Why I Hate Predictive Dialers

So, it’s come up a few times this week – I hate predictive dialers when used for collections.  Like, really hate them.  I say this with full knowledge of how they work.  I've been a collector on one, I’ve then years later built one, and I was also forced into a situation to manage one for 5 years.  I made it the best system I could with the time I had, and I still hated it at the end.

Allow me to explain my extreme dislike.

Hello?  Hello?

We all receive predictive dialer calls now – trying to sell us a cruise, a time share, a collection agency calling a wrong number, a message telling us how to vote, and so on.  The problem is, you pick up the phone, and say hello, and there’s dead air.

The issue is a predictive dialer, by it’s very nature, makes multiple calls per agent.  Most are set at a fixed rate of somewhere between 2-4 calls per agent simultaneously.  It leaves pre-recorded answering machine messages, it notes no answer calls, not in service numbers, and busy signals.  And it can call 200-400 calls per day per agent, rather than someone manually dialing 100 calls.  That seems great, right?

It’s great, until someone picks up.  Or more specifically, 45 people pick up.  Most folks using a predictive dialer crank it into overdrive, to maximize agent time spent talking to consumers – but there’s a lot of bad calls in that wake, people who pick up and can’t be transferred to an agent, because they are all engaged on a call, and there’s no one free.  This is called a dead or dropped call.  Many agencies run a 10-15% drop rate.  So if you call 150,000 people a day, and you get a 0.4% right party contact rate, that’s 600 consumers reached, and 60-90 of them are bad calls.  That’s not a good look, and that’s 60-90 aggravated people who won’t pick up again.


This Is Jessica Speaking, How May I Help You?

So you get a contact!  Hooray!  The call is routed to Jesscia.  Then the agency calls again, and it goes to Mary.  Consumer calls back, and now they get Brian.  A predictive dialer is inevitably a pooled environment, so you can’t build rapport, you can’t hold a consumer accountable to a professional collector with arrangements made.  Now you are relying on notes skimmed over in a few seconds while the call happens, your experienced collectors are hampered by less experienced collectors making follow up calls on promises or broken arrangements.

As a secondary and more insidious problem, a pooled environment creates less pride in workmanship, inside fighting over file ownership and payments coming in, and a generally heartless cubicle farm environment where you can only expect so much quality of work and expertise, because who wants to be plugged into a predictive dialer for the next 5-10 years at the same company?


Excuse Me, Did You Say 0.4% Contact Rate A Minute Ago?

Yep!  If you get an automated call from a number the odds of you picking up after the first call go down – the law of diminishing returns.  So you have 150,000 files and you call them once, great!  You will probably get a 4-5% contact rate (oh, except for those dropped calls we talked about), and then the second time through you’ll get a 1-2% contact rate, etc, etc.  Just because you go fishing with a machine gun doesn’t mean the fish will swim towards the bullets. 

Very loopy metaphors aside, we are human beings and we have patterns in behaviour.  It doesn’t matter how old we are, whether we are employed, or what kind of car we drive, we all for the most part have mobile phones now.  When we get a call and the call display doesn’t show our mom, our best friend, or our auto mechanic, we have a natural disinclination to answer the phone – and if that number calls back every day, or 2-3 times a day we get aggravated and answer the phone less.  So where’s the advantage in building a machine to ruin consumer engagement through battering away?


Can’t You Make A Dialer Better?

Sure you can.  I worked at an agency with a team of 25 collectors representing a client who sent us about 40,000 files a month.  I was told to build a predictive dialer, and I had to in order to keep up with that volume.  Over five years, I sat and made calls with my collectors to learn what would work best, and I lalso earned from my IT team how to code in SQL and write back end code to try to make the predictive dialer behave more like a human collector manually making calls.  Here’s what I did:

I changed the call ratio to fluctuate.  It was set more or less to 2.5 calls per agent not on the phone at the time.  So when all the agents were on calls, the pace of the dialer slowed down, and when they all got free it would speed up.  There was a lag, as calls are already in progress when agents are on the phone, but I got my drop call rate down to about 20 calls a day out of 140,000 calls.

I also wrote some back end code to set call backs at different rates.  Most predictive dialers have a giant reset button, and you can just call the whole list over again.  Or others you upload a new list to call each day.  I wrote code to call a file less – it would track the number of times each file had been called, ant at 1-7 times, it would call every 3 business days.  After that it would call less and less (because you could focus calls on newer files, but still occasionally tap at older ones, in a priority sorting).

I also set the code to call files based on behaviours.  File statused Promise Payment?  Okay, don’t call it for a week.  Call is Busy?  Okay try back in 10 minutes, but only 2-3 times in a day.  Call was no answer?  Call it back in exactly 18 hours, so an afternoon no answer became a morning call next time, etc.

Finally, I focused on inbound calls.  If someone got a message from our dialer, and they were calling in, I wanted to make sure that someone answered.  Everything about outbound calls became secondary to inbound …. Once a file called in, it was given special treatment by the coding to prioritize it over everything else.  And I measured the bell curve of inbound calls which started at 2:00 pm eastern time, maxed at 5:30pm, and trailed off to 10:00pm, and set the outbound call ratio to take a back seat to the wave of inbound.

That was five years of experimentation and about 1,200 lines of SQL coding that took 30 minutes to run every morning reorganizing over and over, about 3 million files.

We had 24 agents, a series of computer servers, enough T1 lines to allow 80 phone lines to run simultaneously for this team, and we paid our Telco about $11,000 a month back in 2006-2007 to run this.  The collectors squeaked out $5,500 in collection revenue per month, and after Telco costs it was a net $5000 per collector.  The client was happy, we outcollected our competitors, we made sure we were as professional as possible to the consumers we talked to … and it was only marginally profitable.  It was soulless work for our staff jacked into the system, and we were only as good as the least experienced member of the team.

There is something to be said about quality over quantity.


A Small Confession …

So I'm on a full rant here, but I want to come clean. 

My company's phone software has the capability to be a predictive dialer, and I still have all the coding I learned years ago.  I don’t use it … well, I almost don’t use it. 

I do sparingly use an IVR (Interactive Voice Response) calling files we have identified as Avoids Contact or No Contact.  To be more accurate, I have an IVR for each collection team member in our company.  So it makes maybe 20-60 outbound calls per collector per day on telephone numbers we have habitually got a no answer or answering machine, and never spoken to the person.  It calls the number once every 1-2 weeks.  We don’t leave a canned answering machine message, but if a human being picks up, it recognizes a voice and plays a message from the individual collector calling and offers the consumer to press “1” to be transferred to that collector – if they press 1, the collector’s direct line rings.  It costs us almost nothing, and generates us maybe 18 right party contacts a day.  And these are people we couldn’t contact by manually calling first … it’s not a primary method of communication, it’s just a supplementary tool for a small subset of our files.

So when I say I hate predictive dialers, it doesn’t mean I didn’t try to take a little bit of the code that made it work as best I could, and use it as a small supplementary tool.


Remember, The Phone Is The Least Used App On Your Phone

I could give you a study showing how many times you look at your phone in a day (average is 90), or how many people under the age of 50 no longer have a landline (approximately 75%), or talk about the dopamine released by your brain when  you respond to a text (it’s roughly the same dosage as a person having sex), but forget studies and statistics … how do you use your phone personally?  Right now?  While you are reading this blog?  If you like it, you’ll share it by email or Messenger or Twitter.  You will not call your friends to tell them about my amazing yet humble writing style. 

Let’s face it, I use my phone when I am driving to check in with my wife on what groceries we need this week, or to call my mother.  That’s it.  Maybe 15 calls a month.  I send about 500 Messenger messages a month, and about 100 SMS texts, and I don’t want to even think about the number of emails I send from my poor, battered Blackberry.

Think about being your target audience.  Communicate with them the way they want to be communicated with.  How do you want to be communicated with?


Welcome To The Future

The predictive dialer was cutting edge technology when I started in collections in the 80s.  It’s time is over.  It’s time to look at how we communicate as human beings outside our collection agency offices, and realize how consumers want to be communicated with.

Oh, and shut down your predictive dialer.  It’s not helping you, our industry, or the consumers we call.  There’s a better way to do it.

That’s my rant, if anyone wants to talk shop, or you are a collection manager trapped in a predictive dialer environment and you have no choice because a client is sending you 30,000 files a month, give me a shout, happy to share what I have learned and what you can do to make your environment a little bit better if possible.

Thanks kindly,

Blair DeMarco-Wettlaufer
KINGSTON Data & Credit
Cambridge, Ontario
226-946-1730
bwettlaufer@kingstondc.com