You have to fight scale with scale https://www.wired.com/2014/01/how-to-hack-okcupid/
If you strip away the machine learning/computer/math part, the story is: After 88 first dates, Chris finds someone he thinks he wants to marry. (Article says from the first 55 1st dates -> 3 2nd dates -> 1 3rd date).
It seems possible that going on 88 random dates would lead to a similarly shaped funnel.
While "use math" sounds like smart strategy, many (most probably?) happily married people have not gone on 88 actual first dates.
With these kinds of numbers working against you you need scale on your side or your getting nowhere fast. But if you are able to contact 10k people what comes out of those small percentages are decent numbers.