Predictive Dialing

Predictive versus Nonpredictive Dialing

From an operational standpoint, what distinguishes predictive dialing from nonpredictive dialing is the elimination of the 1:1 interviewer-to-call ratio. If one call is placed for one available interviewer, then the interviewer will be idle while the call is placed and the signal returned, which could be for 30 seconds or more. Autodialing (mechanized dialing while still using a 1:1 ratio), along with automatic signal recognition for engaged (busy), fax, disconnected, and unanswered lines, will partially reduce nonproductive time and increase interviewers' talk time. Predictive dialers include these autodialing technologies but also allow for the probability that not all call results will require interviewer intervention, so typically there are a great many more calls being placed than the number of interviewers available to take them, with the dialer handling the calls that do not require interviewer involvement.

Predictive dialing algorithms utilize time-series or queuing methods, and the variables utilized vary by manufacturer: for example, interviewers in the queue, those nearing the end of a call, connect rates, average call length, and so on. Ideally, all predictive dialing systems strive to make the number of connects equal to the number of interviewers available. Variation from this ideal has two possible results: first, there are more connects than interviewers, leading to either abandoned calls or “dead air”—and ultimately alienated respondents. Second, there are too few connections, which results in a decrease in a call center's productivity and an increase in its costs. The extent to which the predictive algorithm errs on the side of “wait time” versus increased “abandonment” (i.e. dropping a call that the dialer detects has been answered by a human since no interviewer is available) is determined by setting the maximum abandonment rate parameter on the dialer. At its most conservative setting (i.e. “zero abandonment”), the dialing ratio will be 1:1 (in effect, removing the predictive element). Most survey companies operate with the abandonment rate set under 3% (meaning 3 out of every 100 connects will be abandoned). Of note, research has suggested that setting the abandonment rate too high in telephone surveys is counterproductive, as respondents who have had too many prior contact calls abandoned by a predictive dialer may become frustrated, which increases their likelihood to refuse when a “live” interviewer comes onto the line during a subsequent call attempt.

...