Developer Documentation Library > Interviewer - Server > UNICOM Intelligence Interviewer architecture > Interview Service tier > Dialer Interface component > Call log analysis for the predictive dialing algorithms
 
Call log analysis for the predictive dialing algorithms
Before predictive dialing was implemented in UNICOM Intelligence Interviewer, numerous dialer call logs were examined to obtain an overview of typical outcomes for dialed calls in terms of the ratios of outcome (busy, no answer, connected, and so on), how long before an outcome occurred, and how long a call remained connected. This section describes the results of those examinations and the conclusions that were drawn from them.
Log files used
Six log files were examined altogether. This section covers two of them.
Project A. The call logs relate to one project in a large residential survey conducted in the UK. The project had around 400000 calls.
Project B. The call logs relate to one project in a business survey in the USA. The project had around 100000 calls.
Analysis techniques
The log files were read using a DSC based on mrLogDsc and modified to read IBM SPSS Dialer call.log files. This allowed the log files to be queried using DMQuery and tabulated using Survey Reporter.
Results
The variables in the call logs were crosstabulated to try to establish patterns that might be useful during the development of a predictive dialing algorithm for UNICOM Intelligence Interviewer, and to investigate whether there were any typical call profiles that should be used when testing predictive dialing. The following sections analyze the call profiles that have the most effect in determining how many calls to over-dial:
Call outcomes
Ring time
Call duration
Dial time.
See also
Summary and conclusions
Dialer Interface component