Getting back to pricing methods... I came up with a set of models that when plotted in various ways help to reveal what factors are being used to price prints. Again useful at a service that doesn't publish exact methods or rates. Of course this is somewhat dependent on a service to provide their measured values of material volume, machine space, and surface area after model upload. If they don't you'll need to calculate those yourself - with the machine space potentially being the most difficult to calculate depending on your test model geometries.

I generated two basic sets of models that hold either material volume or machine space (volume) constant and vary the other. The first set is a cube hollowed out by various sizes of smaller cubes with a small through hole from the outside to any inside cavity. Therefore this set varies material volume while holding machine space constant.

The second set is another cube spread out into various shapes to vary machine space, but the material volumes are constant for each shape. The size of this cube (37mm) was chosen to have approximately the same material volume as the most hollowed out cube of the other set to provide an intersection point.

Here are the models used. (The spread4 model is hollowed from one end. The hollowing not observable from the preview thumbnail.)

I generated plots on a spreadsheet for material price as a function of material volume, machine space, and surface area. The idea is that holding some model parameters constant will reveal pricing slopes on certain graphs. Basically a straight line with a slope reveals a cost dependency on that parameter. First up wsf price vs material volume and wsf price versus machine space graphs. We can see that there's a price relation of $0.28 per cm^3 of material and $0.21 per cm^3 of machine space.

Plotting wsf price versus surface area gives a plot that shows no perfectly straight relation between surface area and price (there are slight curvatures). In fact the two model sets show different correlation trends - one positive and one negative. I haven't decided yet if it would help to generate another set of models that holds surface area constant and varies material volume or machine space. So far I'm not sure that's necessary or easy to do.

A fly in the ointment is determining any constant part handling fee or minimum baseline fee that transitions to material/machine/surface area costs. For determining those you need models that are closer to minimum size to show those pricing transitions. The method outlined here is primarily good for determining various individual linear price components although some additional mathematical manipulations might be useful to tease out those other cost additions. Another fly in the ointment is price schedules changing for oversized objects.

I suppose another way to do all of this is to create a wide spread of models where all the basic model parameters have large variations between models. Then generate a basic equation that includes a bunch of pricing methods and run a Monte Carlo analysis to fit the pricing data of all the models (for a specific material) to an equation. I'm not there yet.

Oh, and one more mistake I made was that the 37mm box I chose is too small for ceramic models (minimum would be 40mm on the 2 large bounding box sides). I'll probably adjust those, rinse, and repeat the whole shebang.