Most of the models we test in the UIUC wind tunnel are built by modelers. Construction methods used include glassed foam cores (most common) and a few built up models. Recently we have also been using professionally built CNC models, but these have been used for more specialized applications such as airfoils designed here for specific UAV applications. Most of our models still come from "normal" model builders, though.

Accuracies that these modelers attain (as measured on a coordinate measuring machine), vary from an average distance error of 0.0160 inches on the "worst" models (about 1/64") on a 12" chord, to an average difference of 0.004" or sometimes better on the more accurate foam core or built up models. The CNC models are usually slightly more accurate than that, but of course few modelers have access to that type of equipment. It seems therefore that 0.004" average error is about the best you can get when building by hand. With these accuracies, the models behave very similar to the predicted characteristics in almost all cases.

If you have an overpowered sports plane, I doubt you would feel the difference between a "good" and "average" airfoil. However, if you are a competition level glider or pylon racing pilot and know what you are doing, you certainly will see the difference in performance between the different airfoils. At low Reynolds numbers you can often see a 50% or more difference in minimum drag coefficient between a "good" and an "average" airfoil.

Of course, if you cannot build as accurate as this (lets say within 1/64" accuracy or better on a 12" chord), then your airfoil selection is probably not going to always give you what you want. However, I think most experienced modelers can build to that type of accuracy when they really take care (such as building a model for competition purposes), and therefore I do not see any reason why you should not spend the extra time to carefully select your airfoil(s) for a particular design. Even if you cannot build it that accurate, your airplane performance and flying characteristics will still be more predictable if you build as closely as possible to a known airfoil than just using a random shape. Of course, the more accurate you can build, the bigger the chance of actually achieving the predicted airfoil characteristics. Also, some airfoils are less sensitive to building accuracy - and if you know you cannot build that accurate, it may be a good idea to choose one of these airfoils. An example is the type of airfoils used on many trainer and "first low wing" type aircraft. The NACA 4-digit airfoils also seem to work relatively predictable when slightly off on accuracy.

When you see how much wind tunnel data differ between different airfoils (or similarly, if you start studying results from a program such as XFOIL or Profili), you start realizing how much improvement you can get by a) selecting a good airfoil and b) building accurately. How critical this perofrmance difference is, all depends on what you want to get out of a particular design...