Predictive Dialling - defined
By Michael McKinlay, Managing Director Sytel Ltd
Predictive dialling and predictive dialling practices are going to get a lot of attention during the course of 1997. This note, which has been prompted by an article on predictive dialling published in TeleProfessional in February 1997, explains why.
2. Exactly What is Predictive Dialling?
In popular usage the term has two quite distinct meanings.
An integrated call center application that provides all the functions required to run a predictive campaign, including list management, screen-popping, call progress and telephony functions, and predictive algorithms to manage the dialling process
Historically assembling all these functions together into a single package has been a non-trivial task, not helped by the lack of CTI standards. The provision of such packages has been dominated until recently by just a few companies, all North American-based, who have developed proprietary technology and the experience to build integrated solutions for managing predictive campaigns.
The use of the term "predictive" to characterise these companies has been a natural development since the black magic associated with predictive algorithms is not easy to master, and any "predictive" vendor worth his salt has gone for the glory of being seen to have mastered this process, and thus the ability to deliver the productivity benefits that go with it.
Life in the late 90's has changed. All of the hardware and software components required to manage a predictive campaign are now widely available. And the market for call center solutions, including running predictive dialling campaigns, is now being addressed by many software houses and systems integrators, all seeking to choose best of breed components and integrate them to meet customer needs.
|Call progress monitoring||The aim of this activity is to sift out all non-connects, and in some cases any machine connects as well. This used to be the province of predictive vendors; these functions can now be undertaken by both the PSTN and the switch working together, supplemented by standard DSP technology, if required.|
|Scripting||Good scripting assists the agent to deliver a message to a called party, and can have a substantial impact on agent productivity. Any predictive dialler is likely to benefit from a good scripting system, but scripting is common to many applications, not just predictive dialling.|
|Screen popping is done concurrent with a connect being delivered to an agent's headset.||Nothing to do with predictive dialling. Should be possible with any well-engineered CTI-enabled application.|
|Recycling calls on events such as busies and no answers.||This functionality is usually a part of any good outbound call center solution; again nothing to do with predictive dialling per se.|
So what about the traditional predictive vendors - have they lost out? Not a bit of it. The smart ones are moving with the times and repositioning themselves as general purpose software houses and integrators, with a focus on the call center marketplace. A good example of this is Mosaix's acquisition of a workflow management company during 1996.
And with these changes has been a realisation that the heart of predictive dialling, i.e. the black magic of the algorithms, is itself just a software component, which can be selected on its merits, and integrated with any vendor's offering. And this is indeed the core of Sytel's business, the provision of such algorithms to any software house or systems integrator looking to add an outbound capability to their call center solution.
When you see predictive diallers being discussed in future, increasingly expect to see people referring to soft algorithms, rather than these and all the other components that surround them, and go to make up a full solution for managing outbound campaigns
A set of algorithms for controlling the timing and the pace of dialling for outbound campaigns
As a minimum a good set of algorithms should take account of the following:-
the number of predictive agents on a campaign
the distribution of talk and wrap times both by agent and for the campaign
the distribution of call outcomes (live calls, no answers etc)
ring and detection times for all call outcomes
You will sometimes find products being called predictive, when they are considering only some of these events, and in consequence are limited in terms of their predictive quality. A clear example of this is where the algorithms attempt to predict dialling requirements, based on consideration of talk and wrap distributions only. The scope for good prediction, on this basis, in a modern Telecoms environment, with fast call set-up and response, is very limited, and so such algorithms can offer little productivity gain to their users beyond what auto preview (i.e. one line per agent) systems offer.
Such talk/wrap algorithms, because of their limitations, are not usually called predictive, unless marketing hype is on the loose. Typical names for them are progressive, and predial. Sometimes also (wrongly) called power dialling.
A particular feature of predictive dialling, as well as of some other dialling methods, is that in the bid to drive wait times for agents down, a predictive dialler will dial more trunks than there are agents available, or likely to be available to take a live call. And this means that some inevitably some calls will always need to be abandoned by the dialler.
The aim of responsible predictive diallers is to set sensible dialling rules, including low targets for call abandonment, and given them, seek to minimise the amount of time agents have to wait between calls.
3. What Rules Should Diallers Observe?
Just as there is sometimes confusion in the market as to what predictive dialling is, there is a lot of confusion as to what rules, if any, apply to the way in which they can/should be used.
For example, in practice you will find that a predictive dialler from a given vendor might give 45 minutes talk time under particular campaign conditions for one user, whilst for another user, a similar campaign, using the same dialler might have talk time of say 55 minutes in the hour.
How can this be?
It's all down to the dialling rules which the user, sometimes unwittingly, happens to be using. In practice, differences in performance among predictive diallers probably owe more to the dialling rules employed than they do to the quality of the predictive algorithms
This is not a healthy state for the outbound industry, and is one reason why Sytel has taken a major initiative on dialling rules. An article on this subject, written by us has already been published in both Voice + Magazine in Europe and TeleProfessional International in the past month, and it will also be appearing in the US domestic edition of TeleProfessional. Those interested in seeing a WORD version of this article may obtain a copy by sending an email to firstname.lastname@example.org with "Dialling Article Request" in the header. Further information on the article and the opportunity to rerun the simulations on which the article is based can be had by going to http://www.sytelco.com/dialrules.html.
We believe that the article published on predictive dialling in the US domestic edition of TeleProfessional in February, which looks at one aspect of dialling rules, must be seen as an important contribution to the debate on dialling rules, which is now long overdue.
The TeleProfessional Group in the US intend to take the lead on getting together a number of leading vendors; the American Telemarketing Association, the US Direct Marketing Association, and also ourselves, in the US in the Autumn, to see if a consensus can be reached on a code of practice. We think that this is a bold move on their part, and to be applauded. As yet, an invitation list has not been drawn up, but we understand that it will include at least several of the (traditional) predictive vendors, as well as some of the leading call center software houses now beginning to offer soft predictive solutions.
We expect to be one of the main contributors to this debate, and will be making Oceanic, our workforce (and simulation) management package available to help those present understand the significance of particular dialling rules.
We referred, several paragraphs ago, to the article on dialling rules published by TeleProfessional this month. The vendor responsible for this article is a very well regarded software house, which has made significant contributions to the development of the call center market.
Nevertheless the practice advocated in their article of hanging up on dialled calls that are ringing, and not classifying them as abandoned is one that we think should be reconsidered, not least because it is easily abused by users looking to reduce their wait times between calls. In our own article on dialling rules we have this to say about this practice.
"Some diallers will dial lots of numbers, allocate connects to waiting agents, then when no more agents are either available or imminently available, will hang up on any calls still being dialled Sometimes known as "predictive hang-ups". This is an inefficient way of dialling, and is usually confined to diallers whose algorithms are unsatisfactory. It's got little to do with true predictive dialling, and any dialler doing it should classify such calls as abandoned."
This comment was not targeted at any particular vendor and should not be taken as referring to the vendor responsible the article in the current version of TeleProfessional, since we were unaware of their views, when our article was written.
Nevertheless we see no reason to change our view on this particular dialling rule, and will be seeking to persuade both this vendor and any other vendors not classifying such calls as abandoned, that it is in the best interests of the industry that such calls are indeed classified as abandoned. The feedback we have had from other key vendors shows a lot of support for this position.
4. In Conclusion
We are very pleased to see the debate on dialling rules now opened up, and look forward to moves towards a code of practice for predictive dialling now that the TeleProfessional Group have decided to take the lead on this issue.
In the meantime we would be pleased to hear from any readers of this note and of the other articles referred to, who may have particular views of their own to express. Any such views will be treated in confidence, unless otherwise requested.
Back to the Top of the Page or to Predictive Dialling Overview