Predictive Dialling - what it really means
by Michael McKinlay Sytel Ltd
Who can tell us what the main benefit of predictive dialling is? A sea of hands goes up. No problem there.
Lets try another one. Hands up those of you who know what predictive dialling is, and how it really works? A few hands boldly up, but a striking number are at half mast, and there are quite a few no shows.
Does it matter? It certainly does. An industry/product starts to realise its true potential when vendors, users and other interested parties have figured out what is going on; not just what the benefits are, but how things work, and what the standards are.
By standards I simply mean the measures and practices which allow all interested parties to be presented with a level playing field, for purposes of making effective comparisons and assessments of performance.
Take a related technology. Imagine that you are buying a database to manage your calling lists. A key question is whether the software/hardware configurations being proposed are up to the job. One way of doing this is get vendors to benchmark their solutions against a common standard. Here's a quote taken from www.tpc.com , the home page of the Transactions Processing Performance Council
"The TPC is a non-profit corporation founded to define transaction processing and database benchmarks. TPC benchmarks can be run by any company based on published standards. The benchmarks are audited and reviewed by the TPC."
Of course all database vendors will go for differentiation, or even look to alter the shape of the playing field occasionally to suit their offerings, rather than just be tried by a brute ranking of numbers. But there is little doubt that this kind of standard has been effective in helping the database market to grow in size.
So what standards, if any, apply in the case of predictive diallers?
There is a wealth of legislation and codes of practice governing issues such as agent behaviour, who can be called, and cooling off periods. In the US much of this is covered by the The Telephone Consumer Protection Act, and there are similar rules in Europe, in those countries where outbound dialling is allowed. Some would even argue that the legislators have been too busy protecting consumers rights.
Rules governing dialler behavior
One area, surprisingly, which has been almost totally neglected by the legislators and rule makers is the rules governing how outbound diallers behave.
Research in a number of leading outbound countries including the UK, the USA, Canada and Australia shows that in only one are there any explicit rules at a national level. That country is the UK, and even there confusion reigns as to how the rules should be interpreted.
So what do these rules concern?
Essentially they are to do with dialler behaviour when numbers are ringing, or when people answer and there are no agents available to take a call. This is a feature primarily of power and predictive dialling. If you are not sure why this is, see the highlight on abandoned calls.
Predictive and power diallers typically dial more numbers than there are agents to take calls, on the basis that not all numbers dialled will get a live answer. In the case of a predictive dialler, the extent of over dialling will depend on the factors assessed by it, in its algorithms. The higher the level of over dialling, the lower the wait times for agents between calls. And because diallers are dealing in probabilities and not certainties, sometimes live calls will be made when no agents are waiting, leading to calls being abandoned (sometimes known as a nuisance calls) by the dialler.
These rules are as follows
1. Abandoned call limits:
This is the maximum percentage of calls that may be abandoned. Good practice in all major markets is now within the range of 1% to 5%. Policy among serious users is to set targets at the lower end of this range, but see next rule below.
2. Measurement of abandoned call rate:
There's little guidance for users on this, and it's not uncommon to find major users in the same market using quite different measures. Yet the way in which abandoned calls are measured can be just as important as call limits themselves. For example an abandoned target of 1% that is expressed as a percentage of all call outcomes, may actually generate more abandoned calls than a target of 4%, that is based on live calls only.
3. Abandoned call delay:
This is the time taken before a connect is abandoned by the dialler, when no agent is available to take the call. The UK Direct Marketing Association code of practice sets a maximum hang-up time of one second. Apart from this, there are no recognised standards anywhere.
Ever been rung up by a dialler that kept you waiting on the line for say 15 seconds, while it waited to see if another agent would become free? If so, your blood temperature probably rose a few degrees, and you'll have some sympathy with the limit set by the UK DMA.
4. Dialler hang-ups:
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.
The Rules in Context
If these rules are left to the user to interpret as he sees fit, lets look at what the performance implications might be in respect of the first three of these rules. In all cases the measure of performance being looked at is the agent talk time per hour on a campaign.
Graph 1. 7 out of 10 call attempts result in a live call
Depending on which measure of call abandonment, and which call delay you go for, the spread in the range of talk times per agent hour achievable, is 2¾ minutes per agent hour. What about at lower rates of live calls?
Graph 2. 2 out of 10 call attempts result in a live call
The spread now in agent talk times achievable is almost eight minutes.
Some diallers may allow longer call delays than those shown above. Added to that are the impact of any dialler hang-ups. If this practice is also considered, and hang-ups are not treated as abandoned calls, then in both of our calling scenarios, the same dialler, by virtue of the dialling rules being applied, can easily display variations in average agent talk time, in excess of ten minutes in the hour.
So How Does a User Decide..
If performance is an issue when it comes to choosing your predictive dialler, then how many users have been able to draw up a level playing field for these rules, that enabled them to make like for like comparisons? Ill hazard a guess, and say that they are in a definite minority.
Theres an equally important issue. Having acquired a dialler, how should these dialling rules be applied? Is it really fair to leave this issue to call centers to decide for themselves? The average call center manager doesnt want to be bothered with the fine nuances of dialling rules, he just wants to know whats permissible, and then to get the best performance from that situation, subject to the wishes of his client(s).
What About The People Being Dialled?
Sure they matter - most of all. If dialling rules are too zealously interpreted in the search for competitive advantage, then we alienate the very people upon whom the success of the outbound industry depends.
So Whither Dialling Rules?
A telemarketer flying in from Mars would be pretty impressed with the size of outbound telemarketing, especially in the US, and he would be encouraged by all the interest in it in Europe. But he'd probably look in wonder at the differing ways dialling rules are being interpreted in the market. If we asked his advice on whether dialling rules matter, here's a guess as to what he might say.
Sensible dialling rules do matter -
to call centre managers, so they know what to expect from their diallers and can get on with the business of managing their calling lists
to the owners of calling lists so that they know how their client base is being marketed to
to consumers so that they are receptive to not just business today, but business tomorrow as well.
And last but not least they matter to vendors as well, so that they don't have to waste their time tilting at windmills seeking competitive advantage, and can focus on real differentiators such as product quality, CTI capability and so on.
The results in the graphs were simulated using Oceanic, Sytel's workforce management product for outbound. Key assumptions in the simulations include: 20 agents on a five hour campaign, with breaks assumed to be outside this time; 35% of all live calls averaged 15 seconds only, and the other 65% averaged 2½ minutes; and calls other than live calls were split equally between answering machines and no answers. To get the complete range of assumptions used in the simulations, and to rerun them yourself, if you wish, go to http://www.sytelco.com then 'Call centre Info' and then 'dialling rules'. ( no direct access to the dialling rules)
Bio. Michael McKinlay is Managing Director of Sytel Limited, a U.K. company specialising in predictive dialling algorithms and workforce management software for the outbound market. He can be contacted on +44 1494 793200 or at firstname.lastname@example.org.
This is an article that is being published in a number of call center magazines around the world in the first half of 1997. The exact text published may differ slightly because of editorial style policy.
Back to Top of the Page or to Predictive Dialling Overview