How many positive responses are good enough for building a marketing response model when the response rate is low(0.5%)

We are planning a marketing campaign to collect data and the response rates for a random sample. Total population size is 10 million and historically, response rates are low (0.5 - 0.65 %).

How long do we need to run the test to collect a decent number of positive responses and how many positive responses need to aim for ?

I know there are no general rules for deciding the model data size as it all depends on the attributes and how predictive they are?

I am planning to run the test so that we get 5- 6K positive reponses for our campaign. We may need to send atleast 1 million mailers to get 6K responses. Only reason for proposing 6k number was aiming to collect as much as data and at the same time not spend too much on marketing.

Please let me know your thoughts about test sample size.