Let it spend three to five times your target cost per booked call before you make a decision. For a lower-cost event like a lead, give it more room, seven to ten times your target. In days, that usually means five to seven days before you touch anything.
The reason you give a lead event more leeway is that low costs swing hard early. You can start a campaign at a twenty dollar cost per lead, have one good day, and watch it drop to five. If you cut it on day two at twenty dollars you just killed something that was about to work. When in doubt, sit it out.
Same logic on the ad level once you are past the adset. If an ad has spent three to five times your target cost per booked call and booked nothing, turn it off. If it spent that same amount and booked one call, I let it run to five times, because one or two more calls puts it right where I want it.
The early numbers lie. Give the platform enough spend and enough days to tell you the truth.