How to develop sewing tolerances

spot_image_how_to_establish_sewing_tolerancesBe forewarned that this is closer to an inquiry than a tutorial.

Tolerances are a plus or minus measurement used on a tech pack to determine whether a product meets a specified quality standard. It is usually expressed as plus or minus. For example, one point of measure (POM) for a bust line may say the tolerance is plus or minus 1/2″. This means that for each size (34, 36, 38 etc), the garment bust measure could be 33.5″-34.5″ for the size 34; 35.5″-36.5″ for the size 36, etc.. My inquiry today is how are these tolerances determined? Frankly, it seems like many of them are drawn out of thin air or copied from similar tech packs wherever one can find them.

Tolerances are a new wrinkle and I’m not finding established or good practices to follow. In the olden days, few worried about tolerances because most everyone made their own stuff. Since things have changed, people have been winging it and don’t let anyone tell you differently. That doesn’t mean some people don’t have the right or good answers only that there is no established practice, much less agreement on how to do it.

Some things I’ve seen are using half the grade rule as a tolerance which as a rule per se, doesn’t work. For example, if the grade rule were 2″ and the tolerance were half that (1″), the product could measure an inch over or an inch under and still make the spec. I also don’t agree that the larger the piece, the larger the room for error (permitted tolerance) could be.

In trying to drill this down into something that can be determined in a concrete way, the illustration below is something I’ve come up with. I’m not saying this is the answer, not saying it is the best and I’m not saying somebody else doesn’t have a better solution [but I’d certainly love to hear about it!].


My idea is pretty simple in that it is tied to production processes. In the top pane is a tolerance for cutting the side seam, specifically 1/16th on each seam making for a total tolerance of 1/4″ because cutting is where one can make errors. Ideally a cutting error would amount to one side, once, so I think 1/4″ is very generous.

Still, we don’t measure cut pieces for tolerance, only sewn ones so the lower pane includes a tolerance for sewing. Again, each side is allotted 1/16th. An error or variance occurring at this stage is actually more possible -or should be more possible- as compared to cutting. Adding the two together, we come up with 1/2″ total tolerance for this point of measure laid flat. [Yes of course, as only the half measure is taken this would amount to a full inch; you’ll need to be specific in your specifications as to whether these are full or half (flat) measures.]

In total, 1/2″ (or 1″all around) is too much -in my opinion. Reason is, if you were running numbered sizes with only 1″ difference between sizes, a size 2 could be as large as a 4, the 4 could be as small as a 2 and both would still meet the specification.  So returning to my exercise of attempting to arrive at a solution, I would halve the tolerance from the worksheet to make the final tolerance only 1/4″ (or 1/2″ for total girth).

Ideas? Opinions? Do tell.

Better sewing quality with work instructions

Get New Posts by Email


  1. Jennifer says:

    Your approach to calculating the error propagation is correct–as long as each error is independent, then the total error is estimated simply by adding up the errors in each step as you have done. But to then halve the estimate is completely ad hoc. When you note that an error of 1″ is too large because if so a size 2 could be as large as a size 4 you are really falling back on your own experience that in practice accuracy far better than 1″ is attainable. But this simply indicates you have over estimated the error somehow. The manufacturer of the cutting system ought to be able to quote a spec on how accurate the cutting is, perhaps it is closer to 1/32″. Similarly 1/16″ sounds like a small error to expect per seam but it may be an over estimate, too. Seam gauges are typically marked in 1/8″ increments so a 1/16″ error means consistently allowing the seam allowance to fall halfway past the desired marking. Now anyone who has operated a machine has allowed their work to slip out of position by 1/16″ of an inch many times. But this is usually just a short jog out of position which is then corrected. To sew an entire seam this way is likely far worse than even an inexperienced operator would do. Perhaps 1/32″ is a better guess since allowing the work to slip on average a quarter of the distance between marks is more plausible. I don’t know what the correct estimates are but the thing to do would be study each of the errors individually and get a better feel for what goes on in practice. Regardless, it does seam clear the accuracy decreases in proportion to the number or seams.

  2. Vesta says:

    Thank you, Kathleen. We’ve moved ahead with our ad hoc tolerances, but I would be much happier with a system based on . . . anything logical!

  3. Paul says:

    There is another aspect to this beyond fit/sizing, and that is fit of the pieces as the whole product assembly progresses. So the two broad categories are 1) manufacturing tolerance; and 2) fit tolerance. The first determines the second, but fit tolerance is essential. As one piece grows and another mating piece shrinks due to marking/cutting tolerance stack-up, the seam lengths will now be off, and when this sub-assembly is stitched to another piece (or another sib-assebly made up of one or more pieces) that is over or under due to tolerance, well, this is what ‘toerance stack-up’ means, and can lead to a wasted piece. So really, I see more opportunities for error: Design/pattern (should be +\- 0), marking (if one is hand marking pattern pieces vs. a printed marker), cutting, and sewing, and it can impact both fit and whether the piece even makes it out the door.

    In manufacturing things like engines and cars and complex stuff that shrinks & grows with temperature, this requires complex evaluation of the stack-up, but it would pay, even in our low-tech industry to take a closer look at it.

  4. ticia says:

    In my experience dealing with and preparing spec packages, I have found that one half the grade is fair in most cases.
    However when dealing with larger increments between sizes, for example a four inch chest circumference grade, I only allow 1 inch.
    Most often I have to determine what the final outcome will be or how the variance will influence the fit of the garment.

  5. Paul, I agree with your mention of interdependence of sub assemblies. I refrained from mentioning it or the post would be too long and thought I might follow up with a part 2 if comments warranted it. In this vein, the posts I wrote (the 7 minute cutting test, also pt.2) bear a mention. I completely agree that tolerance in design/pattern should be 0. We routinely expect perfection of airline pilots and brain surgeons so expecting the same of pattern makers isn’t unreasonable.

    Jennifer: absolutely adhoc but with a basis in logic that has been heretofore absent, hopefully.

    The manufacturer of the cutting system ought to be able to quote a spec on how accurate the cutting is, perhaps it is closer to 1/32″

    This made me smile. Even today, most cutting, particularly of DE product lines is done manually. CAM cutting would be closer to 1/64th or less.

  6. Esther says:

    I’ve had to include tolerances in my tech packs. One company I worked for created tolerances for each step of manufacturing including spreading and cutting! The additive problem became very apparent at the end of the line quality control inspection. In any event, I created guidelines for myself. I’m hesitant to post them here, so perhaps in the forum. One thing I will add here, tolerances are dependent on the fabric, product, and price point.

  7. What’s a tolerance?

    Does it mean that an item fails QC if any one of the tolerances is missed at any part in the process? Does it mean that a batch fails QC if 20% of the garments tested are out of tolerance for the completed garment?

  8. You can take my response as coming from someone with no experience telling someone else what the tolerances of their work should be. But it seems to me that it would make more sense to start by figuring out how much variance is allowable in the finished item, and then work through the production process to figure out how to achieve that. Things that are very fitted and made of woven material and have a high retail price would require the tightest tolerance; things that are low-priced, loose-fitting, made of knits, etc. would be more forgiving. In other words, if two examples of a really drapey knit swimsuit coverup that retails for $10 are off by 10%, no one will ever care. If two examples of a very fitted satin bodice that retails for $200 are 5% apart, that would be excessive.

    Any production-based considerations, I would think, would be in the direction of tighter tolerances, rather than the other way around. For example, even if the exact final dimensions aren’t that important to the function of the final product, they have to be close enough for seams to match up acceptably.

  9. Paul says:

    Tolerance is wiggle room. If something is ‘out of tolerance’ then theoretically it is rejected or reworked or scrapped. Your 20% number is itself a tolerance, if you said that you would accept up to 20% of the articles ‘out of tolerance’. We need to set tolerances at each step. It’s a way of setting a limit, so you can say ‘as long as it falls within this tolerance it will still work’. A tolerance is a spec for sizing of the finished product or for any particular piece.

    In your example, you would have to decide whether any specific out-of-spec garment is acceptable. Your 20% example soundsore like a statistical quality control parameter, which is a type of tolerance, but not the same as what I believe Kathleen is talking about.

  10. Kathleen says:

    I think Alison’s question about tolerances was rhetorical but yes Paul, I concur. Measuring tolerances are only one aspect of statistical quality control parameters. For those interested in that (within the context of SOW aka Statement of Work), here’s more:

    Related to this discussion, see the post I linked to at the tail end of this post for an example of a make or break tolerance (of only 1/8″!) that would require repair before proceeding in the work process.

    It’s all relative.
    And people think this business is easy…

  11. Dara says:

    Kathleen, this is a really broad topic, mostly it gets down to customer expectations at a given price point, not what the machines can handle. It does you no good to polish the underside of the bus if no one can see or appreciate. The standard used to be 1/2 inch, but plenty of people (Brooks Brothers springs to mind) have tolerances of 1/32 inch now on finished goods today so the 1/2 inch you state would be way out of line for a target market where customers expect something like that. Maybe not so much for say…Walmart which might have even larger tolerances. However, the same jacket style at Brooks Brothers may sell for a few hundred dollars and $24.99 at Walmart. Every manufacturer I worked for has had different tolerances ranging from 1/2 inch down to 1/300 inch depending on equipment, target market, and worker experience. If you can’t hit the tolerances your competitor can, then you are kidding yourself about who you are hanging with and what your t-shirt, etc. is selling for and you can’t complain if the “designer jacket” you wanted to sell for $590.00 ends up at the end of the season in a clearance rack for $4.98.

  12. Kathleen says:

    I’m not sure I follow Dara.

    it gets down to customer expectations at a given price point, not what the machines can handle

    Customers certainly have expectations but in an era were few know where their waist is, I doubt they’ll be askance over 1/32nd. Frankly, I think if a consumer had access to the specs (not likely), could measure correctly and found a garment to be only 1/32nd off, I think they’d be impressed rather than disappointed. (I’m also wondering if something was lost in translation with respect to BB’s total tolerance being 1/32nd of an inch).

    In reference to the above, You say it doesn’t matter what machines can handle but further along in your comment, you say (emphasis mine):

    Every manufacturer I worked for has had different tolerances ranging from 1/2 inch down to 1/300 inch depending on equipment, target market, and worker experience.

    so I’m not understanding why it was first not okay that machines are a component of tolerances but machines being a component of tolerances is okay with respect to manufacturers you’ve worked for.

    For my part, machines should never affect tolerance. That’s what prototyping is for. You develop allowances that permit successful seam completion with the limitations of a given piece of equipment. Now if someone arbitrarily determines an inappropriate seam specification regardless of equipment demands, that is a whole other subject (which could involve pillorying the culpable party) but is separate and apart from tolerances.

    If you can’t hit the tolerances your competitor can, then you are kidding yourself about who you are hanging with and what your t-shirt, etc. is selling for and you can’t complain

    I definitely agree tolerances can be necessary (lest one suffer the consequences of not having them) which is why I wrote of it. Should I have backtracked to sell the idea? Maybe I erred in failing to sell it first (directly) but instead jumped ahead to wanting to compare notes on how we determine tolerances.

  13. Dia in MA says:

    Tolerances are not just necessary but flunking them can mean bankruptcy for a company. I know because the company I worked for many years ago failed because of them. We were strictly a stitching shop. Errors in the cut goods we received cost the company thousands and finally killed it.

    When one collar piece is cut an inch and a half longer than the other, it is not a “no cut” job. It adds 5 seconds to the job. Costly when the stitcher is supposed take 3 seconds or less stitching the pieces. (Actually that would be a slow stitcher in a factory setting.) Since the job had been bid as “no cut”, this meant that the company lost money on the job.

    Not pattern grading which is the current topic, but examples of how tolerance errors can cost a company.

    For grading, quarter inch off per edge adds as much as an inch variance for 2 seams (4 edges). Make it 4 seams, say for a pair of pants, and you can have the garments varying by 2 inches in the same size. Which makes a huge difference in how they will fit.

  14. Dara says:

    Sorry Kathleen, this was quickly typed out at lunch. Let me try again, according to a brilliant retired patternmaker I knew, he summarized quality as your customer’s perception of your product (Kaizen principle that quality is a vague notion, but the absolute answer in terms of searching for specifications is you must satisfy your customers desire for a quality product as THEY define it.). He said it a lot more crudely “Give the customer what they f*ing think they want, not what you think they do.” The second part of his saying was “Jigs are the lifeblood of a shop, but they cost money. Watch your nickels, they add up.” In other words, you have to balance a customers desire to maximize their dollar against the reality that higher quality standards tend to cost more, especially in tooling costs. Maybe this is what you are calling prototyping? Or you are just that much better than me. The limit of a person’s ability to see fit is 1/64 of an inch, even if it doesn’t effect fit, but ready-to-wear is much more varied as it’s unnecessary/impossible to hit that high mark all the time. Also, pattern piece error grows so more complicated garments are often harder to make hit specs. As more and more computerized systems enter the market people are coming to expect higher looking quality even if it does not affect actual fit. I picked Brooks Brother’s as they are almost entirely automated (they use a Gerber system) and their spec data is public knowledge so it seems like something I could share without it being a company secret. Let’s say you have a client and they want to make a hypothetical leather jacket. Well, I can hand cut it (sample quality) between 1/16 to 1/32 inch with a final variance sewn of 1/4-1/2 inch (Cost 1 hr. time). However, I can also cut off a Gerber file to between 1/200-1/300 inch, sewn to roughly 1/8-1/16 inch final product (3-4 hours). Optimally, I would prefer to die cut it though as our machinist is a stickler and I can do full scale production with parts variance under 1/2000 inch, sewn roughly 1/32-1/64th inch (1 week turn around time and several hundred/thousand dollars, plus sweet talking the senior designer). The die cuts are the most accurate method, but they take the longest and cost the most up front, plus I have to negotiate with 4 people to get them. Which method is the best for which application for you? Don’t know. What you choose will depend on the client in my opinion and how big your run size is. I can tell you shop equipment will vary and you need to design for the actual equipment you have not what you want. I will tell you that I think the industry as a whole is getting tighter and tighter tolerances around me as more and more people buy Gerbers/L:ectras where I am. But what do I know?

    Is that a better explanation for you? I love this blog, but being in menswear and often military not sure how helpful or relevant my experience is for you.

  15. nowak says:

    Not knowing much about sewing industry I agree with Dara and Esther: The “right” tolerance is mostly determined by your price point. Because that’s how it is for technical items and I see no reason why it should be different for clothes.

    So basically you juggle between the biggest tolerance you can get away with (regarding your customer expectations) because the smaller the tolerance, the higher the cost. Because you need to control the goods that arrive at your factory (no matter if screws or cut fabric) if they meet the specs. Which increases costs. And the tolerances for each step of the production have to be set, which again costs money.
    But if you don’t… you might loose money.

    If the finished items (no matter if its a jacket or a… kitchen blender) is cheap the customer is more likely to accept that he will have to try on a big number of jackets with more or less random size tags to find the one that fits him. Or he is more willing to return the blender five times until he ends up with one that works.

    If the finished item is expensive the customer will expect it to fit consistently or to work well at once.

    I can’t say for the US, but in Germany or France it works like that for clothes. If I go to a very cheap store the sizes often will overlap and if a size X fits me in pants type B I can’t expect that pants type C will do the same. Or that I will need size x for next weeks delivery because it fit me this week.

    While if I am shopping more expensive brands I can relay that I will need the same size of the blue or the red pants. And that I will need the same size in spring I had when I shopped my winter clothes. (If my body wasn’t changing, of course.) Even cross brands it will not be more than one size that varies, often it will be the same.

    Otherwise it wouldn’t sell because consumers here expect it. So even middle prized or lower middle prized clothes companies try to remain consistent in their sizing. (Even more important for brands that mostly sell online or by mail order.)

    It might be, that consumers in the US are so low in their expectations that you can get aways with overlapping sizes even in middle prized or expensive brands. Well… lucky producers, the tolerances can be higher. Lower cost with higher prize point is of course good for profit.

  16. Eric H says:

    1/300 inch is 85 microns, about the width of an (undyed) human hair and certainly a lot smaller than a discrete cotton thread. How wide is the pen on your plotter? How accurate is the mechanism moving the pen and paper? What is the kerf of your cutting blade? The resolution of the system aiming the cutting blade? All of these things stack up. It doesn’t matter how badly the customer wants it, you can’t ignore the tolerance stack up enforced by your machinery and processes. I get something like 464 hits when I search Amazon for a book on the subject. The subject of subjectivity is probably mentioned -if at all- only in passing in any of them. Tolerance is the result of your machinery and processes.

  17. Stu Friedberg says:

    In the metalworking machine shop world, tolerances can make the difference between a $10 job and a $10,000 job. Let’s take the example of drilling a hole. The hole has to be in the right place, be round in cross section and be of the right diameter. With sufficiently relaxed tolerances on position, profile and size, someone can burn a hole in the metal with a cutting torch and call it done. Capital investment of $500. With sufficiently tight tolerances on position, profile and size, the work may have to be done on a highly specialized machine in a temperature controlled shop, and definitely must be checked with measuring equipment whose calibration is traceable to the National Institute of Standards and Technology (formerly the National Bureau of Standards). Capital investment of well over $500,000.

    Tolerances are established by the designer, not the sewing contractor or the machine shop. But the sewing contractor and machine shop both have limits on the tolerances they can hold, either economically or at any price. The designer may have to re-design things to make the product at a reasonable cost. Here’s an example. Suppose you’ve got a spiffy patchwork top that requires 24 vertical seams to assemble and you want to hold the finished circumference to a 1″ tolerance on size. If every patch is a separate part, sewn independently, the patch seams must be held to a very tight tolerance because the imprecisions in all the seams (can) stack up. Costly, and impractical. However, if the patches are applied to a conventionally cut foundation, the sizing is controlled by the foundation and is exactly like normal. If you don’t want the extra layer of the foundation, you can work with your sewing contractor to assemble the patches into pieces slightly larger than a conventional pattern piece, to be trimmed to standard size before final sewing. This is still costly and probably impractical, but at least you can control the finished sizing without requiring any unusual tolerance on individual seams.

    In my personal opinion, any brand that emphasizes a distinctive fit should set finished sizing tolerances tighter than +-1/4 of the grade. The typical +-1/2 grade tolerance means the customer has to try on each individual garment to find out how it fits. If fit is a feature, the off-the-rack fit should be consistent, and that requires tighter tolerances.

  18. Patricia says:

    I printed this post to read it tomorrow during lunch break.
    Thank you Kathleen and all who posted their comments. This is extreme-interesting learning.

  19. Katja says:

    As a clothing consumer, I find that sizing tolerance matters more to me than it did in the past. I remember pulling a stack of six pairs of jeans all the same size, style, and color at the Gap and trying them all on, to find that only one or two fit me properly. (I have a long rise for a lady.)

    This was good, in that one could choose the garment where whatever variations in manufacture made it fit one’s body best.

    But now I buy most of my clothes online. I expect them to match the size chart and order accordingly. If they are off, then they won’t fit as expected and I send them back.

    I was disappointed earlier this year with a dress from Anthropologie: I bought a dress in the store and it fit beautifully. I ordered the same dress in another color online and it was almost 2 inches shorter in the bodice. The fit was awful and back it went.

  20. Colleen says:

    And then…there is the discussion with the vendor, who measures the same set of garments you did but finds them “within tolerance.” The back and forth, checking the How To Measure Manual, measuring again, and again…..

    And then….there is the discussion with sales who “need” the garments to sell because they have been shot and the catalog is dropping soon. At this point, they don’t care about the tolerance. If, the garment returns high for poor fit, the same sales staff will demand to know what the heck you were thinking to approve the shipment?!

    And then….there is the designer who’s product is so hot, it doesn’t matter where it falls in regards to spec. The customer’s “gotta have it!”

    And then…there are the buyers who strongly suggest you “find something wrong” with such and such product because they no longer need it.

    Tolerances are one piece of the puzzle. They matter to some companies and are completely disregarded by others.

    To Kathleen’s point – the tolerances are all over the board in the industry! I often adjust the tolerances and add POM to specs just so they make sense to me :)

  21. Adrienne says:

    I feel like there are two main issues here, based on comments, that are similar but different enough that I need some clarification.

    Are we talking tolerances for QC/piece rejection or are we talking ease in regards to fit? It seems like the comments are going back and forth in interpreting the main issue.

    The shop I worked at… we had to make it the exact sizes listed, no room for errors in final sizing. Most of our items were made to fit specific models of weapons, batteries etc so there is no wiggle room. I have brought that same mentality with me in all areas of my sewing. If it is supposed to measure 5″ finished, then it should indeed measure 5″ finished.

  22. Kathleen says:

    I feel like there are two main issues here, based on comments, that are similar but different enough that I need some clarification.

    I wish I could help you. My post was hijacked a long time ago -and in spite of my exhortation that we return to the issues I brought up in the first place which was -how do we determine tolerances? Yeah, tolerances are important, I get that. Yeah I get that price points matter. Yeah I get that styling, materials, fullness and all that is important. What I wanted to get to was how do we develop tolerances in a logical, one-two step fashion.

    Dara: we’ve had this conversation. I’m glad you’ve begun to learn and enjoy its challenges but I’ve specialized in better menswear and leather production for most of my 3 decades in the business. I also explained the accuracies *and costs* associated with die cutting and such in my book (which you have). All that said, it would be regrettable if any readers believed that most of the “tolerances” you mention are measurable to the extent you say they are. Maybe in sheet metal but not fabric and sure as heck not in garment weight leather. All anyone would have to do is run their hand over the cut piece once and surprise! the accuracy of your tolerances evaporate.

    We’ve also had conversations about CAD & CAM which you certainly have seen but do not own or operate yourself. I know for a fact that a Gerber CAM (including Taurus) does no such thing as to guarantee tolerances of 1/2000 and frankly, I do not understand why you mention die making in the same context as the leather CAM solutions as the latter makes dies obsolete. I’m confused. Either you’re cutting [not OEM guaranteed] accuracies with CAM or you’re using dies. And besides, even with dies there is splay over time requiring truing or remaking of dies after however many punches. Granted, this isn’t something you may have had opportunity to observe in the year or two that you’ve been working with leather production.

    And lastly (from another conversation we’ve had), Gerber & Lectra are names you’ve heard of but have not actually used personally and many firms have them but they’re not the only CAD solutions and as to whether they are the best is a matter of careful discussion and comparisons between experienced users of similar systems who are qualified to make these distinctions. As an aside, last year you were beating the drum for generic cad, not understanding the feature set and specialized nature of apparel cad programs and even suggested they were superfluous because you didn’t know anybody who had one. You brought up that the job shop you’re associated with seems to be doing fine without apparel CAD so if you’re bringing it up now, I’m guessing this is a new acquisition (congratulations!). But I digress. I personally do not think Lectra or Gerber are the best. This isn’t the place to say why but besides, Gerber & Lectra don’t have magical tolerances as compared to other CAD programs.

  23. Leslie Hanes says:

    We have this problem daily, because we make custom jackets. Depending on the fabric the customer chooses and who sews it, the tolerance is just one of those things we have to tolerate. As we can’t possibly have every fabric we offer for a sample, the stretch (both ways) of the softshell someone chooses may dramatically affect the fit of the jacket. Despite the 1/4″ rule of seam allowance, if a fabric is quite thick compared to a thin one, that allowance is hard to maintain as we find the thicker the fabric, the harder it is to stay close to an edge, particularly when there may be pile on the underside. Thankfully we are pretty good guessers…and 99% of the time the customer is thrilled. The other 1% we simply remake the jacket with some changes.

  24. CC says:

    What an interesting post. It reminds me of a day in which we were making samples of reversible garments at work and someone along the road did not understand the way the equipment operated… and it was very difficult to make that person understand that the felled seam did not take the seam allowance that person provided with. And all this was a hot mess because the garments came out 2″ off tolerance.

  25. Sohail Farooq says:

    Dear Kathleen,
    Your posted article is so much interesting for me.
    I missed this conversation when you posted but after 8 years your effort is still fruitful for my kind of learners and your article smell is still flourishing just like new spring flower.

    I am really thanks full to you and want to ask so many questions in this regard.

    Measurement tolerance is multidimensional topic I am sure you touch very sensitive vane of garment industry and as per your experience you deserve it very well.

    Please give me permission to communicate with you through this format or via email.
    It would be pleasure for me to learn from you because you are an encyclopedia for garment industry.

    As far as my introduction is concern I have been belonging to garment industry for last 28 years and I had experience in all kind of garment outer wear products. During my career seen 3 countries & 7 cities in the world which countries mainly involved in apparel production export industry.

Leave a Reply

You have to agree to the comment policy.

This site uses Akismet to reduce spam. Learn how your comment data is processed.