Common tech-pack mistakes that cost you time and money with factories

"The factory messed it up" is the most popular verdict in brand-owner chats. Sometimes it's true. But dig into the specific cases, and in most of them the factory honestly sewed exactly what was in the brief. The brief was just wrong, incomplete, or floating around in three different versions at once.

The seven mistakes below are the seven most expensive gaps in a spec. For each one: what it looks like, how it ends, and how to close it.

Mistake 1. Measurements with no tolerances

What it looks like: there's a size chart, but no "±" column. How it ends: the batch arrives and the body length drifts by 3 cm from unit to unit. Defective? In your gut, yes. Legally, no, because no criterion was ever set. A refund dispute turns into an argument about taste. Real example: in a 600-piece hoodie run, the back length ranged from 67 to 71 cm. The factory never touched the pattern; two different operators on two lines simply set their overlock machines differently. With no tolerance to point to, there was nothing to claim, and it went to clearance at a discount. How to close it: a tolerance on every point of measure (typically ±0.5–1.5 cm depending on the point) and a diagram showing exactly how to take each one. Then inspection becomes a tape measure and a table, not emotions. More in the breakdown of tech-pack structure.

Mistake 2. Material "like the sample," with no article number

What it looks like: "fleece like last batch," "same fabric as the sample we sent you." How it ends: on the reorder, the factory's supplier no longer has "that exact" fabric and grabs something similar. Weight 320 → 270 gsm, composition 80/20 → 65/35. The product page promises one thing, the reviews record another. Real example: the second run of t-shirts arrived noticeably thinner than the first. The fabric supplier switched machines, and the factory took an "equivalent under the same name." Customers wrote "see-through" in reviews, and the listing rating dropped from 4.8 to 4.3 in two weeks. How to close it: in the BOM, the exact material article number, the supplier, the composition, and the weight as a number. Substitutions only with your written sign-off and recorded in a new version of the tech pack.

Mistake 3. Colour described in words

What it looks like: "dark grey," "off-white," "like in the photo" (a photo shot on a phone under a yellow lamp). How it ends: a colour-block of two "identical off-whites" from different dye lots; returns saying "the colour isn't like the picture." Real example: a suit was cut from two fabric purchases a month apart. The "off-white" in the second batch came out with a pink cast. On the finished garment, the top and bottom side by side looked like two different shades, and half the run went out as a markdown. How to close it: a Pantone TCX reference for each colour in the BOM, plus a physical standard (a swatch) signed by both sides for critical items. Cheap, and it kills the "that's not the right grey" dispute.

Mistake 4. Edits in chats on top of the sent tech pack

What it looks like: the tech pack is sent, then in chat: "let's make the length 68," "actually, the hood without a drawcord." A month later nobody, including you, can reconstruct the final configuration from the message thread. How it ends: the factory sews to the last file, not the last message. Or the other way around, to a message you've already changed your mind about. Real example: the chat asked to drop the side pockets, but the tech-pack file was never resent. Cutting went off the old pattern, and 500 jackets arrived with pockets. The technologist calmly pulled up the thread: the document said one thing, the chat said another, and they sewed to the document. How to close it: an ironclad rule, "a change = a new numbered version of the tech pack." Chat is for discussion, the document is for decisions. Old versions are marked void. Boring right up until the first 500-piece batch with the "wrong" length.

Mistake 5. Skipping the sampling cycle

What it looks like: "we're out of time, let's run the batch straight away and review the sample from a photo." How it ends: a fit you can't judge from a photograph, you end up judging across the whole batch. The true cost of that "saving" is the entire run going to dead stock. Real example: a dress went into production off a photo of the sample. On a real body it turned out the armhole cut in and the sleeve pulled when the arm was raised. 800 pieces sewn, the fit no longer fixable, only a markdown. How to close it: at least two rounds. The first sample (proto) checks construction and fit on a real person or a dress form. The second (pre-production, made from bulk materials) is the standard for the batch. The approved sample is signed and kept: at inspection you compare the batch against it, not against memory. Yes, that's plus 2–4 weeks. They're cheaper than any dead stock.

Mistake 6. Size grading "by eye"

What it looks like: the chart was built by adding "2 cm per size" evenly across every point, with no base size and no grading rules. How it ends: the base M fits perfectly, but the XL is "somehow off": proportions don't scale linearly. Reviews say "sizing is wrong," and returns rise specifically on the extreme sizes. Real example: trousers were graded with an equal step across all points. The M fit was perfect, but on the largest size the inseam came out short and the waistband narrow. Returns on the extreme sizes reached 40%, while the base size was under 10%. How to close it: grade from a base size using rules for your garment type (steps for chest, waist, and lengths all differ), and try on at least the extremes of the range, not just the base.

Mistake 7. No agreed process for statuses and inspection

What it looks like: "the factory's quiet, they must be sewing." You learn the status by phone, and inspection is "opened the boxes, looks fine." How it ends: you find out about a delay three days before the planned delivery window; defects surface only after the goods have already shipped to customers. Real example: a fabric slip surfaced a week before the delivery slot. The factory had been silent for three weeks, assuming "everything's under control." The slot was lost, a new one came 11 days later, and sales had already slumped at the start of the season. How to close it: agree control points with the factory (material purchase → cutting → sewing → QC → packing → shipping) and a reporting format for each. Inspection runs off a checklist with an AQL sample, a measurement check against tolerances, and a comparison to the approved sample.

The common denominator

Look at the list again: none of the seven mistakes are about sewing. They're about managing information: what's recorded, in which version, and who can see the current state. The factory executes a document. The quality of that document is your job.

The minimum kit for your next order: 1. A tech pack to the full structure, with tolerances, article numbers, and Pantone (template here). 2. A versioning rule: a change = a new number. 3. Two sampling rounds, the standard sample signed. 4. Production control points and an inspection checklist.

From PLMflow. We've built this process into a system: a tech pack with tolerances and versions, a portal where the factory always sees the current version and stage statuses, and a change history that writes itself. "Sewn from the wrong file" simply can't happen in that setup. See how it works →