First of all I measured the background count over a long integration time. I clamped a beta source to fire through aluminium into a GM tube and counter. I measured the thickness of the aluminium in several places with a micrometer screw gauge and calculated the mean. I took a reading of counts over the long integration time and subtracted the background count. I added more thicknesses of aluminium and took more data until I had enough data to plot an accurate curve. It was then possible to work out the thickness that halves the counts. It is best to do this in 3 places and find a mean. This is a thickness that halves the counts so is called the "half-value thickness" - sort of distance equivalent to half-life.