The term norm group and 'comparison group' are used interchangeably. Norm groups are used for benchmarking, and are the best way to see how a score compares with the scores other people got.
A straightforward 'raw score' of for example 16 out of 20 only tells us something if we know what other people got in the same test. Is 16 / 20 a high score, a low, score, or about average for that test? We don't know until we can compare it with what other people got.
This is where percentile scores come in handy. They tell us the percent of scores which are lower than the score being considered.
For example: if a candidate scores in the 90th percentile, their score is higher than 90% of the scores achieved by people in the norm group. In other words they are in the top 10%. A percentile score of the 50th percentile means 50% scored lower than the candidate and 50% scored higher; this score is exactly 'average' compared with the people in the norm group.
So you can see how it is important to know the make-up of the norm group. Whose scores are the candidate's score being compared against? Is it fair to compare their score with scores achieved by managers, or other graduates? It's usually best to pick a norm group comprising people of a similar level to the person you are testing. That's the fairest comparison, and probably the most useful to you.