Cmputer-Assisted Dialing: What will it d fr yu? Dan J. Bernard, Marketing Systems Grup (MSG) 1. Backgrund Cmputer-Assisted dialing has been an increasingly integral part f many market research rganizatins telephne data cllectin peratins ver the last 10 years. While being adpted as a means f imprving interviewer prductivity, an unexpected byprduct has been quality imprvement. This paper will address the ecnmic and quality issues surrunding the use f cmputer-assisted dialing shwing the benefits t scial research. With the adptin f cmputer-assisted dialing by rganizatins such as NORC and ther large scial research agencies, it is rapidly becming accepted in scial science research circles. 2. Overview While the gal f this paper is t review prductivity and quality imprvements that can be realized by cmputer-assisted dialing, it will likely be useful t review hw a cmputer-assisted dialer wrks. Nt s much that there are peratinal miscnceptins it is cmmn that seasned researchers are quite unfamiliar with hw a dialer really wrks. What it sunds like. Hw it interfaces with a CATI system. What cntrls there are t eliminate respndent abuse. 3. Sample Management Mst rganizatins have determined that it is best t leave the sample management t the CATI system r special frnt end sftware develped t d s. Many telemarketing dialers have built-in scripting and number management capabilities, but generally these capabilities are wefully inadequate fr the survey researcher s requirements. Generally, a telephne number r grup f telephne numbers is sent t the dialer fr dialing. Once dialed, a message is sent back t the CATI r sample management system frm the dialer reprting the results f the dialing. In the case f a nn-cnnect such as a nn-wrking number r busy signal, the CATI system will dispsitin the sample recrd as such and send a new number t be dialed. All f this is transparent t the interviewer therefre the dialing f a secnd number ccurs very rapidly minimizing unprductive time between dialings. When the dialer detects a persn answering the phne, the vice is cnnected t the interviewer and a message is sent t the CATI system t display the intrductry paragraph. The measured accuracy f the autmated dispsitin f the sample is cnsidered a significant feature by ne large agency.
4. Methds f Dialing The different terms used fr cmputer-assisted dialing can be very cnfusing. Terms frequently emplyed are: aut, pwer, predictive, adaptive, super, prgressive, preview, etc. There are three fundamental methds f autmated r machine-based dialing: Aut telephne number is dialed by a dumb mdem Pwer dialer can detect and dispsitin certain dialing results such as nnwrking numbers Predictive dials mre than ne number per interviewer using sphisticated statistical algrithms and number knwledge base t deliver a live respndent mre quickly Preview is an ptin utilized in aut and pwer dialing which allws the interviewer t review data frm the sample recrd befre talking with the respndent. Hybrid is an ptin cmbining preview and predictive refined by NORC, Pulse Train Ltd. and MSG and named HATI sm. Hybrid-Assisted Telephne Interviewing. It cmbines the best f bth wrlds delivering the high respnse rates f preview dialing with the efficiency f predictive dialing. 4.1 Aut Dialer Dials ne telephne number under interviewer cntrl via mdem r black bx Accurate dialing f telephne number Dials number much mre quickly than manual dialing Apprximate prductivity gains f 3 t 5% N abandnment f calls N intelligent sensing f dialing result N ability t dial ahead f the interviewer
4.2 Pwer Dialer Dials ne telephne number per interviewer - can be under interviewer cntrl r paced by the prgrammer Builds n all aut-dialer features Can autmatically detect fax/mdem, ring n-answer, nn-wrking, busy signal. Prgrammatically exchanges messages with CATI system t dispsitin ld number and retrieve new number fr dialing Cnservative prductivity gains f 24-50% Less tangible benefits f imprved wrking envirnment N abandnment f calls N ability t dial ahead f the interviewer 4.3.1 Perfrmance Gains - Pwer 1000 23-minute; 94% incidence; Cmpletes per Hur 19% higher 950 23-minute; 6.5% incidence; CPH 53% higher 2000 10-minute; 95% incidence; CPH 68% higher Qualify and transfer t IVR; CPH 100% higher Analysis f 1,700,000 dialings: 31 secnds t cnnect Vs. 56 secnds manually; 43% Reductin Large cmpany yields verall 24% increase 22 minute; 8% incidence; CPH 96% higher 4.4 Predictive Dialer Dials telephne numbers in a rati greater than 1:1 Builds n all pwer dialer features Uses sphisticated statistical algrithms t calculate quantity f telephne numbers t dial Allws adjustment f call abandnment percentage Cnservative prductivity gains f 25% ver pwer dialing
Can cntribute t respndent abuse via call abandnment if nt prperly managed. A research sensitive dialer will allw cntrl f abandnment t be held t ne in 10,000 dialings. 4.4.1 Perfrmance Gains - Predictive Side-by-side cmparisns shw 25-50% imprvement ver pwer dialing with less than 5% abandnment With higher abandnment rates- claims run t 300% 4.5 Hybrid Dialing Cmbines predictive dialing with preview dialing If a number has never been cnnected t a ptential respndent, it will be dialed predictively. If the number has been previusly cnnected, it will be dialed in preview mde. This blended slutin maintains the high respnse rates f preview dialing, but intrduces the efficiencies f predictive dialing. 4.5.1 Perfrmance Gains Hybrid Dialing Prductivity imprvements up t 25% 5. What else can cmputer assisted dialing d fr yu? 5.1 Can replace need fr PBX: All call centers need t cnnect t the public switched telephne netwrk. Calls must be silently mnitred fr quality assurance purpses. Supervisrs must have the ability jin a call as required. Sme peratins require the ability t accept inbund calls. All these features generally assciated with a PBX can be perfrmed by many autmated dialers, therefre eliminating the need fr a PBX and thus saving capital investment n six-figure PBX s.
5.2 The perfect vehicle fr call center decentralizatin: Simple integratin with VIP PBX allwing remte interviewers t cnnect via VIP back t a central ffice and then ut t the PSTN (public switched telephne netwrk). Imprved Quality Cntrl Audi recrding intrduces rbust QC tlset. Live mnitring can riginate in any lcatin Exhaustive call data is available 5.3 Optinal mdules prvide additinal capabilities: Remte Audi Mnitring Clients, prject staff r quality cntrl staff can dial in frm remte lcatins t mnitr interviewers. Digital vice capture f pen ends allws fr mre exhaustive prbing by interviewer. When the respndent realizes interviewers are writing verbatim respnses, they tend t self-edit r abbreviate respnses. When the cmputer is recrding the respnse fr later cding - a richer, mre cmplete pen-end respnse is btained. Whle interview recrding The ability t audi recrd an interview can serve many purpses. On large agency sees it as significant tl in its quality cntrl and training effrts. Caching f the interviewer Dispute reslutin Questinnaire develpment. Integrate with IVR (Interactive Vice Respnse) Autmated inbund/utbund switching, Call blending, Autmated Call Distributin t Spanish speaking interviewers, fr example. 5.4 Interviewer & Prductivity Management Enfrces standardized call rules Eliminates dialing errrs Faster dialing means greater thrughput Dialing mdes can be assigned n a study by study and/r statin basis Real-time graphic and tabular reprting f interviewer prductivity
Full silent mnitring capabilities bth lcal and remte. 5.5 Facilities Management Real-time and histrical prductin reprting, by interviewer, study, shift, site, client, and date Scheduling mdule prvides infrmatin n number f interviewers and supervisrs, and thse briefed Lcal and remte mnitring capabilities Real-time analyses and reprting f truble n telephne lines 6. Hw Dialers Interface with CATI Fr example, the MSG system is a 20 slt, industrial strength Intel-PC with special telephny hardware by Dialgic E1, T1, ISDN, r CO lines plug int bards inserted int backplane Lines frm interviewing statins are punched nt demarc blck and crss cnnected t lines ging t statin bards Dialer is cnnected t CATI server via serial cnnectin r Ethernet using TCP/IP CATI system manages sample file 7. What is Heard by the Interviewer? Pwer Mde: Sme systems can be set t pass call prgress tnes t the interviewer r just the respndent vice n cnnects. The interviewer will usually hear ell The call will sund like a nrmal call t the respndent Predictive Mde: N call prgress tnes can be heard The interviewer will usually hear sme part f the hell The call shuld sund like a nrmal call t the respndent unless the call is abandned
8. Research Vs. Telemarketing Research has a limited sample frame. The telemarketing supply is cmparatively unlimited. A primary gal f research is a high respnse rate Researchers cannt affrd respndent abuse - n a prject OR industry basis Predictive dialing wrks best with mre peple, researchers ften have 5 t 10 peple wrking n a given prject 9. What is Research ding Different We knw mre abut a given telephne number than anyne in the cuntry We pay attentin t call histry We will predict the prbability f cnnectin rather than predict when an agent will be finished We are ffering predictive dialing with near zer abandnment Predicting prbability f cnnectin wrks with just a few interviewers We can dial numbers in fractinal ratis, e.g. 1:1.7 rather than 1:2 r 1:3 like sme telemarketing systems which frces high abandnment rates Traditinal predictive is an ptinal setting as is researchpredictive SM We have the flexibility t d it many ways: pwer, prbability f cnnectin, traditinal predictive Use f autmated answering machine detectin is discuraged. Studies have shwn that sme 20% f true respndent cntacts are misclassified as answering machines. 10. Beynd Prductivity The Imprved Envirnment Empirically, it is fund that new interviewers are successful mre quickly when using cmputer-assisted dialing. Less prductive interviewers find it easier t keep up with mre prductive interviewers. As such, sme rganizatins have cncluded that cmputer-assisted dialing: Imprves interviewer retentin by helping them be successful mre quickly Makes their jb easier, therefre mre desirable Prvides a discipline that isn t innate
Facilitates imprved jb satisfactin VP large agency: I much mre enjy running the phne shp. The tedius part f the jb is gne. Other tangential benefits f autdialing: Cnsistent applicatin f dialing technique. Autmatic and accurate call result dispsitin. Gives the supervisr time t d things ther than push fr prductivity Overall, these factrs help imprve prject quality. 11. Scial Research is Different It s ften heard that scial research is different especially when it cmes t length f interview therefre negating the impact f autdialing. Analysis f 1.7 millin dialings shws that 70% f interviews are cmpleted n the first cnnect. This allws scial research t benefit frm the gains f cmputerassisted dialing. Scial research call rules ften call fr duble and triple the number f dialings dne by market research. This actually gives the advantage t scial research f being able t make use f cmputer-assisted dialing. The mre dialing f telephne numbers yu d, the mre prductive it can be. The adptin f predictive dialing by NORC in their 500 seat call center has, in ne fell swp, legitimized the use f cmputer-assisted dialers in scial research. Tw ther majr scial science research rganizatins in the States are testing and cnsidering cmputer-assisted dialing. 12. Are Dialers Expensive? They are ne f the few things in this industry that can demnstrate a return n investment in under a year. Are yu having truble finding interviewers? Wuld yu like t imprve prject quality?
Cntact Infrmatin: Dan J. Bernard Vice President Marketing Systems Grup 5000 Central Park Drive, Suite 204 Lincln, NE 68504 USA +1-402-489-0000 dbernard@m-s-g.cm www.pr-t-s.cm