Editor's note: Tim Macer is a U.K.-based independent specialist and adviser in the use of technology for survey research.

Most firms faced with an upturn in survey work that forces them to double the capacity of their CATI facility would be out talking to architects, realtors, and placing ads in the vacancies column of the local newspaper. But a growing number of firms using Voxco's Pronto system have found they can take such upswings in their stride, without needing to hire anyone, let alone build the extra booths for them to use. The only changes are to install extra telephone lines and connect the Pronto dialer - a hardware device that sits between Voxco's Interviewer CATI system and the facility's telephone lines. By automating call connection in a highly intelligent way, it enables interviewers to spend upwards of 55 minutes in the hour actually interviewing, instead of the more customary 25 to 35 minutes when the interviewer is in charge of the dialing.

Pronto is a predictive dialing (PD) solution that either connects directly with blocks of ISDN lines, or indirectly through a switch, if one is already in place. At the other end, it communicates with Voxco's CATI solution, though there are plans to create an open interface to allow Pronto to work with other CATI systems too.

While automated dialing systems can increase productivity by reducing the number of misdials and speeding up the connection process, PD goes that extra mile by working out in advance when it needs to get the next respondent on the line and how many actual calls are needed to reach anyone at home. If necessary, it will fire off several calls at once, anticipating that only one of these is likely to result in a successful connection to an actual person - which is why PD systems need more phone lines than actual interviewers. Voxco recommends a ratio of between 1.6 and 2 lines for each interviewer.

A few other firms also offer predictive or power dialing solutions, but what is unique about Voxco's is the way it integrates with the CATI system's call management capabilities, and in particular, the ease by which you can modify the rules and set your own game plan to optimize sample use. Pronto can accurately discriminate between busy lines, fax machines, lines carrying a service message of some kind, and phones simply not answered. In the last year, Voxco has successfully added detection for answering machines and voicemail. What you decide to do with each outcome is up to you, and can be built into your calling rules.

Increase in productivity

A typical Pronto user is political and consumer research specialists Cole Hargrave Snodgrass & Associates (CHSA) in Oklahoma City, Okla., with 35 interview workstations connected to 72 lines. Pat McFerron, director of survey research, describes the increase in productivity using Pronto as "tremendous."

"It has reduced my interviewer hours by as much as 50 percent on short surveys and even up to 30 percent on those exceeding 20 minutes," McFerron says. "It has also allowed me to expand my hours of operation as I can use a limited number of agents to interview during daytime hours, knowing that I will have the capability to re-call all those not reached later in the day."

A key measure of increased productivity is the time interviewers spend between calls. At CHSA, this is typically down to eight seconds, though this will vary a lot according to the kind of project. The improvements are also most dramatic where interviews are short, but long interviews where a high proportion of calls are abandoned early, such as after a few screener questions, also show striking productivity gains. Pronto will slice its way through RDD samples or poor quality samples containing a lot of disconnected numbers. On the other hand, the gains are less dramatic on high-incidence studies or lengthy interviews, and PD is virtually no help with B2B work, since the phone is generally answered promptly.

The Pronto prediction algorithm is optimized for market research interviews, and will take into account the variability of interview length over a study. If two or more studies are live at once, it will operate a different predictive model for each one.

It also gives you great control over what Voxco calls "drop" calls - the calls you have to abandon because more respondents were ready than there were interviewers available to take the calls. These nuisance calls - where someone answers the phone but the line is dead - are a common event in direct marketing, where PD is widely used. In research, we need to take a more respondent-friendly and responsible approach, and keep drop calls to a minimum. Again, Pronto and Interviewer between them allow you to handle these any way you like.

You can get Pronto to switch out of predictive dialing for an instant, to what is called "preview dial," where the number is presented to an interviewer who clicks a button to initiate the call. This is a handy feature, not least because you can route any calls that were previously dropped to a real interviewer - and real quick, if you want - something they tend to do at CHSA. As a final safety valve, if the drop rate rises above a pre-determined threshold, it will automatically step the whole survey back to preview dialing.

In an equally respondent-responsible way, McFerron routes any suspect calls such as answering machines to a live interviewer whenever Pronto is not 100 percent sure.

"This is the advantage of getting a market research provider and not just one for a sales system," McFerron says. "It takes a lot to get the balance right. I really like the flexibility of the callback module, where you can set up your own callback rules so you can get through the least amount of sample."

He gives the example of how he built recall rules to take into account church attendance in his state. "A lot people will just increase the sample to compensate, but this approach means you are not skewing the sample."

Pronto will also provide audio integration, either to play sounds such as advertising jingles down the line to the respondent, or to capture open-ended answers as voice files. Although this is not in use at CHSA, I have seen it demonstrated. Voice capture of verbatim responses will not only increase the quality and amount of data captured for open-ended questions, it will also provide another productivity boost, as the interview is not artificially slowed down by interviewers having to type in the answers.

Ease-of-use

One of the nicest and, at present, unique features of the telephony integration offered by Voxco is in the monitoring capabilities it affords, providing an audio feed to the supervisor's ear and at the same time a screen-pop of the CATI interview being completed, simply by selecting the interviewer from an on-screen floor plan of the telephone room.

This level of ease-of-use is a hallmark of the Voxco package. Most functions are carried out using Windows-based GUI tools, including survey script preparation. McFerron estimates it would take a new user under two days to learn the basics, and in three-and-a-half years he has only encountered two problems that required a call to support. The dialer has also proved to be very stable. The most serious problem CHSA experienced, which required a dialer update, was fixed in just 20 minutes over the phone.

Speaking to other users, there are relatively few problems with the system. At times, complex callback rules can slow the system down, and when problems occur, people have found the documentation to be somewhat lacking in good examples or step-by-step guidance on what to do. Another disappointment is the analysis suite StatXP, which is overdue a major upgrade: many Voxco users tend to use the excellent export facilities to analyze their data in other tab packages.

Cost and quality benefits

It is always encouraging to see an application of technology that brings not just cost benefits but quality ones too. One of the most surprising reactions to using Pronto comes from the interviewers. They love it, as they get to spend most of their time doing what they came to work to do - to interview. Without the constant stop-and-start, they find themselves on a roll, and the shift flies by. It is a puzzle to me that more people are not using this technology.