Auto Dial Level Threshold: when predictive kicks in

Predictive dialing is a numbers game. It works by averaging how your recent calls behaved and betting that the next batch will look similar. With twenty agents that bet is reliable. With two agents, a single bad guess sends a live caller to nobody. The Auto Dial Level Threshold field exists to stop the dialer from making confident predictions when there simply are not enough agents to predict against.

What it controls

You set a minimum number of agents. As long as the campaign has at least that many, the predictive algorithm runs normally and the dial level floats up and down to keep agents busy. The moment the count drops below your number, the dialer forces the level back to 1.0, one line per agent, and stays there until enough agents come back. It is a floor on how small a crowd the predictor is willing to work with.

The field only has meaning under an ADAPT or RATIO dial method. It also needs to be set to something other than DISABLED, which is its default, and the agent count in the pulldown must be above zero, or nothing changes. Out of the box, then, this brake is off and you have to deliberately switch it on.

Which agents get counted

There are three ways to count, and the choice matters more than the number itself:

• LOGGED-IN_AGENTS counts everyone signed into the campaign, even if they are on break or wrapping up.
• NON-PAUSED_AGENTS counts only agents who are waiting or actively talking, not paused.
• WAITING_AGENTS is the strictest, counting only agent seats free and ready to take a call right now.

If you pick LOGGED-IN_AGENTS, predictive can keep running even when most of your staff are paused, which can drive up drops. WAITING_AGENTS is the cautious choice because predictive only engages when you genuinely have spare capacity. For why predictive needs a crowd in the first place, see our .

When to lean on it

This setting earns its keep on small teams, on campaigns where agents drift in and out, and during the tail end of a shift when people start logging off. Without it, a campaign that started with ten agents and ended with two can keep predicting as if it still had ten, and your drop rate climbs at exactly the wrong moment. If that is already happening, our guide to covers this alongside the other end-of-shift gotchas.

Picture a campaign you set to require at least four waiting agents. For most of the afternoon you have eight or ten people ready, so predictive runs normally and the pace climbs to keep them all busy. Then lunch rotations or a string of long calls leave only three agents truly free. The dialer immediately drops the level to 1.0, one line per agent, so it stops outrunning the staff you actually have. As soon as a fourth agent frees up, predictive switches back on. You never had to watch the screen or flip a setting by hand.

There is a sibling field worth knowing about that follows the same logic. The Available Only Tally Threshold can flip the Available Only Tally setting on automatically once your agent count slips below a number you choose, giving you a second safety net layered under this one. Both share the same idea: when the room gets small, dial more carefully without anyone having to intervene.

If you would rather these guardrails be configured for you, the managed boxes on arrive with reasonable threshold defaults so a thinning room never quietly turns into a wall of dropped calls.