A problem of ore dressability study, containing very valuable components, represented both by coarse and fine inclusions, has been considered. It is shown, that on the basis of small laboratory samples it is impossible to achieve sufficiently high repeatability of balances with respect to such ores. Efficiency must be estimated not by valuable component recovery into concentrates, but by its content in final dump tailings and their yield. Reliable estimate of such ores dressability requires pilot-scale testing on large several-tons samples.