Since I don't know what your current state of knowledge is, I will start with the basics. If any or all of this is "old hat" to you, please accept my apologies.
The process making a bow stave bend in the desired way is called "tillering". The bowyer puts the bow stave on a "tiller" or "tiller board" and draws the string. The string can be drawn by hand or with a pulley system, according to how powerful the bow is.
Most tillering is a subtractive process. The bowyer flexes the stave as much as he dares, and then judges how much material to remove, and where it is best to remove it. His goal is to make the stave bend evenly from one limb to the other. He must also keep the strains on the stave within acceptable limits at every point.
If he starts with a stave that is straight, and of an even thickness, he will have a relatively easy time telling when it is bending correctly. In this case, the bow stave will fall on a circle when it is flexed.
Unless it is very thin and light, a steel or wood bow stave for a crossbow can not start out straight. It must have some "deflex" Otherwise, the act of drawing it fully will cause unacceptably great strains in the limbs. That is to say, it will either break outright, or will take a permanent deformation. In the case of steel, there is another possibility; and that is that it will be so close to its elastic limits that it will fail in use. (This last failure mode is one that we need to be especially cautious of. )
If the bowyer starts with a stave that is deflexed, it will be difficult for him to tell when the limbs are flexing properly. He needs to look, not at the flexed shape, but at the difference between the resting shape and flexed shape. An experienced bowyer does this all the time, but it is an acquired skill that builds on an innate ability to "see" the shape.
The inexperienced bowyer might find it easier to employ a "tool" here. That tool is the model of which I spoke. While a powerful (wood or steel) crossbow stave must have deflex, a light and wimpy one does not. If the stave is modeled in a thin springy material, there will be no need for deflex. The model can be tillered to a nice graceful curve, and the resulting shape or "plan" can be copied directly in the thicker stave.
The model removes the need to visualize the difference between the final curve and the initial deflex. Instead you only have to see the difference between the final curve and straight.
The model also lets you do the tillering on something that only requires a few pounds of force to flex. This is much easier than having to flex the leaf spring you are starting with. Also, if you overdo it and destroy the model, there is little harm done and not much investment lost.
It is a difficult thing to write clearly about a process that happens in such a "wordless" place in the brain, but I have done what I can. If I can clarify anything, I hope that you will ask.