Same list, same hours
Run the bots on a portion of the list while your fronters work the rest, in the same hours. A pilot on last month's leftovers, dialed at a different time of day, tells you nothing you can act on.
Agree the number first
Transfers per hour, or transfers your closers accept, pick one before you start. Two weeks of argument about which metric counted is the most common way a pilot ends without a decision.
A few hundred dials is enough
Enough to see the distribution, small enough to stop. If the answer is not visible in the first few hundred dials, the problem is usually the script rather than the sample size.
