In all courses where soil testing is discussed, the importance of collecting a representative sample is stressed. Sample collection is usually cited as the weakest link in the soil testing scheme. Good soil sampling is difficult concept to demonstrate, especially in introductory classes. Ideally you would go to a site, sample according to two or more sampling schemes and compare the results. Unfortunately the number of sampling scenarios that can be tested are usually small and by the time the soil test results return from the laboratory, soil testing may no longer be the lecture topic. We have developed an on-line soil sampling exercise that runs in an HTML browser. Users click on a map to indicate where a soil core is to be taken. It is possible to write field notes, collect several cores into a single sample and collect multiple samples and at a site. An established spatial structure is kriged to create a spatially dense dataset which is then converted to a x,y,z response surface function, where x and y are spatial dimensions and z is the soil test parameter(s) of interest. This function is used to determine the ‘true’ average soil test value for the site and to calculate the value that would have occurred had an actual sample been taken at that point. When users click ‘Final Result’ a soil test report, complete with field notes is generated immediately. With this application it is relatively simple to test the effects of core number and/or sample number on the quality of the resulting soil test recommendation, or any other concept where the spatial variability of soil properties is of interest.