New Guidelines For
Predictive Dialer Use


This article was printed in Call Center Solutions magazine in April 1999.

The telemarketing guidelines for predictive diallers announced by the U.S. Direct Marketing Association (DMA) in January 1999 represent a major step forward for outbound activities in the U.S., and indeed the world, by sending a clear signal that high levels of nuisance calls will no longer be tolerated.

   1) Vendor Vs. User Responsibility. Perhaps the most significant aspect of the guidelines is that the responsibility for implementing them falls almost entirely on users, as opposed to vendors. This is not totally unexpected since, given the current lack of agreed standards, the alternative would be to oblige vendors to ship different versions of their software, depending on what dialling guidelines a user was expected to observe. For example, it is not just an issue of DMA members vs. non-DMA members in the U.S., but of different practices among countries to which vendors ship. Well, that is the argument for now.

   2) Nuisance Calls And Abandoned Calls. The heart of the guidelines are the stringent limits the DMA has set on nuisance calls. Vendors and users alike talk about both nuisance calls and abandoned calls as the things diallers do, in seeking performance improvements. They are not the same; abandoned calls being just a subset of nuisance calls. What the guidelines set out to do primarily, and admirably, is not just to limit abandoned calls, but to curtail nuisance practices that diallers have resorted to in the past, specifically what we refer to in the table as “predictive hang ups” and “call delays.” These are practices that diallers have indulged in to gain extra productivity, whilst avoiding having to hang up on a live party and declare an abandoned call.

Predictive Dialer



Predictive hang ups

(The dialler hangs up on a ringing call, before a reasonable time has been allowed for a person to answer.)

“Allow the predictive dialling system to ring at least four times or for 12 seconds before disconnecting.” Putting this dialling rule into a set of guidelines is a world first, and represents a major step forward in eliminating this practice.
If a person’s intent is to answer the phone, rather than let an answering machine click in, then this time is on the short side. But there is little debate required here, since independent tests can easily be run to determine exactly how people do answer the phone, and it is to be hoped that the DMA will do this in their first annual review of the guidelines.
Call delays

(Once a person has answered and the dialler is ready to transfer the call to an agent — after call progress analysis has been completed — the dialler keeps the line open hoping that an agent may become free to take the call, to save it from being abandoned.)

“If a live agent is unavailable to take any call generated by the dialler, abandon the call and release the line after not more than two seconds.” Again this is a major step forward for the DMA. Many diallers have allowed very long delays in the past, causing considerable nuisance and leading to many people hanging up before an agent becomes available to talk to them.
Although the wording may seem open to interpretation, the DMA has confirmed that the intention of this rule is that it should be measured from when the phone goes off hook, and not from any later point; for example, from when call progress analysis has been completed.
The DMA accepts that this rule may be tough for some dialler vendors and users to comply with when call progress analysis is being undertaken. Diallers often use specialist cards for this purpose (which is acceptable for faxes, modems and nonconnects, but answering machine detection can take time). Our expectation is that the limit may not be too rigorously enforced in the short term, for those who have difficulty in complying.
But the real issue being addressed is one of shifting focus to the consumer away from dialler efficiency. So the two-second maximum may be there to stay and enforcement may get tougher, with perhaps a reduction even occurring in a future version of the guidelines. This could well lead to a decline in the use of answering machine detection on telemarketing campaigns, which might not be a bad thing (see separate box).
Abandoned calls

(A live call is abandoned by the dialler because no agent is free to take the call.)
“Abandoned or ‘hang up’ calls should be kept as close to 0 per-cent as possible, and in no case should exceed 5 percent of answered calls per day in any campaign.” Given adherence to the other three guidelines described in this table, most diallers will have extreme difficulty coaxing material productivity benefits (i.e., beyond doing auto preview dialling — one trunk per agent) at anywhere close to 0 percent. Fact, not opinion!
For campaigns with large numbers of agents and high call throughput, there should not be that much of a problem, but for campaigns at the 5- to 20- agent level, “close to 0 percent” is a very tall order, if users are looking for significant gains from going predictive.
Measurement of abandoned calls

(Live calls abandoned by the dialler may be measured in several ways. The DMA guidelines set a single standard.)
Abandoned calls are measured as a percentage of… “Answered Calls — calls which are answered by a live consumer (not an answering machine).” Historically, different measurements have been used by vendors. In addition to the basis defined by the DMA, a measure based on ‘all calls’ has often been used. The difference between the two measurements is illustrated as follows:
Example: Assume for every 100 calls:
   50 percent are live,
   50 percent are busies, no answers, answering machines, etc.
An abandoned call target of 2 percent is set.
If the measurement basis is ‘live calls’ then one abandoned call is allowed per 100 calls launched by the dialler.
If the measurement basis is ‘all calls’ then two abandoned calls are allowed per 100 calls launched by the dialler.

Call Delays & “The Answering Machine Detection Dilemma”

There will be some concern about the limit put on call delays due to the time required to perform answering machine detection. Forget the claims you hear for high detection levels in several hundreds of milliseconds. The fact is that two seconds from the time the line goes off hook will be tough for some diallers performing answering machine detection, as opposed to other kinds of call-progress detection. Here are four reasons why users may want to forgo this kind of detection and hence be sure to comply with the call delay rule.
Predictive diallers are increasingly used for marketing campaigns, away from their traditional home base of debt collection, where concern for caller reaction was not always high. Today in the U.S., many consumers simply hang up when they
know a dialler is doing call progress analysis on them to determine whether or not they are an answering machine. Or their blood pressure rises if the answering machine detection is slow, and the agent is then in for a difficult call.
   - The agent misses the first “hello” and perhaps the second as well.
   - If the speed of detection is in-creased to avoid these problems, then it is pretty certain that some live calls will be mistaken for answering machines and dropped.
   - If the agent does the detection, then there is the scope to leave a personalized message.
For users worried about the impact on productivity, measured as minutes of talk time per agent hour, it is probably a lot less than you think.

3) Impact On Nuisance Calls And Productivity. Sytel recently conducted some research in the U.S. to determine the current level of nuisance calls. The sample size was small and hence open to significant variance. Nevertheless, our results suggested that for every 100 live outbound calls made to consumers in the U.S., there might be approximately 50 nuisance calls of all kinds. If this is true, then effective implementation of the guidelines in the U.S. could result in at least a tenfold reduction in nuisance calls. Given that predictive performance and the level of nuisance calls are clearly linked, how can it be that such a large reduction in nuisance calls would leave much room at all for achieving performance gains from predictive dialling? The response is:

i) There is a law of diminishing returns at work in predictive dialling, in that the more nuisance calls that are made (of all kinds), the less the additional benefit in terms of higher agent talk time per hour or lower wait times between calls.
ii) It is quite likely that the way some diallers are being used means that nuisance calls do little for productivity, meaning that a reduction in them may not be noticed.

Although probably unintended, the guidelines make dialler efficiency a real issue. So expect to see a lot more serious discussion in the future on what’s good (and bad!) in predictive dialler design.

   4) Future Impact. Outside the U.S., the U.K. led the way five years ago with its code of practice, and several other countries have followed suit (e.g., South Africa). Many countries have yet to address these issues, so don’t be surprised to see the U.S. guidelines providing a model for widespread adoption internationally. If and when that happens, expect the onus of compliance to fall squarely upon vendors. Otherwise, if users are free to make their own choices, then some, unwittingly or otherwise, may choose less rigorous standards than those recommended in the guidelines. The DMA has made a bold move, and the outbound market in the U.S. and elsewhere can only benefit if the beachhead of these guidelines is both sustained and, in due course, extended.

Michael McKinlay is the managing director of Sytel Limited, a U.K. company which specializes in outbound software and supplies its soft predictive algorithms to some of the world’s leading call center vendors. Sytel has campaigned consistently in all major markets for self-regulation of diallers and was a key adviser to the U.S. DMA in the formulation of its guidelines.


Return to the The top of the Page, PD Overview, Home Page, The Case Files or What's New.