Thanks to Chris Hunter, both when I am vetting or creating a WODB, I use a table like this where there are 4 criteria/characteristics labelled along the top.
I consider it a strong WODB if 3 of the items share a characteristic that is not present in the 4th. For example:
The criteria used for Shape 5 could be:
I find that working through one like this with students is a really important step for them to understand that "It's the only one that's pink" is not the depth that we are looking for (unless you are looking at colours!).
I also strongly encourage you to do some Incomplete Sets before your students create their own. Here is an example:
I love that there are many different options for the missing number here. And students (may) quickly see how helpful the table is to ensure a strong WODB.
What have I missed? Let me know in the comments!