Typically I used to do a simple ten-round set and used a ladder test to find a sweet spot load (low SDs, probable accuracy) and then second / tweaking sets when I loaded my 'barely major' semiauto rounds.
With my short-barrel reloading project, I've strived to develop a broad-based set of data for each powder. That is, the bullet weight is fairly well set, and a given bullet is used. Then I build 10-round sets through a published charge range. Accuracy is not the driving paradigm; velocity is, within generally-accepted parameters, and / or a recoil "feel" similar to desirable SD rounds.
Typically procedure involves 20 cartridges, fired in two sets of ten. The first ten is fired for "subjective recoil" assessment and usually goes over the chrono as well. Those are fired from a 640 (2"), and the second set then goes over the chrono from the M60 (3") or is used in the 640 if there has been unusual or incomplete readings from the first set.
The ones replicating a factory SD round are then loaded again for verification of the first data set and for accuracy and POA shooting. These tests are done at 15 yards.
Once tweaked, then a set of 50 rounds are built and fired one last time in four barrel lengths--1&7/8", 2&1/8", 3", and 4". Two of those are j-frame lightweights, and the 4" barrel is a 686.
Occasionally, I chase a particular load completely--recently I fired a set of 38 Spl / SR-4756 / 158LSWC cartridges throughout its reported loading range--from 4.8 gr. (current Hodgdon min, 38 Spl) through the two "FBI" variants (6.4, 6.8 gr. in 38 Spl) and on up through 8.0 / 8.2 gr. (the Speer #8 loads).
As you can see, OCD is a reloader's hazard. The above regimen also explains why I wanted to move in with either The Bushmaster or Walkalong in April, when Minnesota's winter refused to end.
Jim H.