Quality Control and SOW pt.2

Posted by Kathleen Fasanella on Oct 15, 2009 at 3:59 pm / Operations, Quality / Trackback

~sigh~ I had another entry planned for today but it just didn’t gel nicely. I really wanted a topic break between this entry and the one that preceded it but it doesn’t work out sometimes. You know, writing is hard work. Or no, it’s easy. Just sit down and open a vein (Red Smith) …but I digress.

Returning to our subject matter (part one, part two), how do you sample a lot for testing? What is sampling a lot? How do you know the whole lot is bad or good without examining every single piece? Or maybe, you do want to examine every single piece, how will you know? Then, you have to document each inspection of each unit.

What is sampling?
Sampling a lot means to pull out a few garments to inspect them. The quantity of garments you decide to inspect shouldn’t be random but the selection of items should be random. Most people think they’re pretty good at deciding what’s random but few are. I love this random number generator which is perfect for this (or you could buy this book -the reviews are hilarious and worth a click). You put in the size of the lot (number of pieces) and it will tell you which one to pull out to inspect. Yes I know that regenerating the generator resets the timer but it’s a start.

What kind of inspections?
Mr. Fashion-Incubator and I discussed this at length because we actually like these things. We’ve decided you need to do two levels of inspection we’ll call A and B. Level A inspection is brief; you check the obvious such as label placement, seam matching, and stitch quality. With the level B inspection, you do a level A plus you check everything from finished lengths such as sweep, sleeves etc and compare them to the dimensions listed on your tech pack or pattern. This is one way you can get toward the goal of weighted criteria I mentioned in the second entry (this entry is technically the third). Errors in level B are more egregious. Not to say you can’t find a way to work around those but you’ll definitely need to go back into product development (patterns, grading and such) to figure out how things went wrong.

How to know how many to inspect?
We had a lot of problems answering this question due to all the factors. Here are some questions we came up with:

1. What is a lot?
2. How large is the lot?
3. What’s your market and who is your customer?
4. Is this a first delivery from this contractor?
5. Will you be shipping to a new customer?

What is a lot?
A lot should be one SKU. I’ve written about this before, likely to your dismay. If your shipment is 100 items in 4 different sizes and 3 colorways each, that’s  about 12 lots with 8 units in each one -assuming orders were in equal size and color quantities. The matter of sizes and colorways leads to complex formulas in the garment industry, so onerous they make my head hurt. In essence, each size/colorway is its own lot. Small red is totally different from a medium green. If there were 100 medium blues, that’s one lot -which is why the number you select to inspect from each lot will vary. Speaking of, do you know how to inspect a colorway?

How large is the lot?
Large and small are relative. If it’s a thousand pieces, you’re not going to be able to inspect every single one so you might want to invest in an inexpensive membership at the random site. The percentage of units to inspect matters with large lots. You might think 10% is a good figure but a lot size of 1,000 pieces works out to 100 units so something like 3% may be better (caveats below). If the lot is only ten pieces, you might inspect half of them but again, it depends on other factors.

What’s your market and who is your customer?
If your items are loose fitting, casual, moderately priced attire, you can inspect fewer pieces. If you sell bridge priced dresses to high dollar boutiques, you’ll want to inspect more. For a small lot (10 pieces), higher priced brand, I’d probably do a level A inspection on each unit and a level B on maybe as many as 30%.

Is this a first delivery from this contractor?
If it’s a first delivery from a new contractor, you should probably select more pieces than you normally would, probably double the number you’d normally select. If during the process of inspection, a pattern of consistency develops either negative or positive, you can either lighten up or give up, depending. For example, if the quality appears stellar and uniform, you can relax a little and go to the standard inspection lot size. If it’s bad, well, it’s back to the whole matter of pass/fail.

Let’s say the lot is 100 pieces and you’ve pulled ten out for inspection. This means the value of one piece equals ten units. If you’ve decided to weight the criteria and have a figure in mind for lot failure (>70%) and your first three fail, then you’ve likely got a bad lot and may throw up your hands. If you want to prolong the agony of continuing the lot inspection, it wouldn’t hurt to have more data with which to go back to the contractor for repairs or redress.

Will you be shipping to a new customer?
If this lot will be shipped to new customers, you might want to take a bit of extra care. Do your usual inspection but be mindful in the packing process. To prevent yourself from re-inspecting, get some of those little dots or find a way to unobtrusively mark the packaging label, say on the poly bag if you use those.

Documentation:
There’s no set form out there you can use (unless you have something like StyleFile), so create a form to use and edit it as you get better. If a garment fails, you’ll need to attach the form to the item and set it aside. Garments that pass go back into the pool and need no further controls beyond the unobtrusive mark or sticker I mentioned already. Ideally, you need two copies of each inspection report. The documentation will help you write a report to present to the contractor. In writing your documentation, be sure to mention your sampling method, as in, how many you decided to pull from each lot and the level of inspection for each. A lot of existing research is bogus because researchers used poor sampling processes or failed to document it.

You need to decide what you’re going to do. If the item can’t be sold, it’s already a loss. This is the stage where most people make analytical mistakes (Kahneman won a Nobel for proving that people will spend more money in an attempt to salvage a loss, than they’ll spend to get a new gain). If you want to repair a problem, this constitutes a new loss. You really need to look at the gravity of the situation and other intangibles to come to the complete equation. Your choices are to let them repair it (maybe done well, maybe not) or send it to someone else and bill the original contractor for repair. If you aren’t going to keep that contractor, it may be better to use another contractor that specializes in repairs or a local sewing shop because an incorrect repair will simply move the problem from one party to another. If it’s a contractor you want to build a relationship with, work with them to resolve it.

• Share this post
• Email to a friend
• Add a comment