Parallel dialing places several calls at once and connects the rep to whoever answers first. That is the easy half. The half nobody writes down is what happens to the second person who answers, how often that happens, and the rule that puts a number on it.
A dialing round launches one attempt per line and costs a fixed number of seconds. When somebody answers, the rep stops dialing and talks. Those two facts are the whole model, and they pull against each other.
For n lines at connect rate p, with a dial cycle of c seconds and conversations lasting t seconds, an hour holds 3600 / (c + n x p x t) rounds, and each round yields n x p conversations. The talking term sits in the denominator, so every conversation you win takes time out of the half of the hour that does the dialing. Adding lines stops helping well before it stops costing.
A round is n independent trials at probability p, and the rep can take exactly one. The expected number of connects arriving with nobody free is n x p - (1 - (1 - p)^n): the expected connects, minus the chance that at least one of them arrives. It is exact rather than a rule of thumb, and it climbs faster than throughput does. That single expression is the argument for running fewer lines than a vendor will sell you.
This is the part the category leaves out. Every explainer describes the collision from the rep's side, where a line simply drops. From the other side, somebody stopped what they were doing, said hello, and got silence or a click. They do not know a dialer was involved. They know your number called and nobody was there, which is what the rule in the next section exists to limit.
US federal telemarketing rules set an explicit ceiling on abandoned calls and an explicit definition of what counts as one. Both are worth reading in the original, because the definition is stricter than most people assume.
“Abandon more than three percent of all telemarketing calls that are answered live by a person, as measured over a 30-day period for a single calling campaign.”
“A call is ‘abandoned’ if it is not connected to a live sales representative within two (2) seconds of the called person’s completed greeting.”
Two seconds after the greeting, not two rings. Put that beside the collision formula above and the line-count decision stops being a matter of taste. Whether this particular rule governs your campaigns depends on who you are calling, on what basis, and in which jurisdictions, and that is a question for your counsel rather than for a vendor page. We are pointing at the rule, not advising you on it.
A power dialer places one call at a time from a list, automatically, as soon as the last one ends. A parallel dialer places several at once and connects you to whoever answers first. A predictive dialer places calls ahead of your reps based on a forecast of when one will be free. Parallel and predictive share the same failure mode: both can produce a connect with nobody available to take it.
It depends entirely on your connect rate, which is why a rule of thumb is worthless here. Run your own numbers through the calculator: the collision share is computed exactly, and you will usually find it rises faster than the conversation count does, so the useful line count is lower than the maximum on offer.
Under 47 CFR 64.1200(a)(7) a call is abandoned if it is not connected to a live representative within two seconds of the person's completed greeting. A collision where nobody is free plainly meets that description. Whether the rule governs your particular campaigns is a separate question that turns on who you are calling and from where, and it is one for your counsel.
No, and the arithmetic says why. Conversations take time out of the same hour the dialing has to happen in, so throughput flattens. Doubling the lines does not double the conversations; it does roughly double the collisions.
Four, which is deliberately at the lower end of this category. The collision formula above is the reason. We would rather explain the ceiling than sell you a bigger number.