Multisize me more
More on grading a knitting pattern for multiple sizes

Refining the selected target sizes

In the last issue, we looked at how a knitting pattern, written for one original size, can be rewritten for multiple sizes. The example worked in that article, however, was a simple example: there was an implicit assumption that there would be little, if any, difficulty in upsizing/downsizing the pattern to different dimensions. In reality, design details like stitch pattern repeats -- whether small, like a finely textured stitch pattern, or large like a cable panel or stranded colorwork pattern -- will dictate where, and by how much, a garment piece can be altered without throwing the overall design off-kilter.

You may have determined, in advance of actually computing the numbers for additional sizes, that you wished to write up instructions for a given range of sizes, with a given frequency; say, from 30 inches to 60 inches full bust measurement, with sizes incremented by 6 inches: this would give you target finished full bust dimensions of 30, 36, 42, 48, 54, and 60 inches.

It's easy to adjust the width of a garment piece when there is only a single focal design element on the body. The diagram at right shows a garment front that is wider than the one at left.

If your garment design consisted of only a simple pattern repeat or texture (say, stockinette or a small-sized repeat such as moss or seed stitch), writing the pattern instructions for each target size is relatively easy.

Even if the design consisted of a single focal design element -- for example, a single, centered cable panel, set off on either side by a simple background stitch -- resizing the garment will still be easy, because most (if not all) of your size adjustments will occur in the background stitch area, and the construction of the garment incorporating the design element will be left more or less untouched.

But let's say that the design feature of your garment repeats horizontally, and is not merely a single centered occurrence -- for example, a series of 4-inch wide cable panels, each separated by half an inch.

Your original design, written for a 48-inch full bust measurement, might have a front (and back) designed like the one at left. In this diagram of a sweater front, four-inch-wide cable panels are separated by half-inch columns (which may consist of a background pattern or a dividing pattern stitch), except at the edges, where they are framed by one inch of background. The total width of the piece is 24 inches; doubled for both the front and back, the full bust measurement would be 48 inches.

Minor changes in the width of the garment pieces can be easily accommodated in this design.

For example, say that the next smaller size was to have a 46-inch finished bust measurement, which is only a difference of one inch on each of the front and back. Where you choose to subtract (or add) the extra width will be determined by how the garment is supposed to fit and look.

If the extra width is removed from the left and right edges, that extra width would probably (although not necessarily) be eliminated in the armscye shaping so as to keep a cable panel right at the sleeve-body seam; the cross-shoulder width of the garment would be the same as the original, 48-inch size.

If the subtracted width is distributed throughout the front and back by omitting a stitch in some of the background columns (if the design can accommodate that), then you might leave the armscye shaping as it was in the original 48-inch size, but the cross-shoulder width of the front and back will be greater than for the 48-inch size.

When the front or back requires only a small adjustment in width (here, one inch), the change can be easily made without interfering with the overall garment design. The original at left; at center, half an inch is removed at either side (in the underarm region); at right, one-quarter inch is removed from each of four locations between the cable panels.

Larger changes to the garment size, however, might not be as easy to make.

Consider the next smaller target size of 42 inches: assuming the front and back follow the same patterning and are of equal width, this means that each of the front and back must be reduced by 3 inches.

One solution is to remove one cable panel, and increase the width of the background framing the left- and rightmost cable panels, as shown at right.

However, by removing one cable panel, the overall arrangement of the cables has been altered: there is no longer a centered cable panel, which may or may not wreak havoc with the neckline shaping. And if the sample garment for the pattern is shown for a single size only, it will be misleading to knitters working a different size.

A more radical alternative is to remove two cable panels, and respace the remaining panels accordingly:

At left, an extra-wide border of background stitches frames three cables, spaced as they were in the original design. At right, the three cable panels are repositioned from their original locations.

These solutions might make the design look a little odd -- there is now a lot of "empty" space.

A further solution would be to add "filler" cables to the extreme left and right, or even between each of the cable panels:

Additional filler cables are added in these examples to fill in the empty space resulting from the removal of two of the original cable panels. At left, two filler cables are added at the sides; at right, four narrower filler cables are inserted.

But each of these alternatives is a redesign of the original cable pattern that will also require a rethink of the original design -- if you insert filler cables between each original cable panel, then for consistency, your original size should have these filler cables, as well; if you simply add filler cables to the left and right edges, it may not be necessary for every size to have that filler.

In other words, the larger the constraining pattern repeat, the more difficult it may be to rewrite the pattern instructions for a specific target size while maintaining a consistency of design for all sizes. This is probably one reason why you might find that some patterns for garments with big cable or colorwork repeats are offered in only a few sizes, and are typically oversized -- it's the easiest way to maintain a consistent design for each size, without having to create alternative filler patterns or respacing the original pattern repeats. (This is often compounded by the fact that many designs with large pattern repeats are oftened offered in drop-shoulder or other styles that are typically loose fitting to begin with.) However, reducing the number of sizes offered isn't always a desirable solution; in those cases, think about adding filler design elements instead, if the design can bear that kind of alteration.

While the example here deals with cables, the same logic applies to other techniques with pattern repeats, like colorwork, other textures, and lace. Colorwork can be more flexible when it comes to multisizing: you can often (although not always) truncate the repeat at the sides of a garment (whether there is actually a side seam or not); if the pattern is mirrored on the other side of the side "seam", the design may look deliberate (and even attractive). If the colorwork pattern repeat is just a little bit short of the required width, a different panel can be added at the sides, just like a "filler" cable or background in a cable design. A cable panel, of course, cannot always be cut off as the colorwork pattern is in the left-hand example without significantly altering the cable design itself. A lace or other texture pattern may or may not survive this kind of treatment.

Other ways to make the grade

Grading by number-crunching, as described in the previous issue, is not the only way of multisizing. One could knit to fit the hypothetical fit model for each size of the pattern, which is more or less like test-knitting the pattern in each size -- but this does drive the cost of pattern development up significantly in both supplies and labor.

You could also take a practical approach by drawing the garment pieces in the prototype size to scale on paper, and by grading the pattern for different sizes by drawing the pattern shapes directly on the paper; the new shapes would then be measured at key points, and knitting instructions written to match these shapes. In effect, this is like the approach described in the last issue, but with an intermediate step of creating a scale model of every piece.

Another option is the soft solution: using knitting pattern software to carry out your computations. There are a number of software products available now that will work out the numbers and instructions for a knitting pattern in a single size, based on the parameters input by the user and a starting size. By applying the same or similar parameters to a different starting size, numbers can be generated for multiple sizes, then consolidated into a single set of pattern instructions.

Knitting pattern programs are a type of specialized computational tool -- unlike a calculator, they're preprogrammed with certain assumptions about garment shapes. Because of those assumptions, this type of software can do a good job of generating the numbers to be used for the basic garment shapes for which it was programmed, and save you the labor of computing the numbers yourself; however, because of those same assumptions, the software is potentially limiting when it comes to drafting your knitting patterns.

Some software is capable of generating numbers and instructions only for simple sleeves, such as trapezoidal set-in sleeves or simple raglan sleeves, which isn't too useful when you've designed a pullover with a fitted sleeve cap (the type with an S-curve or point of inflection) or a compound raglan seam involving multiple decrease angles. Some software may come pre-programmed with certain fitting assumptions, like the amount of ease required for a "fitted" silhouette or whether armscye depth should be a fixed value for a given size, regardless of style. If you don't share those assumptions, then the numbers generated by the software should be treated with caution. (In addition, the written instructions generated by pattern software may not be the best way of expressing your instructions to other knitters, and are probably not in the proper format if you are planning to submit your pattern for publication by a third party -- you'll probably find yourself rewriting the instructions generated by the program as well as revising the numbers it calculated.)

Even if pattern software was used to generate basic numbers as a starting point for a prototype garment, the process of design and pattern writing often involves changing those numbers to fit the designer's vision and to accommodate sizing requirements. Plus, the more complex the design (in terms of construction method or stitch patterns), the less likely it is that pattern software will perform adequately; you'll probably find yourself reviewing all the numbers it generates, and rewriting them to fit your design parameters. This may negate any time savings you might have otherwise realized by using the software in the first place.

Alternatively, you can create your own labor-saving, customized computational tool using a spreadsheet program -- for the initial investment of time and effort in setting up the spreadsheet and formulas, you can automate many of of the calculations required to multisize your pattern. These calculations will also be tailored to suit your design and fitting assumptions, which is an advantage over using somebody else's off-the-shelf pattern software.

Estimating yarn requirements

Estimating the amount of yarn required for the various sizes in a pattern is another exercise in basic mathematics. Assuming that knitters may be substituting yarns and will require as much useful information as possible, it's best to express the yarn quantity as a length (in yards or metres), as well as provide other information about the yarn's characteristics if they are important for yarn selection (for example, the yarn's weight -- that is to say, gauge -- is important).

1. Determine how much yarn was used in your prototype.

You might have kept track of the yardage of the yarn you consumed in making your prototype. If you didn't or couldn't do this, you'll need to determine the yardage a different way: for example, you can compute the length per unit weight of your yarn (e.g., yards per pound or metres per kilogram), and weigh either your finished product or the remaining yarn (if you knew the starting weight) as accurately as possible. A postal scale or a kitchen scale will probably work nicely; the more accurate, the better, although you'll be rounding up this value later.

If you are weighing the item and it has a zipper, ignore its weight contribution; just include it in your measurement -- it's safer to slightly overestimate the weight of the finished garment. However, if the item is beaded or has particularly heavy buttons, it would be better to discount their weight, or to weigh the remaining yarn instead.

2. Estimate the increase/decrease in surface area for each additional size as a ratio.

For each size in the pattern, there will probably be an increase or decrease in the lengths and widths of each garment piece, and consequently an increase or decrease in the total surface area of the garment, when compared to your prototype size.

For the purpose of estimating yarn estimates for other sizes, you can approximate the surface area of garment pieces using rectangles and triangles.

First, work out the approximate surface area of your original size. Fortunately, you don't need to be precise; you can approximate the garment shapes with rectangles and triangles. You might want to be more precise if you actually needed to know the surface area of the garment, but in this case we're not interested in the actual surface area numbers -- we're more interested in the percent increase or decrease for each size.

For the simplest body shapes of all -- a rectangular body, such as that used for a drop shoulder pullover -- the surface area computation is simply the width of the body times its length. In a more complex shape, like a set-in sleeve pullover, we can break up the front or the back into two rectangles. The upper rectangle has a width equal to the cross-shoulder width of the upper body (after all the armhole decreases are done); the length is the armscye depth. For this approximation, we can just ignore the fact that there's a bit of surface area at the lower armscye and at the shoulder not covered by the rectangle (see the diagram), and the fact that the rectangle covers a lot of empty space at the neckline -- it's just an estimate.

The lower rectangle is given the maximum width of the lower body (one of the hem, hip or the bust measurement), and the length is measured from the hem to the beginning of the armscye shaping. Any waist shaping or other shaping in the lower body is ignored. So, the front (or back) of the garment has a surface area roughly equal to:

	cross-shoulder width times armscye depth
plus
	widest body measurement times length from hem to armscye

A set-in sleeve can be represented by a triangle and a rectangle. The triangle represents the sleeve cap, with a height equal to the height of the sleeve cap, and a base equal to the width of the sleeve at the bicep (usually the maximum width of the sleeve, unless the sleeve is flared). The rectangle represents the rest of the sleeve: its length is the length from sleeve hem to the beginning of the sleeve cap shaping, and its width is the average width of the sleeve (here, the average of the bicep and the hem measurement: add them together, and divide by 2). The total surface area of the garment is therefore approximately:

	bicep width times sleeve cap height divided by 2
plus
	(width at hem plus width at bicep) divided by 2, times length from hem to bicep

The total approximate surface area is, of course, the sum of approximate surface areas for each major piece of the garment: front(s), back, sleeves (don't forget, there are two sleeves!). Minor bits such as collar bands can be ignored, but if the garment has a hood or a large collar, you might want to add those in as well: hoods and big collars take more yarn than you think, and depending on your design and sizing may or may not scale up or down at the same rate as other parts of the garment.

Once you've worked out the approximate total surface area for each size, you can also express these numbers with reference to your original size. Let's say that the original had a surface area of approximately 1150 square inches, and took 1310 yards of yarn. If three other sizes of your pattern had approximate total surface areas of 1121, 1172, and 1201 respectively, then you could express these sizes like this:

3. Apply this percentage to your prototype quantity.

These ratios can then be used to estimate the yardage requirement for the other sizes. If the original size required 1310 yards, then size 1 would require approximately 0.975 times 1310, or 1277 yards. Sizes 3 and 4 would require 1336 and 1362 yards, respectively.

4. Add a "fudge factor", and (optionally) round off your yardage requirements to the nearest 5 or 10.

Now, add an extra percentage (the "fudge factor") to account for variations in knitting styles, because not everybody will use exactly the same quantity of yarn, and some people will knit larger swatches than others. For example, you might choose to round up the quantities by 10%. So, if you estimated a yardage requirement of 1362 yards, with the fudge factor added in you'll have 1498 yards.

You can see that the selection of this fudge factor is not insignificant: we've just added 136 yards, which can be equivalent to one or two skeins of yarn, to the yardage requirements. The purchase of a couple of extra skeins can add a substantial cost to the project; on the other hand, underestimating the yardage requirement can be fatal to a project if the knitter runs out of yarn and can't get more in the same dye lot. While you might feel comfortable adding in only 5% extra, a 10% allowance will probably provide enough yarn for a good-sized swatch (which can be ripped out and used if the knitter runs out of yarn), and/or some leeway in lengthening the garment.

Next, you can choose to round off your yardage requirement to a tidy-looking number ending in 5 or 0. This isn't necessary, but it might seem kind of silly to be precise about a number like "1183", when you've already added in a fudge factor. You may as well express this value as "1180", "1185", or "1190".

Finally, if you are specifying a particular yarn brand or put-up, convert your quantities to a number of skeins (rounding up to the nearest whole skein).

References

Marnie MacLean has written a series of detailed tutorials for using Microsoft Excel for knitting patterns, in three parts: one, two, and three.