For more distant relationships, the relationship listed on the DNA Relatives feature will often be a range. The randomness of DNA transmission across generations makes it difficult to pinpoint the exact relationship for more distant relationships. The relationship range should be treated as the approximate bounds of the relationship.
It's important to note that some populations have higher than average DNA sharing, in which case the relationship range will indicate closer relationships on average.
Each match will also have a specific relationship prediction that you can view by clicking the match entry. DNA Relatives estimates a predicted relationship and range using both the number of segments and percent DNA shared. In general, for the same percent DNA shared, longer segments give a closer predicted relationship than a greater number of shorter segments. The estimate is based on a number of assumptions about population growth and characteristics of specific populations.
There can be more than one type of relationship that shares the same percentage of DNA. You can edit this specific prediction in the compare view if you know the actual relationship.
For more distant cousin matches, the predicted relationship should be treated as a best guess. Closer relationships tend to have a tighter range, reflecting the greater amount of information found in closer relationships. The accuracy of the degree of relationship predicted increases as the relationship gets closer.
Matches are labeled as "distant cousins" when the degree of relationship is difficult to estimate, due to the small amount of DNA shared. For every degree of separation in a relationship, the average percent DNA shared drops by half, so that the percent DNA shared remaining is quite small when you get to distant cousins.
You can find a list of the average percent of autosomal DNA shared by relatives on the International Society of Genetic Genealogy (ISOGG) website.