Newsletter Sixteen

Dear Reader,

Welcome to Issue 16 of Outbound Focus, a free email publication of Sytel Limited.

Predictive dialing has been under some scrutiny in the US recently by two regulatory bodies:

In this issue, as promised, we comment on this legislative process, and make some predictions of our own.

Jamie Stewart

Predictive Dialing Started to Grow Up in the US in June 2002

In the first week of June, the FTC signaled its intention to have a closer look at the way dialers are being used in the US by including a session on the subject in its review of the Telemarketing Sales Rule (TSR). Then at the end of the month, we had the decision by the CPUC on how to implement the outbound legislation passed by the California legislature earlier in the year.

Let’s have a look in more detail, starting with California.

The CPUC Ruling

The agenda item which the CPUC commissioners voted unanimously to approve on 28 June 02 was this:

“Effective July 1, 2002, the allowable ‘error rate’ will be 3% of all predictive dialer calls answered by a live person. An “error” is defined as a call answered by a live person in which (1) the predictive dialer disconnects the call after the called party has answered, or (2) the called party does not receive a response from the calling agent or telemarketer within two seconds of the called party’s completed greeting, or, alternatively, no agent or telemarketer is available within four seconds of the called party’s telephone going off-hook. The four-second off-hook standard is a transitional one that will be phased out in six months. The “acceptable error rate” will be reduced to 1% on January 1, 2003, although this standard is subject to further review in an industry workshop that will be conducted by the Telecommunications Division within 90 days.”

At first sight this decision looks pretty swingeing from an industry viewpoint. Dialing at a 1% ‘error rate’ is a Holy Grail that most dialers will probably struggle to reach in any productive way. But there is some (at least temporary) relief given in two ways:

  1. Early Hang-ups
    There is no mention of what should be done in respect of early hang-ups. This is the notion, supported by some dialer vendors, that it is reasonable to hang up on calls that have been ringing for just a few seconds, without recording such calls as abandoned calls. The US Direct Marketing Association (DMA) explicitly rules against this practice in its Guidelines for Users of Predictive Auto Dialing Equipment.

    This is an issue we think was probably overlooked by the Californian legislators. The CPUC may have been aware of this practice but were powerless to rule against it, since their role is to interpret rather than fashion legislation. We think it likely that this practice will now increase substantially in California.
  2. ‘Dead Air’ Calls
    The CPUC has decided to legitimize ‘dead air’ calls, with up to two seconds holdup on calls, following the called party’s greeting. The main reason for the original bill was consumer concern over ‘dead air’ calls. Now, two seconds is not a lot, but will it tip the balance with a lot of consumers and send them scurrying for the Californian ‘do not call’ list (going live on 1 January 2003)? This is/will become a key issue in the FTC deliberations, too.

The FTC Telemarketing Sales Rule (TSR) Review

The right way to view the progress on predictive dialing that was made at the TSR Review in Washington was as a preliminary skirmish. The interested parties eyed each other up, showed a bit of fire, but kept most of their powder dry – while the judges(!) kept their own counsel.

Given the time allocated for discussion, it was probably unrealistic to expect a lot more. Here’s our reading on some of the key issues raised, and what we think you can anticipate from the FTC:

  • As some parties rightly pointed out in their submissions, some areas of activity, for example political campaigning, fall outside its remit. We are not sure that this matters. What does matter is that some body, probably Federal, comes up with a template for how predictive dialing should be done in the future, before all the States come up with their own rules, all different and in some cases (as in Kansas) likely to exacerbate rather than reduce the extent of non-agent calls. So if the FTC wants to take this mantle on and set a standard for other regulatory bodies to follow, our view is all strength to their arm.
  • In its initial review paper the FTC noted that technically the TSR prohibited any abandoned calls. Despite this, we absolutely do not believe that the FTC will seek to ban predictive dialing. Rather, as has happened in California, we expect them to seek a balance which recognizes consumer concerns, whilst allowing industry to still reap the gains that predictive dialing can bring. This could be by allowing some abandoned calls to be made, or could possibly follow the Californian model of allowing some calls, for which no agent is available, to be kept open rather than abandoned.
  • If abandoned calls are allowed, then expect them to be measured as a percentage of ‘live calls’ and not ‘all calls’. Even though the ‘all calls’ measure is still widely used in the US, this should surprise no one.
  • Unlike California, we also think it highly likely that the FTC will follow the lead of the DMA and place a restriction on the minimum ring time for calls.
  • We also think it likely that anyone doing predictive dialing will be obliged to provide Calling Line Identity (CLI) information to called parties, with perhaps a discretionary period to allow call centers to choose carrier links that will pass this information through.
  • The issue that we think is likely to take up most time in the FTC’s deliberations is the use of ‘dead air’ and answering machine detection (AMD) by dialers.

‘Dead Air’ Calls and Answering Machine Detection (AMD)

The heart of this issue is whether it is reasonable to use technology to decide whether the call that has been answered is a machine or a person. There are two distinct camps on the issue:

  1. In California, there was a strong push in a number of submissions to the CPUC that AMD was/is essential if any gains in agent productivity are to be realized from predictive dialing. Hence, the ‘dead air’ period of first four seconds, then two (from 01 Jan 2003), at the end of the greeting to allow AMD to do its stuff and then route live calls to a waiting agent.
  2. Then there is the view, widely held outside the US but less so within it, that the important thing is to have an agent respond to a called party at the time they say ‘hello’, effectively ruling out ‘dead air’ and the use of AMD.

We think that there is a lot of sympathy for this second view in the US, but there is a reluctance among many people to break ranks, and go against what has been seen as best industry practice.

One heavyweight did break ranks at the FTC Hearing: Art Conway, the President of Dial America, whose company is one of the top outbound bureaux in the US. For the full text of his views, see pp 19-22 of the Dial America paper (PDF format).

Dial America are suggesting that the right (only?) way to treat consumers is to have agents available to talk to them when they answer the phone. If they are right, but AMD gets legitimized by the FTC, as in California, then as we noted earlier, we may see a wholesale flight of consumers to ‘do not call’ lists, and once gone they may be difficult to win back.

We very much hope that all parties in this debate will be given an opportunity to argue their case in more detail with the FTC than happened at the June Hearing. And we hope that the proponents of AMD will make their case on how use of AMD, and the ‘dead air’ this involves, improves the quality and productivity of their outbound campaigns, whilst being kind to consumers and not sending them off in droves to ‘do not call’ lists.

The Case Against Answering Machine Detection (AMD)

Right now our sympathies, in the absence of reasoned arguments from the AMD lobby, are with Dial America. These are the issues stacking up against AMD:

  • Many consumers in the US are clearly aware that any ‘dead air’ on their phone is usually a sign that they are being predictively called, and they hang up before there is a chance to connect an agent to them. Our estimate is that this could amount to as much as 25% of all live calls on some campaigns. It could be argued that this is a good thing, since such people are unlikely to be receptive to being called and sold to, even if an agent is immediately available to talk to them, but we cannot think that anyone would condone this as good practice.
  • The argument that AMD is essential to predictive dialing productivity looks like sixth grade math gone wrong!

An Example

Let’s say that you are getting 45 minutes talk time on a campaign using AMD but you have two problems:

  1. some consumers hang up before you get to them
  2. as noted by Dial America in their paper, some consumers are dropped by the AMD, in mistake for being answering machines.

And you want to know what will happen if you get the agents to do the detection:

  • No more ‘dead air’ for anyone, and no one hanging up before an agent talks to them
  • No live calls lost, being mistaken for answering machines
  • If your agents are properly trained and monitored (it’s not hard!), then they should be able to drop an answering machine in no more than around 3-4 seconds. Multiply the number of answering machines each agent will get in an hour by this time in seconds, and subtract it from the talk time per agent hour you get with dialer detection – and that’s the talk time you can expect when agents are doing the detection. We will let you do the math. If you get it right(!) then with agent detection, you should see a reduction of around a couple of percentage points only in talk time per hour.

    If we believe the arguments made by companies such as Dial America who eschew AMD, then the improvement in the quality of the calls made, more than compensates for the reduction in agent talk time. And there is an added advantage for call centers seeking to maximize their reach; because they are not subjected to ‘dead air’, consumers dialed in this way are less likely to sign up to ‘do not call’ lists.

Someone told us recently that the above figure of 3-4 seconds should be more like 15 seconds. Just not so. The only figure that should matter for the calculation is the time spent by the agent detecting the call. If…

  • your system involves long latency times in transferring calls to agents
  • you are forcing agents to go into a drop down menu to select a disposition code to stop the call
  • you don’t monitor the agents to make sure that they aren’t going to sleep listening to answering machine messages

…then you have room for improvement! And don’t make the mistake of adding in call setup time and ring time too, (as well as the agent detection time), as one call center manager suggested to us a few weeks ago!

Non-Agent Calls in the US

Some research that we did of US-based readers of our newsletter earlier in the year, suggested that the real levels of non-agent calls are far higher than those talked about – anywhere. For example, a casual glance through some of the submissions to the FTC and the CPUC might suggest that levels of non-agent calls are running at a maximum of say 5%. Simply not so, as any beleaguered consumer in the US who can do the math will tell you. True levels of non-agent calls are many, many times this.

Our view is that if independent research is forthcoming about the real levels of non-agent calls (c.f. Art Conway at the FTC hearing when he suggested that abandoned calls were running at around 40% of live calls in the US [probably on the low side – Ed.]) then the FTC might be more inclined to go for a higher rate than the 1% ‘error rate’ California is currently going for – allowing better predictive dialing productivity than will otherwise be the case.

Will The Patient Recover?

Outbound (and especially predictive) dialing is in a state of transition at the moment in the US, but actual and prospective legislation on fronts such as CLI, ‘do not call’ and control of predictive dialing should go a long way to restore the patient back to life, albeit in a form that may be quite different from the market we see today.

Some of our actual and potential contributors have been scared off by the battering that outbound has been getting in the US. As the patient recovers – and we expect him to – we hope some of these commentators may get active again, in defense of responsible dialing.

We would welcome contributions from US-based people in particular who would like to give us their crystal ball view. All/any well-argued views welcome, especially on the AMD issue.


What is the Champagne Challenge?

Write for Outbound Focus, win a crate of Champagne!

How do I enter?

Click here to enter >>

All newsletters >>