Receivable/Accounts - Information for Credit and Collection Issues

Sunday, September 30, 2012

Building A Better Predictive Dialer

Ten Tips for Operations Managers

Earlier this week, I had an interesting discussion with a creditor about the strengths of a predictive dialer vs. a manual dialing process.  She was assigning her agencies thousands of files per month, and they were using a predictive dialer to (what I believed) to be mediocre results.

Now, I’ve always been a huge fan of a manual collection strategy, because it can create a sense of urgency, present a consistent professional image to the debtor, develop expertise in the staff, and generally receive higher liquidation than a dialer program.  However, a manual process requires a great deal of manpower, man-hours, and does not process a large number of files quickly.

In my time, I have led a project in building an Asterisk-based open-source code predictive dialer with extensive back-end SQL coding to make it tuned to a specific collection strategy.  The final product was a dialer program that made 120,000 calls per day, and along the way, I learned many things that I thought might be worth sharing.

1.       Your Dialer Is Not Intuitive

A predictive dialer is incredibly powerful.  However, it is not intuitive, and cannot make a myriad of decisions like a human being.  Unless programmed otherwise, it will call the same number over and over and over.  However, you can build a code or script that will allow behavioural changes to your dialer, whether it is call times, frequency of calls, the output message for answering machines, and so on.  Keep this code open-ended to change with the inventory, liquidation and staffing levels.

When we built our dialer, I set up a series of independent sub-campaigns with different behaviours, and routed files to those campaigns dependant on their call count and status of the last file.  By doing this, I was able to extract promise payments, call-backs, and other right party contacts from the answering machines, no answers, and disconnected numbers, and focus the predictive outbound campaign (and maximize the dialer team’s calls) on those files.  While your machine is not inherently smart, you can build in some decision-making code, and make the best use of your staff’s time.

2.       Do Not Put Your Dialer Into Overdrive

While we set our dialer to rotate our files on a reasonable 2-business day to 8-business day turnover, depending on the client, previous status of the call, and call count on the individual file, we continuously heard complaints from our clients of other agencies calling the same debtor multiple times in the day, every day.  Not only is this illegal in the majority of provinces and states, it is a colossal waste of bandwidth and your telecommunications budget.  Just because you CAN call 200,000 files in a day or call back the same file multiple times in the same day does not mean you should.

To control our outbound flow, we set up the dialer to make outbound calls at a ratio of 4-1 or 5-1 per collection agent, but only when those agents were available – the dialer ramped up and down as collectors were logged into calls.  In the late afternoon, when the inbound calls ramped up, our outbound traffic went down to almost nil, and we avoided dropped calls.  The agents averaged 20 seconds between calls with this strategy.

3.       Manage Your Traffic Carefully.

When we were making our calls, we generated in the neighbourhood of up to 6000 inbound calls per day – after analysis, these inbound calls proved to have a higher liquidation ratio than outbound (almost double), and were received by our team in a specific time window, rather than evenly spread across the day.  Be prepared to change your staffing plan to field your inbound calls, and minimize or eliminate any inbound dropped calls, as these could be far more important than keeping your predictive dialer ramped for outbound calls.

4.       Every Client Is Different

When we introduced a new client to the predictive dialer, we had a base-line set of coding to handle the files.  However, we often altered the code by client to deal with unique status codes required by the client, small balance vs. large balance files, higher turnover of high liquidation paper, and so on.

One US client required a “Validation of Debt” status for us to request proof of debt from the creditor and then mail it to the client.  We set up a specific dialer code for this that halted calls until the validation letter was sent.

5.       Be Able To Change Your Pace

If a client is suffering low liquidation on a manual campaign, it’s a simple matter to redirect staff to the inventory failing to perform – make sure your predictive dialer can do the same.  Build a high-priority or red-flag list that underperforming files can be moved to for greater focus, and can be measured closely until the expected liquidation is met.

6.       Be A Debtor.

One of the best things we did in building our dialer is put the management team into the dialing list to receive calls.  Have your dialer set up to call a handful of the manager’s cell phones every day – monitor what the calls sound like, how they are answered by your collection agents, your answering machine when you don’t pick up, and so on.  Measure not only your collection staff’s tone and quality of call, but the volume, quality of telephone reception, background noise, time to agent pick up, answering machine detection times, and so on.

7.       Train Your Staff To Understand The Machine.

Not only do you need to have staff who can follow all the collection laws, represent the creditor and collect the account, you need to have them understand how their predictive dialer behaves, so they can re-route problem files, properly status escalated calls, or understand why a file is called when it is called.  Have regular training sessions with your staff and go over the routing and results of your dialer program.

By involving the staff, and sharing reports, they will give management feedback and suggestions to further improve liquidation – our staff came up with a targeted response to promise payment files, and built a troubleshooting team around the collection of those accounts.

8.       Be Creative

We found, through experimentation, that setting up an escalation inbound group consisting of team leads and the shift supervisor not only allowed agents who were struggling with uncooperative debtors an inbound extension to re-route these calls (rather than manually asking a supervisor for help), we were also able to build a custom outbound campaign on refusals that was only manned by the team leads and supervisor.

9.       Sometimes A Human Being Is Better.

When dealing with calls to places of employment, repeated broken promises, or programs that are heavily skip-trace files, a dialer can only do so much.  Have a plan on how to pull files away from the dialer to a manual process if necessary.

Our predictive dialer team had some enterprising staff who noticed a number of promise to pays were not liquidating – they asked to form a team that would work two evenings a week, manually calling these failed promises.  Not only did these two collectors liquidate on these files, we were able to build the code of the predictive dialer to automatically route files that were promises over ten days without being worked, or high balances that required skip-tracing.  We moved these staff off the dialer to a manual program that each of them was able to generate 150%-200% of what a typical predictive staff member was able to do in a month.

10.   Sometimes a Machine Is Better

At my current office, I have built a similar predictive dialer campaign for a specific program for a client – their program is heavily skip-trace files with wrong numbers, but large volume.  When we first received their batch of files, after lettering them we performed a brief initial dialer campaign, and reported back to our collection database the results of those calls.  In turn, we were able to extract the files with human contact for an initial campaign, and immediately begin a trace program on the not in service numbers and wrong numbers, leaping past our client’s lifetime liquidation expectation in two short months.  It wasn’t really a predictive dialer campaign that collected the files, but it certainly was a supporting factor.


A predictive dialer, if used properly, is an incredibly powerful tool – but it is just one tool available to a collection agency or internal collections team, that can be refined and used in concert with other resources.

If you have any questions about Asterisk phone systems, SQL-based predictive dialers, or call campaign strategy, I would be pleased to offer what advice I have.  Feel free to directly contact me at Kingston Data and Credit.

Blair DeMarco-Wettlaufer
Kingston Data and Credit
Cambridge, Ontario