Posted by: Malcolm Jarvis

Confession time: I used to swear blind that predictive diallers were terrible for business. I’d tell people that they annoyed customers, burned through data and were prohibitively expensive to install and run. What you really wanted was a progressive dialler (like the one that I’d made) that worked through records one at a time, ensuring that as every customer answered the phone, an agent was guaranteed to be ready to speak to them.

That was a long time ago though, as 2011 saw the launch of our own predictive dialler, and I have no problem admitting that it’s much, much better than its progressive predecessor. In my defense, there were, and still are, a lot of really bad predictive diallers out there, and the damage they’ve done to our industry through wonky answer machine detection, silent calls and inoperable do not call lists is maddening. To paraphrase the columnist Bill Vaughan, to irritate potential customers is human, to irritate millions of customers in a short space of time takes a badly configured predictive dialler. Or something like that.

This article is all about why predictive diallers such as our own are so much more complex than the alternatives, and also gives you a flavour of how they go about increasing agent productivity by the impressive amounts claimed. As with most things IT-related, dialler technology is a field plagued by jargon, so we’ll try and get some of that cleared up along the way too.

Let’s start off with a quick roll call of the different types of non-predictive auto dialler and the pros and cons associated with each, then we’ll take a look at why predictive diallers stand apart in terms of effectiveness, complexity and regulations.


Non-Predictive Diallers


This is the most simple version of auto-dialling out there. It usually takes the form of a button or menu option that’s bolted on to an application that places a call to the selected customer when clicked. There needs to be a link configured between the application and the phone that it’s going to place the call using, and from experience this can be an arduous process when working with traditional office-based phone switches.

On the plus side, the user has full control of dialling, and it does save a few seconds per call as you don’t need to press the buttons to enter the number you’re calling on your phone, but for situations where you need to place lots of calls to lots of numbers, there are far more efficient options available. On top of that you still need to open each record one at a time and your agents will need to sit and listen to each call ring. This makes click-to-dial better than just dialling the number yourself, but not by much.


Preview dialling

Preview dialling is similar to click-to-dial, but includes the ability to provide a list of records to call and then have one or many agents systematically work through the list. Each record is presented in turn and and the agent is given the option to call the customer or move on to the next record without calling them. Like click-to-dial, the process of initiating the call is manual but preview diallers do save some time by managing the list of records on behalf of the agents. As agents have the option to skip dialling records if they don’t fancy phoning them, a degree of responsibility and trust is required in order for this to be an effective means of getting through a list of records.

Both click-to-dial and preview dialling are suitable for situations where agents need time to properly review the contents of the customer record prior to placing the call. The main downside of this is that the customer then doesn’t answer the phone, the time spent reviewing the record is effectively wasted, but this approach certainly does have its uses.


Progressive dialling

Progressive dialling builds on preview dialling by placing a call to the customer at the same time as the record is displayed on screen. As calls are placed one at a time there are still limitations on the amount of time agents actually spend talking on live calls as a lot of time is spent simply listening to the phone ring. Still, it’s certainly a lot more efficient than leaving agents to initiate calls themselves and there are none of the regulations of predictive dialling to worry about.


Predictive Diallers

To properly maximise agent talk time and productivity, predictive diallers are by far the best option. They are subject to the most regulation, and it’s down to the end-user to ensure compliance, not the company providing the dialler, so you need to know what your obligations are and ensure that your system adheres to them.

The concept of predictive diallers is simple enough: during the course of a shift, the system should ensure that enough calls are constantly being placed so that as soon as an agent ends a call, there’s another live call ready to connect to them. In a perfect world, agents would be constantly talking to customers throughout their shift, and talk time would be nigh on 100%.

The first snag is that up until recently, this was a very expensive system to run. Predictive diallers like the venerable Amcat were notoriously pricey and way out of reach of most businesses. Even once the dialler was installed, the monthly cost of just leasing the telephone lines, never mind paying the phone bills, was eye-watering. The advent of VoIP (voice over IP) coupled with the abundance of reliable, fast Internet connectivity has changed much of this, and cloud based diallers are much less restrictive and an order of magnitude less expensive than previous generations of dialler.


The predictive algorithm

The other snag is that predictive dialling requires your dialler to make some pretty accurate predictions, and telephone calls are anything but predictable. Let’s have a think about what can vary from one call to another:


  • How long does the call ring before it’s answered?
  • Does the call get answered at all or does it just keep ringing until the call was ended?
  • If answered, how long does the call last before it’s hung up?
  • After the call is hung up, how long does the agent wait before finalising the call (i.e how long are they in wrap up)?
  • Was the call answered by a real human being or an answer machine?
  • Was the number engaged or an invalid number?
  • Was there a delay before the call even started ringing?


Early diallers (and some that still lurk around today), didn’t really worry about any of this too much. They worked off a simple multiplier that the dialler manager could vary up and down depending on how productive agents were being. If the call centre was quiet, they could crank up the dialling rate to a 5x multiplier, and if lots of calls were answering they could lower it to, for example, a 2x rate. This would mean that 5 ready agents would cause 25 new calls to get initiated at the higher dialling rate, and 10 new calls at the lower rate.

The problem with this approach is that it fails to take into account the fact that circumstances can change suddenly, and that humans tend to get distracted with other things (such as food, or going to the loo, or irresponsibly going home for the night). If the dialler manager was to set the system to a high dialling rate mid-afternoon when the answer rate is low and then go home for their dinner, the system will blindly continue dialling at a 5x multiplier into the early evening when many more people are home and the answer rate is high.

This may sound like something from the dialler dark ages, but we received an enquiry in 2014 from a business owner who’d had exactly this happen to them with their existing dialler, and had come in the next day to a 90% abandoned call rate (current Ofcom regulations stipulate 3% is the absolute maximum)! Sensibly, they decided to look for a new dialler the next day.

At the heart of a predictive dialler is its dialling algorithm. This is the formula that it uses to periodically calculate how many new calls it should place to keep agents as busy as possible whilst hopefully ensuring it’s operating within the Ofcom rules. Some diallers have just the one algorithm, others give you a selection to choose from.

The purpose of the algorithm is to try and be a little more clever than a straight “ready agents multiplier” and use current and historic statistics to make more accurate predictions regarding:


  • how many agents will be available in the next few seconds;
  • how many calls that are currently ringing are likely to be answered by a real person, and;
  • how many new calls the dialler should place in order to get the available agents speaking to live customers


And it has to do all this without more live calls being answered than there are agents available to connect them to.


Predictive dialler regulations

Any time the dialler system has more live calls answered than there are agents available you have an abandoned call on your hands and the system has strict instructions from Ofcom as to what to do with them. Firstly it must play an “abandoned call message” within two seconds of the call being answered (the content of which is strictly guided), then no further calls can be placed to that number within 72 hours unless a live agent is guaranteed to be available to speak to the customer.

You’re allowed a maximum abandoned call rate of 3% in any 24 hour period, which needs to be observed across all call centres conducting the same campaign and also across all campaigns running in the same call centre.

There are also additional rules governing the use of answer machine detection (AMD) and rules dictating how individuals can remove themselves from further calls, not to mention the regulations concerning suppression of calls to numbers listed on the TPS register to customers who haven’t “opted in”.

There are a few other rules that Ofcom have put in place over the years to prevent companies abusing predictive dialling technology and these look set to become gradually more restrictive as the powers that be continue to crack down on those abusing the system. Even so, it’s still possible to operate within these parameters and create a predictive dialler system that leaves all other types of auto dialler in the dust.

As you can see, there’s a lot for a predictive dialler to take into account, and that’s before you start worrying about call backs, blended inbound calls, call transfers, and all the other facilities that your call centre solution needs to cover.


Predictive dialler benefits

The main benefit of a predictive dialler is in agent productivity. Let’s take the example of an outbound campaign where agents are using click-to-dial to make their way through a list of potential customers. In this scenario, you’d be lucky for each agent to get 90 minutes of time talking to customers out of 7 hours dialling. Conservatively, a predictive dialler can get closer to 5 hours talk time in a 7 hour period, although this will vary depending on the nature of your campaign. This means that if your monthly staff wage bill was £10,000, using a click-to-dial solution means that only £2,140 of your wage bill is being spent on agents actually talking to customers. The predictive solution increases this to £7,142, giving you more than 3 times the value for money from your investment in staff, not to mention tripling the number of opportunities your agents have to make a successful call.

There are other benefits too. While predictive diallers do get through more data, which is more expensive to source especially at the start of a campaign, data penetration is much better than other methods as the system should automatically manage retries to answer machines, no answers and engaged numbers as well as automating call backs and removing wrong and invalid numbers.

I’ve also seen a strong, positive effect on agent morale that comes from having the right tools to do the job. While agents will usually end up chatting to the person next to them if they have a long delay between calls (even longer if they need to click a button themselves), in general if they’re given more opportunities to talk to people they’ll be more successful and enjoy their work more. I particularly enjoy watching call centres where there are 40 or more agents on the one campaign as time between calls is usually around 4 seconds and the buzz is electric!


So, to recap. Predictive diallers are subject to a lot more regulation than other types of auto-dialler, and while your call centre software provider should take care of all that for you, it’s important to know the right questions to ask as it’s your business that’s on the hook if the ICO comes knocking, not your systems provider.

Cloud based predictive dialler that use VoIP to place part of calls over the Internet are now much more affordable than old-style diallers. Predictive diallers do go through data faster than other types of dialler, but as they should be generating more successful calls, this shouldn’t be a problem provided your cash flow can accommodate this.

The main benefit is agent productivity, which will be considerably higher than using any of the alternative types of dialler available, giving you much better value for money from your wage bill than you’d otherwise realise.

If you’d like to learn more about our predictive dialler, this article will give you some more details to get stuck into. If you’d like to know more about how our dialler works in a fully blended environment and how it interacts with the rest of our call centre solution, then take a look at our features page or get in touch - we’d love to hear from you.

Copyright © Greenlight Innovation Site by Radiator Digital