OPEN PREDICTIVE DIALING: By Design Or Via A Kludgy Workaround?


We are hearing more and more about the migration from proprietary to open predictive dialing solutions.  However, all too often, a crucial question remains unasked: Just how open is open?

Before taking a claim of openness at face value, determine whether you are looking at a system that is  intrinsically open (that is, open by design), or a quasi-open system that links together components originally intended for proprietary or standalone solutions. The information in this article will help you make the distinction.

Over the last decade or so, and very much in parallel with the PC revolution, call center has developed  rapidly. Various methods of automated outbound dialing have been implemented, starting with simple  modern-based speed or auto dialing (still a popular feature of PC-based contact management programs.  Today, automated outbound dialing has evolved into sophisticated, high capacity digital telephony platforms linked to computer database applications dialing numbers based on the results of complex algorithms changing in real-time.

One critical element which has affected the development of dialing systems is call progress analysis ability.  By implementing techniques to analyze the signals sent back in response to dialing a telephone number, the best systems can accurately determine the results of a telephone call before connecting the call to the agent.  In these system, agents will only be connected to live people, not busies, answering machines, bad numbers, etc. Standard signals can usually be detected with complete accuracy; however, answering machine detection is less perfect.

Although the rate of accuracy in detecting answering machines is excellent and is constantly improving, beware of claims which are too good to be true. In cases where specific claims of performance have been made to the customer by a company prior to sale, almost invariably the customer states that the numbers are not being met. This problem (which pertains not only to call progress analysis but to productivity enhancements resulting from the purchase of dialing technology in general) is due to the vast number of variables in the real world which remain unknown until actual customer trials are performed. 

In today's market there is a constant flow of vendors and products entering and exiting. The principals upon which these dialing products are based vary, and may or may not be stated correctly in the product literature. The hardware platforms these systems use are equally varied ranging from PCs with simple modems for the dialing to mainframes linked with expensive switches. The quality and performance of the systems however is not linked directly to price or complexity.

Traditionally, a predictive dialer has been thought of as just that, a predictive dialer, one complete product from one vendor that handles dialing and switching, list management, reporting, and a user interface (a scripting package, for example). Data from this dialer has to he uploaded to another computer system for any specific business need beyond the most elementary application. Otherwise, significant customization of the dialer must he performed by the vendor to meet the client's particular requirements.

We call any turnkey system that requires the vendor for modifications and/or service proprietary. The components from which the system is built may or may not be proprietary to the vendor, but the way in which these components are assembled (both hardware and software) is proprietary in almost every case, resulting in a system only the vendor has total control over.

Performance considered equal, any predictive dialing system requires at lease two hardware components and two software components: Telephony hardware and software (for the dialing and switching functions) and computer hardware and software (for data processing, meeting list management requirements, and running the user/agent interface application). Regardless of what type of solution is being implemented, the aforementioned elements or their functional equivalents must be present.

Questions not normally answered by proprietary system vendors regarding the basic elements are: who makes them, what can / can't they do, can they be modified, what is the future growth capability and/or salvage value. Because these questions rarely need to be answered (when you buy a car do you ask the dealer who manufactures the seats, wheels, or light bulbs?) vendors have incentive to provide the least costly components.

This is particularly true of vendors who sell turnkey systems because most customers are analyzing total (turnkey) system cost on a per agent basis, without comparing the value or long-term implications of the specific hardware or software architecture involved. Many vendors also have an interest in using "nonstandard" (less standard) components when building complete systems to increase the customer's dependence.

The most state-of-the-art dialing solutions available today and tomorrow will not include any vendor proprietary hardware. PCs and dumb switching systems are the lowest common denominators. Just as the LAN has replaced larger hosts throughout the corporate world, so too will it replace centralized processing in the call center. Low maintenance and acquisition costs along with robust availability of hardware and software has already made PCs the leader in small and medium-sized centers, with many larger centers adding or convening to PCs from terminals.

Today, sophisticated, customized, and reliable call center applications can be designed for PCs more easily and inexpensively than for any other platform. That covers the data-processing side.

On the telephony side, an analogous situation exists. The utilization of modular telephony hardware to build a telephone switch has become a reality. In most cases, these switches are built to handle very specific tasks such as predictive dialing. The use of modular components by other companies to build entirely different types of switches limits the research and development cost on the hardware side. In addition, the quality and usefulness of the hardware components are beyond those available in situations where the system vendor manufactures its own proprietary telephony hardware. The bottom line is that with today's open, component-based architecture, complete high-performance solutions can be achieved with very flexible and inexpensive hardware.

Open is one of the hottest buzzwords in computer telephony today. But what attributes make a system open?  If a product possesses any possibility of communicating or sharing data with a third-party product, this may prompt a vendor to tout its "open architecture." Thus, smart shoppers need to do more than study glossy marketing brochures. It is necessary to distinguish between systems that approach openness and systems that are truly open.

Recently, some systems that were primarily developed as standalone or proprietary solutions have begun to make strides toward opening up. Typical techniques to approximate openness in predictive dialing include the following:

Serial Port Links: The most typical way to introduce some openness is to base your predictive dialer on a multi-user computer system such as UNIX with dumb terminals or PCs running terminal emulation. The dialer is basically closed, but it allows the upload of numbers to call via a serial port link. As a next step in openness, data collected through the proprietary terminal application may be retrieved through the serial link, usually daily.

Hot Key Methodologies: Another feature which may be touted as being open but which is actually just a workaround for non-open systems is a hot key methodology for "integrating" the dialer with other
applications. An example of this is to have a terminal emulation program running in one window on a PC and a separate in-house application running in another window. This allows the basic record to pop up on the dialer application.

However, productivity is not improved unless the agent is able to access the application designed in-house which contains much more extensive information about the called party. The agent "hot-keys" from the predictive application after saying "Hello, Mr. Jones..." into the "real application." There may be a cut and paste of some kind which takes some record lookup data from the predictive screen and places it into a lookup field on the main application. Otherwise, the agent typically retypes a customer number or the like to bring up the record.

Combining A LAN And A Standard Database: Other so-called open features include using a LAN rather than a more proprietary networking architecture and using some type of standard database file format with published specifications. This permits access to the dialing system data files by third-party applications, and makes it less necessary to pay the dialer vendor for  customization of reports and other data-processing functions.

Truly Open Predictive Dialing
The only truly open dialing systems are those which permit real-time, low-level integration of third-party
applications with the predictive dialer. Any system which requires batch uploads/downloads of data, or a hot-key architecture of some type, is one which requires by nature duplication of hardware and software resources, as well as more complex and labor-consuming management. 

Low-level integration is provided through programming libraries, or application programming interfaces (APls), which are accessible using a broad range of hardware and software environments. Systems supporting popular programming languages like C or Visual Basic and supporting the most standard communications protocols like TCP/IP, NetBios, or IPX will prevail in this arena.

In most cases, what we have referred to as true openness will only be achieved in cases where a system has been designed to support open architecture from inception. This type of product is one which is analogous to MS Windows versus MSDOS. That is, this type of product is more than a major revision. Rather, it is a new product that has been designed to replace an obsolete predecessor.

This is a difficult step for any technology firm to take if a successful and profitable product already exists,
so often, too many elements of the obsolete architecture survive in the "new," "open" system. The most likely case is that the surviving components will be on the telephony side being masked by an improved data-processing system. This will place limitations on the system in the long run, rendering the "new" system obsolete more quickly than necessary.

Because predictive dialers incorporate several required hardware and software elements, there are an equal number of fronts to cover with respect to being open by design. Either of the two hardware elements can be based on modular (open) or proprietary components, while the controlling software will have varying degrees of openness (as per the preceding section). As a general rule, the more open the better, as far as your long-term investment is concerned, because if some aspect of your system does not work out, it can be replaced or repaired without total loss of investment on the system as a whole. The disadvantage to open systems is that you as the customer will have a higher degree of responsibility to make the system work. You can find vendors who will provide turnkey systems, but the typical proprietary arrangement will lock you in tightly.

There is no question that open systems are the trend in information systems. The historical example of the PC (which was open), in contrast to that of the Macintosh (which was not open), demonstrates how a system allowing outsiders to participate in developing complete solutions will prevail. Open systems for specialized telephony applications are just now becoming available. Keep your eyes open for the unique open architecture product that will plug in and fill the needs of your call center.

David M Friedman is vice president of Marketing and Sales at Calltrol Corp.  Calltrol is the developer of OAPDE, the Open Architecture Predictive Dialing Engine, designed as a truly open dialing component system, and a leading CTI software manufacturer serving the call center industry.


Reprinted from CTI(tm) magazine, Volume 1 Number 3, published by Technology Marketing Corporation, One Technology Plaza, Norwalk, CT 06854 USA. Copyright (c) 1997 Technology Marketing Corporation, all rights reserved. For information about annual subsrciptions, call 800-243-6002 or 203-852-6800 or visit the publication's Web site at