View the original community article here
Last tested: March 2021
The relationship parameter only tells Looker whether or not to use symmetric aggregates. So, the worst case scenario from an incorrect relationship is that Looker does not apply symmetric aggregates when it really needs to. Thus, you can have fanout and potentially incorrect data.
If you leave the "relationship" parameter blank, then Looker will apply "many_to_one".
If you don't know the what the relationship is, "many_to_many" is the safest choice. The worst it will do is add unnecessary symmetric aggregates, which may cause slower queries but guarantees that fanout will not cause incorrect data.
This content is subject to limited support.