Load distribution is where a ten-account setup either runs for months or collapses in a fortnight. The deciding factors are not obvious, because the intuitive approach — divide everything equally — is the one that fails.
Why equal shares are wrong
Accounts differ in age and history, and tolerance scales with both. An even split runs the youngest accounts at the same rate as the oldest, so the young ones burn out first and the old ones are underused.
The workable rule is proportional: new accounts carry a fraction, and their share grows as they age — see warming up accounts.
The channel list splits, not the work
Divide the sources, not the volume. Each account gets its own set of channels and stays with it. Two of your accounts appearing under the same post is the clearest signal of a scheme there is, and it happens automatically as soon as the split is by count rather than by list.
This also caps the fleet: you cannot usefully run more accounts than you have live channels — how many accounts you need.
What no single account should carry
- The whole flow. If losing one account stops the work, the setup has a single point of failure regardless of how many accounts exist.
- Your main channel's identity. The account that owns the channel should not be the one commenting in fifty places.
- Both the outreach and the replies. The one that talks to interested people is the one you least want restricted.
Timing is part of the split
Accounts that wake, work and sleep together are linked by that schedule whatever their proxies say. Different windows per account, and none of them round the clock — quiet hours.
When one account starts failing
Reduce its share rather than moving the same load onto its neighbour. If the cause was the pattern rather than the account, the neighbour is next — what to do about a restriction.
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.