How to Iterate on Samples Efficiently (and Stop Wasting Months)

Sampling rarely takes months because the factory is slow. It takes months because the buyer iterates badly. Feedback arrives in drips. Every round changes ten things at once, so nobody can tell which change caused which result. And nobody wrote down what "approved" means, so round four gets argued on vibes.

The difference between a program that closes in eight weeks and one that limps past six months is usually a protocol. Disciplined teams close sample approval in two or three rounds; teams without one often stretch past six, according to one OEM factory's engineering team that has watched the pattern repeat across categories. This guide gives you that protocol: acceptance criteria before round one, feedback a factory can act on, what counts as a round, fair rules on sample charges, and when to walk away.

If this is your first program with a Chinese factory and you have nobody on the ground to hold the process together, a sourcing agent like CN Ally can review samples against your criteria and keep the iteration loop tight while you focus on the product.

Why sampling actually takes so long

The delay is rarely one big event. It is five small failure modes stacking up, round after round.

Vague feedback. "Make it look more premium." "The color feels off." A factory cannot manufacture a feeling. Every subjective comment triggers an interpretation round: the factory guesses what you meant, builds to the guess, and you reject the guess. Each cycle costs a week or more.

Bundled changes. When a sample needs work, the instinct is to fix everything at once: the color, the handle, the lid fit, the logo placement. But changing every variable in one round makes results impossible to attribute, so the next round starts from confusion instead of information.

No acceptance criteria. If "approved" was never defined in measurable terms, approval becomes a negotiation. The factory thinks the sample passes; you disagree. Both sides argue from memory instead of from a document.

Drip feedback. You send comments on Monday, the factory starts the revision, and on Thursday you remember three more things. The factory restarts. One round just became two, and neither side logged it.

A weak brief. Most iteration pain is brief pain surfacing late. A loose specification turns every round into a re-briefing session. Tighten the front end and the back end gets shorter, which is why our guide to writing a factory-ready design brief exists.

Fix these five and sampling stops being a black hole. Leave them and no factory on earth is fast enough to save your timeline.

How many sample rounds are normal?

Two to three rounds is the healthy range for most programs. One to two for simple adaptations, three to four for genuinely custom products. Past that, something in your process is broken, not something in the product.

Product type · Typical rounds · What drives the count

• Logo or color change on an existing factory design: 1–2 · Artwork approval, color matching
• Moderate customization of an existing product: 2–3 · Fit, materials, decoration
• Full custom OEM with new tooling: 3–4 · Engineering validation, DFM corrections
• Apparel with fit approvals: 3–5 · Fit samples, size grading, fabric behavior

These are planning bands, not guarantees. A bag manufacturer publishing its real process numbers reports first samples in 6–10 working days with typically 1–2 revision rounds at 5–7 days each. The pattern holds across categories.

Treat round four as a diagnostic event, not a disaster. Ask which of three things is true: are we changing too many things per round, is the feedback still subjective, or is the factory simply not capable of this product? One of those is the answer almost every time.

The iteration protocol that closes in three rounds

Each step below is small. Together they are the difference between three rounds and six.

Step 1: Write acceptance criteria before round one ships. Before the factory cuts anything, define "pass" in measurable terms. Colors as Pantone codes or physical chips, not "our blue." Dimensions with tolerances per feature, not just overall size. Function tests with a defined method: how many open-close cycles a clasp must survive, what weight a handle must hold. Agreeing on criteria before sampling starts removes the most common delay of all: the subjective argument at the approval stage.

Step 2: Review against the criteria, not your memory. When a sample arrives, open the criteria document first and check each line. Left alone, the brain compares the new sample to an imagined ideal and finds fresh things to dislike every time. The document anchors the review to what was actually agreed.

Step 3: Send one consolidated feedback document per round. Every change in one document, each numbered, each tied to a photo and a measurement. No follow-up chat messages with "one more thing" after it goes out. Late thoughts wait for the next round.

Step 4: Limit the variables per round. Change a defined, limited set of things each round so both sides can attribute results. A practical sequence: appearance first (colors, decoration, finishes), then function (fit, mechanisms, performance), then consistency (does the second unit match the first). Controlled iteration feels slower than the change-everything instinct, but every round produces usable information instead of noise.

Step 5: Number every version and keep a log. Version 1, 2, 3, each with a date, the changes requested, and the review outcome. When the criteria are met, sign off in writing and move to the pre-production sample. The log doubles as evidence if a dispute ever arises about what was approved.

What actually counts as a "round"?

A round is one complete revision cycle: you send consolidated feedback, the factory remakes the sample, it ships, you review it. Four steps, one round.

Things that are not rounds but still burn calendar time: clarification photos the factory requests, questions about your feedback, a video call to talk through a detail. They don't reset the count, but they explain why "three rounds" can still take two months when the feedback was unclear enough to need three clarification calls per round.

Factories count rounds by remakes, and their patience and pricing follow that count. If you count every message exchange as a round, you'll think you're at round eight while the factory thinks you're at round three, and you'll be frustrated about different things. Align on the definition early. One sentence in your first feedback document does it: "We count each physical revision as one round."

One more rule that saves entire rounds: approve artwork before the sample is made, not during sample review. If decoration is still being designed while the sample is cut, the sample tests nothing and the round is wasted.

How to write feedback a factory can actually use

The feedback document is the highest-leverage skill in this process. Factories rarely ignore good feedback; they simply cannot act on bad feedback. Here is what good looks like.

Annotated photos, not paragraphs. Photograph the sample, circle the problem in red, add an arrow and a number. "Photo 1, item 3: stitching is 2 mm from the edge; spec says 5 mm. Move to 5 mm." A worker can act on that in seconds. Nobody can act on "the stitching looks a bit off."

One document per round. A PDF or a single shared doc with all photos, all numbered changes, and the criteria checklist. Not twelve chat messages, not a voice note, not a call nobody minuted. If it isn't in the document, it wasn't requested.

Every change tied to a measurement or a reference. "Darker" means nothing. "Match Pantone 186 C; the current sample reads closer to 1788 C" means everything. Without a measurement, send a physical reference, a color chip or a competitor product, and photograph the sample next to it.

Separate "must fix" from "try." Factories prioritize what you emphasize. Mark each item: blocking (approval depends on it), preferred (fix if possible without retooling), or exploratory (only if it's free this round). This keeps one round from ballooning into a redesign.

Set a response deadline and a shipping method. "Please confirm receipt within two working days and ship the revised sample by [date] via express." Calendar discipline is part of feedback. Without it, your perfect document sits in a queue.

Who pays for samples, and when to push back

Factories charge for custom samples, and rightly so. A sample consumes setup time, materials, and skilled labor for a product that may never reach production. The common structure: the buyer pays per sample round, and reputable suppliers often refund or credit those fees against the bulk order once production is placed. Get the refund terms in writing before round one, not during the production deposit negotiation.

The nuance is which rounds deserve full price:

• Round one and genuine design changes: pay. This is development work, and it's yours.
• Revisions fixing the factory's own errors: push back. You specified Pantone 186 C and the sample arrived in the wrong red. That remake is on them. Cite the numbered feedback item they missed.
• Repeated failures on the same item: renegotiate. If round three is still fixing a defect the factory acknowledged in round one, you're funding their learning curve. Say so politely, with the version log as evidence.

Watch for a pattern, not just a price. A factory that charges full price for every round including its own mistakes is showing you how it will handle defects during production. For the full economics, see our breakdown of sample costs from China and the overview of how to source samples.

When to kill the project instead of ordering another round

Most guides skip this section. It saves the most money. Every extra round feels cheap, a few hundred dollars and two more weeks. But failing programs have a recognizable pattern, and spotting it early is a skill worth building.

Kill or pause when:

• The same critical defect survives three rounds. The factory has shown its capability ceiling for this item. Another round won't raise it.
• The price drifts upward every round. A quote that keeps climbing as "complexity is revealed" means the factory underquoted to win the order or can't build your product at your target cost. Both are terminal.
• Your feedback is repeatedly ignored or misread. One misread is friction. Three is a capability or attention problem, and production runs are always worse than sampling.
• The timeline has passed the point where the product still makes business sense. A seasonal product approved in March for a December launch is a different product than one approved in October. Do the backward math honestly.

The sunk-cost trap is the real enemy. Four rounds and $1,200 in samples feels like too much to abandon, so buyers order round five. But round five of a failing program is how you end up with a container of defects worth fifty times the sample spend. Killing at round four with a clear log of why is a professional outcome. That log also becomes the precise brief you hand the next factory: here are the three versions, here is what the last factory couldn't achieve, can you?

One legitimate alternative to killing: pause and re-brief. Sometimes the factory is fine and the brief was wrong, because the design as specified genuinely can't be made at the target cost or two requirements contradict each other. A two-week pause to fix the brief beats two more rounds of a factory trying to build the unbuildable.

Budgeting your calendar: the real time math

Each round consumes three chunks of time, and buyers usually budget for only the first:

1. Factory remake time. Simple products: 5–7 days per revision. Complex or tooled products: 2–4 weeks. This is the number factories quote, and the only one most buyers plan around.
2. Shipping time. Express courier from China: 3–7 days. Standard post: 2–4 weeks, which quietly doubles the program. Always express samples; the courier cost is trivial next to a lost week.
3. Your review time. Unpacking, testing, photographing, writing the feedback document: 2–4 days when disciplined, two weeks when the sample sits on someone's desk.

Realistic math for a standard product: 6 days remake plus 5 days shipping plus 3 days review, roughly two weeks per round. Three rounds means six weeks minimum, with no holidays, no clarification loops, and no re-briefing. For full custom OEM work, plan three to four weeks per round.

Plan backward from launch. If production must lock by September and production itself takes six to eight weeks, sampling has to finish by July, which means round one ships in May. Most "urgent" sampling crises are just this arithmetic done too late.

Frequently asked questions

How many samples can I request before a factory loses patience?

There is no fixed number, but tolerance is for productive rounds, not infinite ones. Two to three rounds with clear, consolidated, paid feedback is normal business; no good factory blinks at that. What exhausts patience is round six with feedback that contradicts round four, or repeated rounds caused by the buyer changing direction. If you're paying and your feedback is clean, most factories keep going.

Who pays for revised samples?

You do, by default. Sampling is development work. The standard structure is per-round fees, often credited against the bulk order once production is placed. What you shouldn't pay for are remakes caused by the factory's own errors: the wrong color against a specified Pantone, dimensions that ignored your drawing, brief features that simply weren't built. Cite the version log and the original spec when you push back.

Can I change the design between sample rounds?

Yes, but price each change in time as well as money. A material change can restart the testing logic, since the new material behaves differently and earlier approvals on fit or finish may not carry over. Small tweaks like logo size or stitching color are cheap mid-iteration. Structural changes, a new mechanism or a different material family, effectively start a new round one for the affected parts.

How do I keep track of sample versions?

Keep a simple version log: version number, date received, changes requested that round, review outcome. Photograph every version from the same angles so V2 and V3 can be compared without relying on memory. Keep the physical samples too, labeled, until mass production is done. They are your evidence if the bulk run drifts.

What if the pre-production sample differs from the approved sample?

Reject it in writing, referencing the numbered approved version. The pre-production sample must use the exact bulk materials and production methods; that is its entire purpose. A difference means either substituted materials or a process that can't replicate the approved sample, and both need fixing before you authorize the run. Never approve a PPS that is "close enough." Close enough at the sample stage becomes visibly wrong across ten thousand units.

Should I visit the factory during sampling?

For simple products, usually not. Good photos, video calls, and a tight feedback document do the job. Consider going, or sending an agent, when the product is complex, when two rounds have failed on the same issue, or when the order value justifies the trip. An in-person visit mid-iteration often resolves in two days what messaging can't resolve in two months, because both sides can point at the problem together.

The one-page iteration rule

If this guide had to fit on a single page taped to your wall, it would read:

Three numbered rounds. One feedback document per round. Acceptance criteria written before round one. A limited set of variables changed per round. Every version logged.

After round three, decide instead of drifting. If the sample meets the criteria, sign off in writing and move to the pre-production sample. If it doesn't, diagnose honestly: is the factory incapable, or is the brief wrong? Fix whichever is broken. And if the same defect has survived three honest rounds with clear feedback, walk away with the log intact. It is the most valuable document you will hand the next factory.

Sampling is the cheapest quality control you will ever buy, and a disciplined process tells you with certainty whether the factory in front of you can actually build your product. If you'd like someone on the ground in China running that process, reviewing samples against your criteria and holding the timeline, email hi@cnally.com and tell us what you're building.