How can I improve soil sampling to get a representative composite sample?

[Sofia33]

I'm trying to design a robust protocol for my field soil sampling, but I'm getting inconsistent results from adjacent plots. My concern is that my current method for collecting a representative composite sample isn't accounting for subsurface variability properly.

[Adam.L]

I tried a grid and pulled samples from many spots in each sub area yet the results still jump between adjacent plots.

[MarkYM]

Maybe the trouble is how we mix and ship the soil to the lab or how we dry and grind it before analysis.

[OliverS]

We did a test where we kept everything the same but collected a single composite and the numbers still shifted between plots.

[Jerry_R]

Is the subsurface variability the real blocker here or is something else at play in the field I am missing?