Currently, the Duplicate Management Tool seems to match constituent records based on any similar first or last name. For this tool to be effective, it really needs to be able to increase and prioritise the matching criteria as it's impossible to check through thousands and thousands of individual records.

An example could be, match based on;
Priority 1 - first name + last name
Priority 2 - first name + last name + email address
Priority 3 - last name + email address
Priority 4 - last name + phone number