Editor’s note: Joe Rafael is chairman of Opinion Access Corp., Long Island City, N.Y.
Most market researchers who have anything to do with telephone interviewing have heard of predictive dialers. Most are not sure exactly what they do. Some love them. Some hate them. Some think they are great. Some think they are terrible. Everyone thinks they are expensive. And, of course, every telephone call center that has them proudly mentions them in their advertising.
During the past 15 years, in various ways, predictive dialers have been part of my working life. Beginning in 1982, I lead a programming team that designed a market research CATI system. In 1986, based on client requests, we built an interface between our CATI system and predictive dialers.
Our marketing strategy was to be a turnkey vendor, which meant that we provided not only the CATI software but also the computer, the terminals, the cabling, training and other services needed to get the calling center up and running. In keeping with this strategy, we became resellers for three predictive dialer companies and when one of our CATI clients decided to add predictive dialers, they purchased them from us. During the years, we sold over 2,000 predictive dialer stations.
Last year, we sold the CATI software division of our company. Today, we operate a market research telephone call center in Long Island City, N.Y. Yes, we are a CATI center. Yes, we are a predictive dialer center. Yes, we love them. Yes, we think they are great. And yes, of course we proudly mention them in our advertising.
A primer on predictive dialers
Without predictive dialing, when an interviewer is ready for the next call, they dial the telephone number - either manually or with an autodialer. Many of these calls do not result in a live person answering the phone but are a busy, no-answer, disconnected, out of service, changed number or an answering machine. It is not unusual for an interviewer to spend the majority of their time dialing just to find a live person. And remember, a live person does not translate into a complete. It is just the first step - there is cooperation, callbacks and, of course, incidence to consider.
When using predictive dialing, interviewers do not dial the telephone - it is done behind the scenes. The CATI system "feeds" telephone numbers to the dialer. The predictive dialer knows how many interviewers are waiting for a call and what percentage of the recent dialings were live people. Based on these factors, it dials more than one telephone number for each available interviewer - the goal being that when interviewers are finished with a respondent, they will wait no more than 10-15 seconds before they are connected to a live person. When the call is finished, the CATI system informs the dialer that the interviewer is available for another call. The interviewer just waits until they hear the respondent say hello.
Myths - some true, some not true
Myth #1: The calls will come so fast that the interviewers won’t be able to keep up with the telephone calls. Also known as the "Lucy in the Candy Factory" syndrome.
FALSE. There is a perception that the interviewers will be so overwhelmed by the number of calls that they won’t even have enough time between calls to catch their breath. Ask any interviewer who has worked on both a manual and a predictive system and just about 100 percent will say they prefer the predictive system. The reasons are because with manual dialing they have to (and with predictive dialing they DON’T have to):
- dial the telephone hundreds of times a shift;
- listen to fax and modems screeching in their ear;
- listen for 30 boring seconds, which is what three rings take, before they can dispose of the call as a no-answer;
- listen to the endless telephone company tri-tone messages, which are those three tones followed by a message informing you that the number you called is out of service or disconnected or changed; (If you want an accurate call report, the interviewer has to listen to the tones and the recorded message before they can dispose of the call.)
In predictive mode, when a call has ended, the interviewer does nothing but wait until they are connected to the next live person. The amount of time they wait depends on many factors, but the range is between 10 and 15 seconds. If the interviewer wants to take a break, unlike Lucille Ball on the assembly line in the candy factory, all they need to do is hit one key on their computer keyboard and the dialer will stop connecting calls to their station.
Myth #2: They are very expensive.
TRUE. Depending on the dialer company and the number of stations, the cost is between $3,000 and $5,000 per interviewing station. This is in addition to the cost of the CATI system. Another cost is the additional telephone trunks. Unlike manual dialing where you need one telephone trunk for each interviewing station, predictive dialing requires more trunks - usually two for each interviewing station.
Myth #3: When a respondent answers the phone, it will take so long for an interviewer to be connected that the respondent will hang up.
Sometimes TRUE, sometimes FALSE. It’s 9 a.m. Saturday morning and your telephone rings at home. You pick up the phone but there is no one there. But, you know someone is there because you didn’t hear a hang-up click. Finally, a few seconds later, a voice says "Hello, may I speak with . . . ?"
Could a predictive dialer be responsible for this annoying silence? Yes, guilty - but with an explanation.
A good predictive dialer will transfer a live person to an available interviewer in about 1/20th of a second - the effect being that the respondent will hear no silence and the interviewer will hear enough of the respondent’s hello to know if they have reached a male, female, boy or girl. In the world of "you get what you pay for," some dialers just don’t transfer as fast and that is part of the reason for the delay. The other reason for a delay happens when a not-so-good predictive dialer tries to detect answering machines. The only way for a dialer to detect an answering machine is by the cadence of what the person answering the phone says.
Think about it: When you answer your phone at home, you say hello and then you wait for a response. But your answering machine probably says something like "Hi, you’ve reached us but we’re not home, yadda yadda yadda." The predictive dialer knows the call is either a live person or an answering machine, but which is it? The only way for the dialer to determine is for it to hold off transferring the call to an interviewer and continue listening for the cadence of what is said until it can determine whether it is a live person or an answering machine. After hello and nothing more, the dialer has enough information to know it is a live person and then begin the process of determining which interviewers are available and then transferring the call. A not-so-good predictive dialer which is also trying to detect answering machines is the cause of the many seconds of silence.
Myth #4: Predictive dialers don’t make business sense because they are only effective if the calls are very short. And most market research projects are 15 minutes or longer.
FALSE. In our industry, we refer to the length of the study as the number of minutes that one completed interview takes. However for each completed interview, there are many calls that don’t result in a complete -- and these calls are very short. In fact, it is not uncommon for 90 percent of the total calls dialed to be one minute or less. Consider the mythical 100 percent incidence project using RDD sample. For each complete, about six pieces of sample will be needed. At 50 percent incidence, about 12 pieces of sample will be needed. At 25 percent, 24 pieces of sample will be needed. At 10 percent, 60 pieces of sample will be needed. You get the idea.
Using 50 percent incidence as an example, of the 12 pieces of sample, about seven will be busy, no-answer, answering machine or out of service. Of the approximately five live human beings that answer the phone, three of them will refuse or tell the interviewer to call back at a more convenient time and two of them will cooperate and complete the screener. At 50 percent incidence, one will terminate and one will qualify and hopefully complete the 15-minute questionnaire.
Score: 12 calls, 11 short, 1 long.
This example does not reflect what happens as you make multiple attempts on the same piece of sample because of a busy, no-answer, answering machine or callback to try to find the right person in the household at home. Answering machines and no-answers on the first attempt tend be answering machines and no-answers on subsequent attempts which makes for even a larger percentage of short calls. One big advantage of predictive dialers and one that is rarely mentioned (for reasons that I don’t understand) is that you can better penetrate the sample without an increase in interviewing hours.
Myth #5: Predictive dialers change the results of the study.
FALSE. The key word is change. I have never heard anyone say that they produce incorrect results. Predictive dialers do not change the sample that is dialed. They help to dial it faster and, when used properly, allow for multiple attempts without a major increase in interviewing hours. So if it can penetrate the sample better, any change in results are attributable to the fact that the predictive dialer is helping to find the hard-to-reach respondent. Not such a terrible thing, but that’s only my opinion.
This statement about changed results reminds me of stories we have all heard about a study that has been done year after year after year. The client says that the company that has been doing the interviewing has done a terrible job - poor interviewing, always late, interviews don’t validate as well as they should, inconsistent data, etc. And, although they would like to change the interviewing company, they are afraid to because, you guessed it, it might change the results.
Myth #6: Since they dial more numbers than there are interviewers available, there will be times when a respondent answers the phone and there is no interviewer available.
TRUE. This is true and it is called an abandoned call. When this happens the dialer hangs up and the respondent will hear dead air when they answer the phone and nothing more. A good dialer keeps statistics on how many numbers have been needed for one live person in the past few minutes and dials accordingly. By doing this, the abandon rate can be as low as 2 percent of the dialings.
Love those dialers
I admit it: I love predictive dialers. Just like the traditional type of love, the reasons for being in love change as we mature.
As a software developer of a CATI system, I loved being at the leading edge of the emerging area of telephony integration. Watching our CATI software and the predictive dialer working together at our first few client sites was a thrill.
As we moved into the marketing of the CATI and predictive dialer combination, I loved the fact that the sale of the predictive dialer doubled or in some cases tripled the price that a client paid us.
Now, as a telephone interviewing center, I love the fact that the production gains we get from using predictive dialing allow us to offer high quality interviewing at competitive prices.