A proxy, not a measurement

Almost every pay-per-call agreement contains a billable duration: the number of seconds a call must last before it counts. It is treated as a quality control, and in a rough sense it is one. A caller who hangs up after eight seconds was not interested, or reached the wrong place, or was never real.

But duration does not measure intent. It measures how long somebody stayed on a line, which is influenced by the agent's opening, the hold time before a human answered, the clarity of the routing, and the caller's patience. A well-qualified caller routed into ninety seconds of hold music will frequently hang up before the threshold. A poorly qualified one held by a persistent agent will sail past it.

Why the default number is usually wrong

Sixty seconds became a convention without ever being derived from anything. In some verticals it is far too short, a mass tort qualification conversation has barely started at a minute. In others it is too long, because the qualifying facts are established in thirty seconds and everything after that is the sales conversation, which is the buyer's job rather than the publisher's.

Setting the threshold against the actual shape of the conversation in that vertical, with that buyer's script, produces a number both sides can defend. It takes one conversation at onboarding and removes most of the disputes that would otherwise arrive monthly.

What to pair it with

Duration alone is a blunt instrument. Paired with source-level reporting it becomes useful, because the question stops being whether a given call qualified and becomes whether a given source produces calls that consistently do.

That is the level at which quality problems are actually fixable. A source with a normal duration distribution and poor conversion has a targeting problem. A source clustering suspiciously just past the threshold has a different problem entirely, and it is worth catching early.