A collected list is normally used as a single block: one message to everyone. It is the easiest thing to do and the least effective, because the people in it arrived for different reasons.
What to split by
Source
Someone from a chat about choosing contractors and someone from a general industry group are at different stages. The first has a job; the second has an interest. Same message to both wastes the first.
Signal of activity
People who wrote something rank above people who merely joined — collecting commenters.
Recency
A comment from last week and membership from two years ago are not comparable, even from the same group.
Reachability
Split off the rows you cannot write to before counting anything, or every later metric is diluted by them — why part of your list is unreachable.
What not to split by
Guessed attributes. Inferring gender, income or seniority from a display name produces confident nonsense, and messages written against it read as misdirected.
How many segments
Two or three. Each one needs its own text, and texts are the expensive part; ten segments means ten messages nobody has time to write well, which ends as one message sent to ten lists.
Clean before you split
Splitting an uncleaned list produces segments of uneven quality and makes every comparison meaningless — cleaning a database.
What to write once the segments exist: cold outreach.
How Neurogram does this
Everything above is manual work that runs into the number of hours in a day. Neurogram does it on your own accounts and without you: it collects channels and audiences, writes comments and replies with a language model, and keeps limits and quiet hours so the accounts survive. See pricing or look at your case.