shipping problem, which is more expensive and less negotiable.
For a first order with a new supplier, we plan seven working days between "goods packed" and "container cut-off" as a working assumption, and treat anything less as a deliberate risk. That number is our own planning rule, not an industry standard; what matters is that it is written down before the order starts, because on the day it will feel excessive to everyone involved.
4. Who books and who pays twice
Three commercial questions get settled in the same email as the booking, and they are worth settling there rather than after a failure:
- Who appoints the inspector. The buyer appoints when the inspection is a condition of payment; the factory appoints when it is a service the buyer has asked for and does not want to administer. Both are normal. What is not normal is appointing nobody and assuming the forwarder will notice.
- Who pays for the first inspection. Market practice is that the buyer pays for the first visit as part of the cost of buying, and the seller pays for any re-inspection caused by a failed result. This alignment is the reason the clause is worth writing: it makes the factory's cheapest path to profit identical to the buyer's cheapest path to a clean shipment.
- What "failed" means for the balance payment. If a failed inspection has no commercial consequence, it is a survey. Tie it to something concrete — release of the balance, permission to ship, or the seller's cost of re-inspection — or the report becomes advisory.
The clause that causes the most argument later is not the fee. It is the definition of "failed". A lot that fails on one major defect class and passes the others is not a failed lot in every contract, and the difference between "rework the affected units" and "produce the lot again" is usually a sentence somebody meant to write in the PO and did not.
5. Put the sampling plan in the PO, not in the email
AQL is a contractual acceptance limit, not a legal threshold and not a percentage of the shipment. Two consequences follow, and both are routinely missed.
First, the plan has to name the standard, the inspection level, the sampling type and the AQL values per defect class — otherwise the inspector applies a default and the result is measured against a plan nobody signed. Second, the plan has to name which characteristics are critical, major and minor. That classification decides the outcome far more often than the sample size does, and it is a commercial decision about what a buyer can live with, not a technical one.
For orientation, these are the sample sizes at general inspection level II with normal, single sampling, which is the default almost everywhere in consumer goods. Acceptance and rejection numbers are shown for the two AQL values most often used for major and minor defects.
| Lot size | Code | Sample | AQL 2.5 (Ac/Re) | AQL 4.0 (Ac/Re) |
|---|---|---|---|---|
| 151 – 280 | G | 32 | 2 / 3 | 3 / 4 |
| 281 – 500 | H | 50 | 3 / 4 | 5 / 6 |
| 501 – 1,200 | J | 80 | 5 / 6 | 7 / 8 |
| 1,201 – 3,200 | K | 125 | 7 / 8 | 10 / 11 |
| 3,201 – 10,000 | L | 200 | 10 / 11 | 14 / 15 |
| 10,001 – 35,000 | M | 315 | 14 / 15 | 21 / 22 |
Values follow the ISO 2859-1 series, which GB/T 2828.1 adopts identically in China and ANSI/ASQ Z1.4 mirrors in the United States. The standard itself is a paid document and the tables run further than this; the point of reproducing six rows is not to replace it but to make one thing obvious: sample size is a function of lot size, not of how nervous the buyer is. A 1,000-piece lot at level II is 80 pieces whether the order is worth ten thousand dollars or a hundred thousand.
Two further terms belong in the same paragraph of the PO, because they change the numbers: severity (normal, tightened, reduced) and whether critical defects are on a zero-acceptance basis. Zero acceptance for critical defects is common in toys and is not a statistical statement at all — it is a decision that one occurrence is enough.
6. The document pack that has to be on the table
An inspector who arrives to find the paperwork incomplete has three options: wait, inspect with what is available, or write up the gap. None of them is good, and the third one is the most expensive, because a report with an "information not available" line is a report a buyer cannot release against.
The pack below is what a third-party inspector will ask for on a crochet or soft-goods lot. It is short enough to assemble in an afternoon and long enough that almost nobody has it ready on the first order.
- Purchase order and amendments, including the sampling plan, the AQL values and the defect classification.
- Approved and sealed reference sample, present physically, with its measurement sheet — not a photograph of it.
- Packing list that matches reality, counted from the pallet rather than typed from the order. Carton count, units per carton, gross and net weight, and the carton-mark specification.
- Colour reference: the approved lab dip, yarn card or Pantone reference, plus the record of which dye lot went into this run.
- Test reports for the applicable markets, with the report number, the model or SKU they cover and a validity date that has not passed.
- In-process records that the factory already keeps: gauge checks, pull-force results on attached parts, in-line measurement samples.
The most frequently missing item is not the test report. It is the sealed sample — because the sample was approved by email and never sealed, or was sealed and then borrowed for a trade fair. An inspector who cannot compare against an approved reference can only report that the goods are internally consistent, which is a much weaker statement than the buyer thinks they have bought.
7. The 72-hour self-check
Everything below can be checked by the factory without a standard, a laboratory or an inspector. It is the list we run before inviting anyone, and it exists because each item has, at some point, turned a bookable lot into a wasted man-day.
| # | Check | Why it fails on the day |
|---|---|---|
| 1 | Lot is finished, packed, palletised and counted | Packing is the last step and the easiest to leave at 95% |
| 2 | Carton count on the floor equals the packing list | The list was typed from the order, not counted from the pallet |
| 3 | Carton marks match the approved artwork, all four sides | An old print file was used, or the side panel was skipped |
| 4 | No stock from a previous colourway or PO in the inspection area | Changeover happened early and the residue was never removed |
| 5 | Every carton sealed and strapped to the agreed pattern | The tail cartons were packed in a hurry and left open |
| 6 | Sealed reference sample and its measurement sheet on site | The sample went to a fair, or lives in the buyer's office |
| 7 | Colour record: which dye lot, and a retained hank from it | Nobody wrote down the lot, so a shade question has no answer |
| 8 | Test reports printed and matched to the SKU being inspected | The report covers a different size or a withdrawn model |
| 9 | A clear, lit inspection area with tables and a power point | The only free space is the loading bay, in use that morning |
| 10 | A named person who can answer questions and sign | The merchandiser is travelling and the line leader cannot sign |
| 11 | Access arranged: gate pass, PPE, safety briefing | The inspector waits forty minutes at the gate |
| 12 | Opening cartons identified in advance | Nobody knows which cartons may be cut, so nothing is cut |
Item 12 is the one factories underestimate. An inspector who is not allowed to open cartons can only inspect what is already open, and a sample drawn from open cartons is not a random sample. Agreeing which cartons may be opened — normally spares or a nominated pallet — costs nothing and removes an entire class of argument.
8. Findings that were visible a week earlier
Across a few hundred lots, the same five findings account for most failed first inspections on handcrafted goods. Every one of them is detectable without an inspector.
- Mixed lots. Two colourways in one lot, or a previous PO's balance left in the corner. The sampling plan then describes a population that does not exist.
- Carton marks that do not match the artwork. A shipping mark is a document; a wrong mark means the consignment does not match the documents presented to customs or to the retailer's DC, whatever the goods are like inside.
- Measurement drift on the tail. Hand-crocheted pieces drift when the maker changes, when the yarn lot changes, or when the pace changes near the end of a run. A gauge and dimension check across the first, middle and last fifties catches it; a check on the first fifty does not.
- Colour without a retained reference. If the only record of the approved shade is a photo on somebody's phone, a minor shade difference becomes an unwinnable argument in both directions.
- Attached-part security treated as cosmetic. Eyes, noses, ribbons and small accessories on a toy are a safety characteristic, and the test is a pull test with a recorded value, not a visual tug. Where the record does not exist, the inspector has to test destructively on the sample, and the report says so.
The pattern is consistent: the finding is rarely a surprise to the people on the floor. It is a surprise to the paperwork. That is precisely why the self-check is worth more than an extra inspector-day — it converts a known deviation into a fixed one before anyone is paid to write it down.
9. Reading the report
An inspection produces one of three outcomes, and only one of them is an ending.
| Outcome | What it means | What to do next |
|---|---|---|
| Pass | The sample met the acceptance numbers for the classes inspected | Release against the plan, and record what was not covered: characteristics outside the plan, cartons not sampled, tests not performed on site |
| Fail — reworkable | Defects are concentrated and fixable within the lot | Agree the rework scope in writing, fix, re-present the whole lot, and book the re-inspection before the rework starts |
| Fail — systemic | The defect runs through the production method, not through a few units | Stop. Re-inspection of the same lot will reproduce the same result, and the decision is now about remaking, substituting or renegotiating the date |
Two things a pass does not mean, both of which get assumed. It does not mean every unit is good: a sampling plan accepts a defined risk, and the accepted number of defective units is inside the cartons you are about to ship. And it does not mean the goods conform to the specification — it means the sample conformed to the characteristics that were checked, against the limits that were agreed. Where the plan was thin, the pass is thin.
Which is the argument for spending the effort on the plan rather than on the number of inspections. Two well-defined inspections on a properly prepared lot will tell you more than five visits to a lot nobody defined.
10. What we do before we invite anyone
Our own sequence, in the order it happens: the lot is closed and counted; the carton marks are checked against the approved artwork on all four sides; the sealed reference sample and its measurement sheet are brought to the inspection area; gauge and dimension checks are run on the first, middle and last fifties; pull-force values on attached parts are pulled from the in-process record rather than re-tested; and the packing list is recounted from the pallet the afternoon before the visit.
None of that is inspection. It is preparation, and it is the part that decides whether the man-day buys a decision or a description. If you are booking an inspection for a crochet or soft-toy lot and want the self-check as a one-page sheet, send us the lot structure and the destination market and we will return the version we use, with the document pack listed against your order.
Related reading: the twelve checks in a factory audit, which covers the capability question rather than the lot question; progress reporting, which is how the completion date in the booking gets to be trustworthy; and sealed samples, without which the inspector has nothing to compare against.
How far in advance should a pre-shipment inspection be booked?
Work backwards from the shipping cut-off rather than forwards from the completion date: presentation buffer, the inspection company's booking lead time, and one to two working days for the report. For a first order with a new factory we plan seven working days between goods packed and cut-off. Booking lead times are set by each inspection company's current rules, so confirm the window when you book rather than assuming last year's.
Can an inspection be booked before production is finished?
It can be booked, but it cannot be performed on an unfinished lot. Inspectors sample finished, packed goods because the sample has to be drawn from the population being shipped. A lot that is ninety per cent packed is not ninety per cent inspectable — the un-packed tenth is usually the tail, which is the part most worth examining.
What is the difference between a factory audit and a pre-shipment inspection?
An audit asks whether the factory can make the product, and looks at systems, records and capability over time. A pre-shipment inspection asks whether this lot is acceptable, and looks at the goods in front of the inspector on one day. A factory can pass an audit and still present a failing lot, and a good lot can come out of a factory that would fail an audit. They answer different questions and neither substitutes for the other.
Does a passed AQL inspection mean the shipment is defect-free?
No. A sampling plan accepts a defined number of defective units as consistent with the agreed quality limit, so a passed lot can still contain defects inside the cartons. AQL is also not a percentage of the shipment: it is the acceptance limit applied to the sample, with acceptance and rejection numbers taken from the plan. Reading it as "2.5% defects allowed" is the most common misreading in the trade.
Who pays for a re-inspection?
Market practice is that the buyer pays for the first inspection and the seller pays for any re-inspection caused by a failed result. That allocation is worth writing into the purchase order, because it aligns the incentives: the cheapest route to profit for the factory becomes the cheapest route to a clean shipment for the buyer. Where the clause is silent, the cost is negotiated after the failure, which is the worst moment to negotiate anything.
Can a video call replace an on-site inspection?
For a repeat order with an established supplier, a structured video walk-through of the packed lot, the carton marks and the retained sample catches a useful share of problems and costs nothing. It cannot replace sampling: the camera shows what the factory chooses to point it at, and it cannot draw a random sample or measure a pull force. Use it as a pre-check before booking, not as the inspection.
What happens if the lot fails and the vessel cannot wait?
The decision moves from quality to logistics, and it should be made explicitly rather than by default. The realistic options are to ship the conforming portion and hold the rest, to accept the lot with a price adjustment documented in writing, or to rebook the shipment. What should not happen is shipping a failed lot while the rework is still being discussed, because the inspection then has no commercial effect and the next one will not either.
The packing list is the one document an inspector can test against the cartons in front of them. The document pack that travels explains why a list typed from the order rather than counted from the pallet is the failure that survives every review.
Related: carton shipping marks and numbering — the four fields, the denominator that has to be closed after the count, and the version code that decides whether a carton is accepted at goods-in.
Related: yarn dye lots and lot-change control — why a tolerance figure without an equation is not a measurement, the four places lots break, and the written notice that turns a complaint into a decision.
Want the pre-arrival self-check as a one-page sheet? Send the lot structure, the destination market and the inspection company, and we will return the version we run, with the document pack mapped to your order.