Predictive dialer on-premise: pacing, abandon rate and AMD
How progressive, predictive and ratio dialing work on your own server. Abandon-rate control, answering-machine detection, and why cloud dialers get expensive at volume.
Three outbound modes in plain language
- Progressive — roughly one live dial per free agent; safest abandon rate, lower agent util
- Predictive — dials ahead of free agents using recent answer statistics; higher util, needs abandon control
- Ratio — reserves an agent and fans out several parallel attempts; first human answer wins
Abandon rate is the compliance dial
Predictive overdial without a hard abandon target burns brand trust and can breach local telemarketing rules. A serious on-prem dialer exposes desired/allowed lost percentages, warm-up until enough attempts exist, and automatic fallback to progressive when abandon spikes.
Wivoxa’s pacing engine updates dial budget continuously and can exclude machine answers from abandon KPI when AMD is enabled on the queue.
AMD: do not park voicemail on agents
Without answering-machine detection, predictive dialers treat voicemail as “answered”. Agents hear beep-and-tone; abandon math looks better than reality; overdial gets more aggressive.
On-premise AMD (RTP analysis via sniffer, not a third-party cloud) drops machine legs before they reach the queue and keeps human connects in the pacing denominator.
Why run the dialer on your server
Cloud dialers often meter attempts or minutes. High-volume collections, insurance renewals or B2B outbound make that unpredictable. On-premise software plus your SIP trunk keeps cost tied to licence and carrier — not to how hard you dial today.
Wivoxa — self-hosted PBX and contact center