Predictive dialers, commonly used in call centers, are computer systems that automatically dial several phone numbers simultaneously and then connect agents with those calls that are answered by a person. Hosted predictive dialers are predictive dialers that reside on servers and that are accessed through an internet connection. Predictive dialers, whether hosted or not, use a number of different algorithms to predict whether agents are available, and whether an actual person has answered. They then adjust the calling procedure to match the number of agents predicted to be available when a new batch of calls are predicted to be answered.
A hosted predictive dialer evaluates the answers to calls it places and then detects how the calls are answered. Calls that receive a busy signal, unanswered calls, fax pick-up, answering machines, or other non-human responses are discarded. Calls answered by people are handed off to waiting sales representatives. The reason predictive dialers are so popular in call centers is that they keep agents from manually dialing numbers, listening to ring tones, and dealing with calls answered by machines. This frees up a lot of time and allows call centres to make far more calls than they would otherwise.
Hosted predictive dialers can increase the proportion of time that an agent spends actually talking to another party. In fact, it more than doubles the amount of time agents spend in communication, according to a 2002 survey. Predictive dialer systems are suitable for the so-called low quality lists – lists where there are not expected to be a high proportion of answers by people. If a predictive dialer suddenly encounters a high number of legitimate human answers, the dialing system can be overwhelmed by the volume.
To ready a predictive dialer system to begin working, a call list is loaded into the dialer program. Call lists might be culled from a phone directory, or from Customer Relationship Management (CRM) software. Some predictive dialer call center software packages create call lists and reports from call attempts. Sometimes, dialer software can analyze unsuccessful calls to decide whether the number should be called again, or if it needs to be manually dialed in order for an agent to listen to the call recipient’s outgoing voice mail or answering machine message.
In addition to being used by telemarketing call centres that do business-to-consumer (B2C) calls, hosted predictive dialers are used by marketing survey companies and debt collection services. They are also increasingly being used to automate calls such as appointment confirmations and reminders and customer service call backs. One example would be an automated call from a pharmacy to let a customer know that his or her prescription was ready to be picked up. Another might be a mortgage dialer reminding a customer of an upcoming automatic bank draft for a mortgage payment.
Percentage-wise, only about 30% of calls reach an actual person. Out of the rest, around half aren’t answered at all, and about 10% are picked up by answering machines, modems, and fax machines. Around 5% result in busy signals, and the few remaining may turn out to be invalid numbers or may be involved in network errors. In the older manual dialing call centers, an agent will spend 80% of his or her time listening to a phone ring, dealing with disconnected numbers or answering machines, and 20% of their time talking to people. Predictive dialers filter out the unproductive calls, keeping the agent from spending so much time dialing and waiting.
Algorithms in hosted predictive dialer call centers cause a large number of call attempts to be made – larger than the number of agents available to handle calls. The predictive dialer call center software attempts to match the number of available agents with an estimated number of available calls where people answer their phones. The better the predictive dialer algorithm, the better the call center productivity will be.
In a typical call center using a hosted predictive dialer, the telephone auto dialer may make calls on an attempted call to agent ratio of anywhere from 1.5 to 1 to 2 to 1. In the latter, it would mean that for each agent available, the system may dial two numbers. As the calls are placed, the dialer monitors the calls and determine the outcome of the call. Calls receiving busy signals are re-queued, and non-answers and disconnected numbers are discarded. Some call center software attempts to determine whether an answering machine has picked up. This is why those receiving such calls may experience a delay of a second or two between saying “Hello” and hearing the voice of the call center agent.
The more intelligent predictive dialer call center software packages will try to dial the ideal number of people so that those where an actual person answers are immediately connected to call agents, while unsuccessful calls are discarded. Many predictive dialers “learn” patterns of answering behavior for a particular call campaign and adjusts its dialing rates accordingly. To do this, the call center software collects statistics on connection rates, connection time, the geographic region called, and other particulars. The call center software is able to take the statistics to fine tune its calling so as to minimize the time agents are idle without allowing the system to be overwhelmed with calls where no agent can be matched to an answered call.
The biggest flaw with predictive dialers perhaps is that there is often a pause of a second or two between the time the person answering says “Hello” and the actual connection with a live sales rep. Not only do many people find this annoying, they realize they have a chance to hang up before the sales agent even comes on the line. If a call is answered but no sales rep is available, the call is often disconnected by the predictive dialer call center software, and in some countries, disconnecting like this is against regulatory codes. Most countries limit the number of these so-called silent calls that a company can make in a given period of time and impose hefty fines on companies that make too many silent calls. For example, in the UK, 3% of the calls answered by people may be dropped, so hosted predictive dialers in the UK must monitor live calls versus silent calls closely to keep from going over the limit.
In addition to hosted predictive dialers, there are also software predictive dialers, hardware-based predictive dialers, and “smart” predictive dialers. Some predictive dialers combine call center software and hardware. Software-only dialers are cheapest because they do not require telephony hardware. Call center software dialers can manage functions like call recording and speech recognition. Generally, call center software dialers are connected to a private branch exchange (PBX) system. Open source dialers are another low cost option for call centers of all sizes.
The most recent call center software dialers are voice over internet protocol, or VoIP dialers. They can connect with VoIP-compatible PBX systems just like traditional software predictive dialers. They can also connect to VoIP services over the internet. Software-only predictive dialers are low cost, and their flexible architecture is easily adaptable to a number of environments. However, some of the older PBXs won’t work with call center software dialers, and older dialers aren’t as accurate in classifying calls as to whether they’re picked up by a fax machine, a modem, etc.
Hosted predictive dialers typically only require that a firm using the service have a computer with a high-speed internet connection plus a telephone line for each agent. There is no required investment in telephony hardware, and no required purchase of software or licenses. The dialing is handled by a third party, as are administration and support. Because the links into the hosted predictive dialer system are remote, agents can connect from any location. Any call center software updates are done by the predictive dialer host.
But there are some downsides to hosted predictive dialers too. When an internet connection goes down, the service can’t work. VoIP technology is still new enough that there can be performance and reliability issues. But for those who are good candidates for hosted predictive dialers, setting up is easy and can usually be done in less than 24 hours. The host will need the calling list (typically in CSV format), and the party using the service will have to have a USB headset and call center software. Often there is no setup fee and no long-term contract required. Charges are paid by the month and generally include unlimited dialing within the host country.
Computer requirements are not that onerous, either. Hosted predictive dialers are usually compatible with PCs, Macs, and Linux machines with at least 1 GHz of processor power and 256 MB of RAM. The internet connection must have a minimum of 100 kbs download and upload speed. The dialers can make on the order of 250 calls per hour on average. However, this number will be different depending on whether the call center is small or large. The number of phone numbers dialed at one time depends on how many leads are put into the dialer, what time of day it is, and how many no-answer or disconnect calls take place. The dialer constantly adjusts to calling conditions to optimize the ratio of contacts to calls.
Some hosted predictive dialers will vet call lists against national and state Do Not Call lists, and can display any valid phone number on the called party’s caller ID system (within legal limits – It is illegal in the U.S. to show misleading or false information on caller ID boxes.)
Telephone auto dialers have come a long way in recent years, and have come down in price with the advent of software-only dialers and hosted predictive dialers. Call centers do not have to invest in expensive telephony hardware in order to place large numbers of calls. With internet technology, virtual call centers are a reality, and dialer services are within reach of even very small companies. Predictive dialer systems can easily be tailored to the size of the call center, the number of virtual call center workers, and the number of calls typically made in a day. And it can usually be done with standard, off-the-shelf computer systems.
Mail this post