• Amcat 2000 -ETS Teleprofit
• Amcat Excelerator
• Amcat Medialynx  SQL and Access 2000
• Amcat CCS  - Call Center Solution
• ETS Sigmaworx
• OPC Spitfire/TouchStar
• Noble Systems
• InfoTel -
• Marketel
• Razor - VOIP Dialer
• DataTel

A predictive dialer is a computerized system that automatically dials batches of telephone numbers for connection to agents assigned to sales or other campaigns. Predictive dialers are widely used in call centers.

History

The autodialer preceded the predictive dialer. While the basic autodialer merely automatically dials telephone numbers for call center agents who are idle or waiting for a call, the predictive dialer uses a variety of algorithms to predict both the availability of agents and called party answers, adjusting the calling process to the number of agents it anticipates (or predicts) will be available when the calls it places are expected to be answered.

The predictive dialer monitors the answers to the calls it places, detecting how the calls it makes are answered. It discards unanswered calls, busy numbers, disconnected lines, answers from fax machines, answering machines and similar automated services, and only connects calls answered by people to waiting sales representatives. Thus, it frees agents from the task of manually dialing telephone numbers and subsequently listening to ringing tones, unanswered or unsuccessful calls.

A predictive dialer can dramatically increase the time an agent spends on communication rather than waiting: a 2002 survey indicated an increase in talk time from twenty minutes in the hour to almost fifty. The system is most suitable for low quality lists and large numbers of agents, however an unexpectedly high contact rate can overwhelm the system leading to call abandonment.

Call list data is the loaded to the dialer before commencing a campaign. The data is generally derived from a large database such as a telephone directory or similar listing from CRM software. Some predicitive dialers generate call lists and report call attempts. Unsuccessful calls are often analyzed to determine if the number called needs to be called back later or needs special treatment, such as a manual or autodialed call by an agent to listen to an answer machine message.

Predictive dialer systems are commonly used by telemarketing organizations involved in B2C (business to consumer) calling as it allows their sales representatives to have much more customer contact time. Predictive dialers may also be used by market survey companies and debt collection services who need to contact and personally speak to a lot of people by telephone. More commonly predictive dialers are now being used as a quick and easy way to automate all sorts of calls which would otherwise be made manually by a call center, such as welcome calls for new customers, customer service call backs, appointment confirmations / reminders, or even for the automation of large numbers of ad hoc calls that might need to take place (such as by a taxi company, or parcel delivery service etc).

They generally rely on the fact that if a person was to sit down and manually dial 1000 people, a large percentage of these calls will not result in contact with someone at the other end. Out of 1000 calls made, typically only about 25-35% would actually connect to a live person. Of the rest, a large number (often 40-60%) won't be answered at all, around 10% might be answering machines, faxes, modems or other electronic devices, around 5% of numbers would be busy and the rest will result in network errors, or be identified as invalid numbers. For call centers that need to make large numbers of outbound calls, this represents a large problem. Typically in manual dialing environments, a given agent will spend around 80% of their time listening to the phone ring waiting to talk to someone, or dealing with invalid numbers or answering machines and only about 20% of their time actually doing what they are really there to do. By using a predictive dialer to filter out these unproductive calls and to spare the agent from having to wait for the phone to be answered each time, call centers can reverse the situation. Agents can now spend on average around 80% [1] of their time talking to customers and only about 20% of their time waiting for the next call - a 300% increase in productivity.

How Does it Work?

The predictive dialer exhibits predictive behavior when its dialing algorithm produces more call attempts (dials) than the number of agents currently logged in and available to handle calls. The predictive dialing happens when the predictive dialer dials ahead of the agents becoming available or when the predictive dialer matches a forecast number of available agents with a forecast number of available called parties. The matching and dialing ahead perspectives provide the large increases in dial rates and agent productivity.

If a system has 100 agents working on it, the dialer will dial a number of calls sometimes crudely based on a phone line to agent ratio of 1.5:1 or 2:1. This means that for each available agent, the system will dial the phone numbers of two potential customers. As these calls are made to the telephone network the dialer will monitor each call and determine what the outcome of the call was. From 150 calls made, the system will immediately strip out any unproductive outcomes, such as busy calls (these are usually queued for automatic redial), no answers & invalid numbers. Some predictive dialers incorporate "answering machine detection", which tries to determine if a live person or answering machine picked up the phone. This is one cause of the typical delays that one may experience before being connected to an agent.

If not enough calls are made ahead, agents will sit idle, whereas if there are too many calls made and there are not enough agents to handle them, then the call is typically dropped. A sophisticated system will throttle calls more appropriately to deal with these situations.

The advanced predictive dialer determines and uses many operating characteristics that it learns during the calling campaign and adjust automatically to the behavior of an ongoing campaign. Examples of such statistics include call connection rates (both current and average for recent past days by hour of the day), average agent connection time, geographic location dialed, etc. It uses these statistics continually to make sophisticated predictions so as to minimize agent idle time while controlling occurrences of nuisance calls, which are answered calls without the immediate benefit of available agents. An advanced predictive dialer can readily maintain the ratio of nuisance calls to answered calls at less than a fraction of one percent while still dialing ahead. However, this level of performance may require a sufficiently large critical mass of agents. Conversely, it becomes increasingly difficult to maintain a high talk time percentage with a lower number of agents without increasing dropped calls.

Dropped Calls

Predictive dialing systems use algorithms to control the ratio of calls to agents. Because a dialer cannot know what proportion of its calls will connect until it has made them, it will alter its dialing rate depending on how many connections it manages to achieve. Occasionally the system will get more live parties on call attempts than there are agents available take those calls. Consequently, the dialer will disconnect or delay distribution of calls that cannot be distributed to an agent. This is known as a dropped call or a nuisance call. The called party hears only silence when the predictive dialer does not at least play a recorded message.  Compliant Predictive Dialers give you the option to set your abandonment rate.  The current legislation sets the acceptable call abandonment rate at 3%.  That means the Predictive Dialer will drop only 3, out of every 100 calls made.  Then the Predictive Algorithm automatically adjusts the flow of call to accomidate the campaigns progress, making sure  to stay within those parimiters.  This is why call centers have to be very careful about the Predictive Dialer they choose, and the proceedures the follow.

The experience for those who receive a predictive dialer call can be less satisfactory. There may be an appreciable period of non-response before a call is routed to a sales representative. This annoys people and also gives them a chance to abandon the call. If no sales representative is available for a successful call, it is often disconnected. In certain countries, this disconnection is a breach of regulatory codes, and most countries now regulate limits on the number of silent calls that a company makes within a certain time frame. In the UK and the USA, these dropped calls have caused concern and outrage amongst the public, and strict regulations now govern how these systems may be used, with threats of large fines for companies that abuse these systems.

Types of Predictive Dialers

Predictive dialers perform the same function but the architecture and delivery methods can vary greatly between manufacturers. There are several types of predictive dialing: Software, Hardware, Smart, and Hosted dialers.

In recent years, 'mixed' type predictive dialers have emerged. These predictive dialers are based on simpler hardware, such as voice modems, and more powerful software for answering machine detection and call progress detection. The biggest advantage of this type of predictive dialers is the substantial lower cost of ownership. Another form of predictive dialer has evolved called the 'smart' predictive dialer combining voice broadcasting and attendant phone agents. Hosted predictive dialing is a service provided by third party providers that connects calls via the internet to agents.

Software Based Predictive Dialers

Software only solutions use ISDN messaging, or a CTI link to provide call progress analysis for calls made. Software only dialers are often cheaper because they do not require expensive telephony components, but in some cases can offer less functionality than more traditional 'hard dialer' solutions, particularly when it comes to detecting answering machines (AMD) and integration with other 'voice' related functions (voice recording, IVR, speech recognition, text-to-speech etc.). Typically a software dialer is connected to an existing PBX system via the PBX CTI link. In most cases, expensive specialized 'call classification' cards are required in the PBX for call progress analysis and answering machine detection. Open Source dialers have proven themselves in the production world and enable call centers of all sizes to lower costs. Other advantages include customization of the software suites to meet the needs on an individual basis.

Pros:

* Low cost without consideration of sunk costs of a fully-provisioned PBX
* Flexible architecture works well in multi-site and distributed environments

Cons:

* A few older PBX's will not work with a soft dialer configuration
* Higher error rates in classification of calls (Fax, modem etc. have to be detected by the agent)
* Lower call processing capacity

Hardware Based Dialers

Hardware dialers use dedicated telephony switches to perform call progress analysis and answering machine detection. Those switches usually have two main types of connections, agent audio and external audio. The agent audio connections are usually simple T1/E1/ISDN etc. telephony spans which are connected directly to an existing PBX (although other connection types that do not require a PBX are available such as Analogue or VoIP connections). When an agent first logs in for the day, the dialer will place a call from the switch directly to the phone on the agent's desk. This open phone call between the agent and the dialer switch is then kept open for the duration of the session. The second type of connection is the external audio connection which is the connection that will be used to make outbound phone calls. These connections are typically ISDN/T1/E1 connections direct to the PSTN. When an outbound call is made and answered, the call is immediately joined to an already open agent audio connection of the agent selected to take the call.

Pros:

* Fewer telephony connections required (In hard dialers external audio connections can go directly to the PSTN)
* Dialer typically will not need upgrading in line with PBX/CTI etc.; since standard telephony connections are the only link between the PBX and the dialer, the dialer is less affected by software changes/versions
* Superior answering machine and call progress detection capability
* Remote agent capabilities - hard dialers can connect to any phone, anywhere on the PSTN. (Highly suitable for home working or remote sites)
* Independent of special CTI communication and proprietary PBX protocols
* Independent of PBX (Can work with a PBX or without benefit of any PBX or ACD)
* Higher capacity for handling calls (up to about 100,000 calls per hour)
* Faster call switching, as both ends of the call are already in place.

Cons:

* More expensive than a soft dialer when the enterprise already has a PBX with spare capacity for processing calls.
* Increased cost of ownership as expensive to upgrade and maintain.

Smart Predictive Dialers

Smart predictive dialers combine auto dialing with voice messaging and phone agents who are prepared to handle calls initiated by the dialer. Answering machines, busy signals, and no answer calls are processed in a manner similar to that of a normal predictive dialing system. However, when a 'live' answer is detected, the dialer plays an introductory recorded message, giving the call recipient the option to talk with an agent to complete the transaction. This message is a consistent greeting that identifies the caller, the nature of the call, and the option to speak with an agent. This process requires a more sophisticated predictive algorithm to ensure that a phone agent is available when the call recipient asks to speak with an agent.

Pros:

* Call lists are pre-qualified without the need for an agent
* Call lists are processed much more quickly and efficiently
* Fewer agents are required to handle far more calls
* Agents are only speaking with interested callers
* Call list processing is far less expensive

Cons:

* Overall list performance may be less due to call recipient resistance to recorded messages
* Certain U.S. states do not allow recorded messaging (unless prior business relationship has been established). This may limit the use of this technique to only certain types of business or consumer campaigns in certain geographical areas.
* Illegal in the UK unless you have received prior permission from the people you are calling

Hosted Predictive Dialers

Hosted predictive dialers (aka Virtual Predictive Dialers, Web-Enabled Predictive Dialers, VoIP Predictive Dialers) use the Software as a Service (SaaS) model to provide organizations and individuals with a predictive dialer capability. Typically, the only requirement for a firm to use a hosted predictive dialer system is a computer with an Internet connection and a telephone line for each agent.

Pros:

* No required investments in computer or telephone hardware
* No required investments in software or licenses
* Administration and support are handled by the service provider
* Links into the system are remote, enabling agents and supervisors to connect from any location
* Software updates and upgrades included.

Cons:

* Service is dependent on an internet connection; when the internet goes down, so does the service
* Providers using VoIP as their primary delivery method experience limited reliability and performance.