Predictive Dialer

This section of our technical library presents information and documentation relating to Predictive Dialers and Auto Dialer software and products. The PACER and Wizard phone systems are PC based call center phone systems that are recognized as premier inbound and outbound computer telephony systems. Features such as automatic call distribution (ACD), interactive voice response (IVR) and call recording have added a new dimension to the predictive dialer and auto dialer capabilities of these systems. These computer based dialing systems can perform various types of auto dialing campaigns simultaneously. These types include Predictive Dialing, Progressive Dialing, Preview Dialing and Dial on Demand.

Executive Summary

Predictive dialers are automated systems for high frequency outbound calling that increase productivity in call centers, i.e. where most interactions with customers take place. Consequently, enterprises continue to invest in technologies that improve the customer contact and management process and optimize the call center’s performance.

Predictive dialers are available as turnkey, pre-packaged solutions, or can be custom developed using toolkits. Regardless of the method chosen, careful design consideration should be paid to ensure optimum performance.

This white paper provides an overview of the factors affecting the performance of a predictive dialing solution, including the connect rate generated, the ability to distinguish between an answering machine and a live person response, and the lead base penetration rates accomplished. The paper also outlines hardware and software related issues that could optimize a dialer’s performance, and subsequently a call center’s productivity.

Introduction

According to the Aberdeen Group, 70% of customer interactions take place at the call center. With such a large percentage, it is no wonder that companies continue investing in technologies that facilitate the interaction with customers, even during the present economic downturn. Spending on contact centers is forecast to grow from $8.6 billion in 2002 to $10.9 billion by 2005, a 13% CAGR. Within the contact center industry, the market for predictive dialers is forecasted to grow from $717 million in 2002 to approximately $1billion in 2005 (source: Telecommunications Market Forecast 2002).

Predictive dialers are automated systems that call a list of numbers and turn the call over to an agent when a human responds. The use of predictive dialers increases productivity in call centers, since agents spend more time talking to people, rather than manually dialing numbers, waiting for calls to go through, or hanging up on busy signals and answering machines.

Enterprises looking to invest in predictive dialers for contact management purposes can either select a readily available turnkey solution, or create a custom application using a development toolkit. Regardless of the chosen method, optimum performance of the predictive dialer requires careful design upfront. This paper outlines the factors that affect the performance of a predictive dialing solution, and highlights some of the design issues related to those factors.

Connect Rate

The connect rate is defined as the total number of calls connected to an agent divided by the total number of calls dialed. The connect rate is affected by several factors, including call progress analysis (tone detection) and call transfer time. The ability of a predictive dialer to correctly identify the call result it receives affects the number of appropriate calls directed to an agent, and ultimately increases the agent’s chances of closing a sale.

A high quality predictive dialer will accurately detect whether a live person has answered the call or whether it encountered a busy signal, operator intercept or an answering machine. When an answering machine or a busy signal is detected, the number called is kept in the database for later contact. However, when an operator intercepts the call or a Special Information Tone (SIT) is detected, the number is discarded from the database, not to be called again. SIT tones are three, precise, sequential tones returned to a dialer when a connection cannot be made with a telephone number. The most common SIT tone is the vacant number intercept SIT (i.e.: “The number you have dialed is not in service”).

This tone detection is performed through call progress analysis at the hardware level. Voice cards recognize call progress tones based on frequencies and cadence (pattern of alternating silence and speech), which are pre-determined and stored in a defined tone table. During call progress analysis, the voice card detects the frequencies and cadence and compares them to the patterns stored in the table to identify their meaning.

Since tone detection is hardware based, the selection of a telephony card can affect the dialer’s performance. Ensure that the card you choose or the card that ships with the prepackaged solution has a utility that allows you to customize the call progress analysis parameters, i.e. the tones and the frequencies to be expected under each call progress condition such as busy, fast busy, disconnect, no answer, and so on. Telephony cards also differ in the number of seconds or cycles of tone pattern they ‘listen’ to before reporting the presence of a specific tone. The longer the card ‘listens’ to the tones, the more accurate the detection. However, you should make sure that your card strikes a balance between call progress analysis time frame and accuracy.

Some of the functionality provided by the telephony card – in particular answering machine and fax detection – can also be performed by software, which disables the card’s tone detection capability and performs the call progress analysis directly. Such software-based solutions can be more efficient since the software analyzes and processes the received information faster than a hardware card.

Agent Transfer Time

If the predictive dialer detects a human response, the transfer to an agent should be transparent to the receiver of the call. This means that there shouldn’t be any delays or silence periods between the instant the call has been answered by a prospect to the time the agent has responded. This not only maintains a high quality of service and minimizes nuisance calls for the receiver, but in some cases it is a matter of abiding by the law. For example, a new California law instituted in July 2002 states that a called party must receive a response from the calling agent or telemarketer within 2 seconds of the called party’s completed greeting (Hello, Hi, etc.).

Your predictive solution should thus minimize the transfer time as much as possible. The agent transfer time has a direct impact on your connect rate since many individual recipients quickly realize that they are receiving a telemarketing call when there is a noticeable pause or silence period after they answered the call. This usually results in hang-ups, and consequently limited list penetration.

Answering Machine Detection

Although the connect rate has a significant impact on campaign results, it is the appropriateness of the connection that will ultimately determine whether the transferred call will enhance an agent’s productivity. Calls answered by a live person, but incorrectly classified as answering machine responses – and thus not transferred - cause call abandonment and eliminate a potential sale from the calls transferred to an agent. (Call abandonment occurs when a recipient terminates a call after the dialer has determined that a live contact answered the call.) On the other hand, the incorrect transfer of an answering machine to an agent also reduces the productivity predictive dialers are designed to provide.

An effective dialer minimizes call abandonment rates by ensuring agent availability before it places a call. Predictive dialers – whether turnkey or custom developed - should therefore provide the ability to account for available resources and ensure that an agent is free to answer a call before a number is dialed.

To decide whether they are connected to an answering machine or a live person, and to potentially pass the call to an agent, predictive dialers usually need to hear a response from the call recipient. Usually, a short ‘Hello’ means that a live person has answered the call, whereas a lengthy response signifies an answering machine. Your predictive dialing application or voice card should enable you to measure the duration of the response. Based on this duration, your answering machine detection strategy will be determined. Software-based detection algorithms represent a faster alternative to hardware-based solutions. However, they sometimes require more configuration and testing efforts. Make sure that your software provides the flexibility to use both alternatives.

Lead-Base Penetration

Lead-base penetration is defined as the total number of completed calls, bad numbers, wrong numbers, sales and refusals divided by the total amount of leads available for calling. The degree of detection accuracy plays a part here as well.

The ability of a predictive dialer to correctly identify SIT tones or bad numbers affects the quality of your calling list. When the predictive dialer detects a Special Information Tone (SIT), the number called should be discarded from the database, not to be called again. If a predictive dialer categorizes SIT tones as ‘answering machines’ or ‘no answers’, the numbers are retained in the database, to be re-called later. Repeat calling to the same bad numbers prevents agents from calling other numbers in the database, limiting the lead-base penetration process. Moreover, it reduces the ability to maintain a ‘clean’ and useful database, and consequently hinders the performance of a contact center.

Another factor affecting a call center’s database is the accurate detection of fax machines. Commonly, predictive dialing applications discard numbers that have been recognized as fax machines. Valid telephone numbers may be deleted from your database because the predictive dialing application has wrongly identified them as fax machine responses, and vice versa – numbers that should have been detected as fax machines may be retained in the database because they have been classified as ‘no answer’ or ‘answering machine’. Make sure that your application employs an accurate algorithm and that it allows for the flexibility of using a hardware and/or software based solution.

Conclusion

Implemented correctly, predictive dialers are an important ingredient in call center productivity. However, the degree of actual productivity experienced by an implementation depends to a large degree on design decisions and capabilities of the underlying product platform. To achieve optimal performance, your predictive dialing solution should accomplish high connect rates, accurate answering and fax machine detection capabilities and high levels of lead base penetration, which are dependent on the accuracy of your solution’s tone detection and call transfer capabilities.