Bans get discussed as bad luck and should be counted as a cost line. Once they are, arguments about limits and proxies stop being matters of taste and become arithmetic.
What is lost with an account
- The purchase price. The obvious part, and usually the smallest.
- The warm-up. A month of ramp that the replacement has to repeat.
- The volume not produced. While the replacement ramps, that share of the work does not happen at all.
- Your time. Buying, uploading, proxy setup, redistribution.
Add those up and an account costs several times its sticker price. Which is why aggressive settings that «produce more now» are almost always negative over a quarter.
Estimating the probability
From your own data: how many of the accounts bought in recent months are still alive. That is your attrition rate. Multiply it by the full cost of an account and you have the monthly replacement spend that belongs in cost per lead.
Cheapest risk reduction
In order of return: one proxy per account, respecting the warm-up allowance, pauses between actions, night mode, no mass messaging of strangers. The first two cost almost nothing and cover most cases.
When an account does get limited
Take it out of rotation rather than pushing on — a limited account that keeps trying moves from a soft restriction to a hard one. Appeal, since some restrictions are lifted and it is free. Then find the cause: if several accounts were hit at once the problem is settings, not a channel, and replacing accounts without diagnosing means paying for the same mistake twice. The protections are on the automation page.