My thoughts - epd's for a calf are initially calculated based on the epd's of the sire and dam. As additional measured data on the individual is submitted (actual birth weight, weaning weight, yearling weight, marbling, height, etc), that data is used to update the epd's. Then as progeny come along, more measured data is available which can be used to further refine the epd's. Resulting in more accurate numbers. Remember those initial epd's were based on the average of the parent's numbers.
We know that not every mating will produce an animal that is the average of the parents. In fact, we hope that they will occasionally produce progeny that are superior to the average. So, I sometimes hope for some variation in the numbers after the dna is done. Of course, the epd's may get better or may get worse, but the main hope is that they get more accurate.
Most of the time for me, the genomic enhanced epd's are pretty close to the originals, but occasionally there is a more significant change. Lot's of examples of flush mate bulls that start out with the same epd's, but end up with significant differences. In growth, birth weights, calving ease and daughters. Looking at markers in the dna profile and correlating those to physical traits is a head start on getting measured data on progeny. It's another tool. Usually tools get better with time and as people get experience with them.