Share with your friends










Submit

Analytics Magazine

Calling the smart way

Big data analytics increase call center productivity and reduce unwanted phone calls.

Douglas A. Samuelson and Olcay YucekayaBy Douglas A. Samuelson and Olcay Yucekaya

If we know anything about outbound telephone call centers, it’s these two facts: 1) People profess to hate all the unwanted telephone calls, and 2) Call centers keep making the calls because many of them actually do produce sales, debt payments and votes in elections. Now new research and development has produced a better approach, increasing response while decreasing the number of calls.

This seemingly paradoxical improvement reflects a profound insight: Most call centers have focused on increasing the number of dialing attempts, while trying to limit calling when no representative will be available to talk to the answering party. In fact, doing this much more effectively revolutionized the call center industry back in the late 1980s. The method resulted in a major patent [1] and a Wagner Prize finalist entry [2], as well as great success for the company that developed it.

Recently, 2Contact, a call and contact center company in Haarlem, the Netherlands, realized that a different focus would work better – not more calls, but calls better timed to reach the right party when the person is most amenable to being approached. This means that the software, trade named Smart-Call ™, uses big data and current high-speed computing capabilities to schedule call attempts, based on past responses by the called person or similar people. Neither the data nor the high-speed computations were available in the 1980s, but that was then, and this is now.

Ironically, this was initially a tough sale within 2Contact. Olcay Yucekaya, the inventor of Smart-Call, says, “It may sound hard to believe, but I’ve had more failures and rejections than you can imagine. In the end, ignoring the rejections and always thinking in solutions resulted in the fact that we are the leading call center in this idea.”

In fact, many U.S. companies have tried and are trying to integrate better data about called parties into their dialing systems. Since most of these approaches are closely guarded trade secrets, companies that claim to do superior list management are reluctant to disclose few if any substantiating data. We surmise, however, that these newer methods still rely on a predictive dialing method like the 1986 original, possibly modified to take advantage of cloud computing and real-time massively parallel simulations [3]. Such systems can adapt more quickly and smoothly to changing conditions (there’s no steady state in outbound calling, so traditional queuing models don’t help – it’s an adaptive automatic control problem, and rapid smooth adaptation is critical), but they still lack the shift in focus of Smart-Call.

Better-timed calls reach the right party when the person is most amenable to being approached

Better-timed calls reach the right party when the person is most amenable to being approached.

In tests using closely matched pairs of lists, 2Contact has observed that the list using the new method had a substantially higher rate of sales – 31 percent more, in the most rigorous test tried. The test group had a 4 percent higher answer rate – that is, somewhat more people were willing simply to answer the phone and talk to the representative if they were called at a more convenient time. But the difference in sales was much larger, indicating that the effect of time of call was much more than just whether people would answer.

Another performance statistic is the proportion of records in a contact list eventually used in the campaign. Most campaigns have a preset limit of number of times a party will be called. If this number of calls does not result in a conversation, the record is just abandoned for the purpose of a particular campaign. Usage rates around 60 percent to 70 percent are common; with a well-targeted campaign using older methods, it is unusually good if as many as 75 percent of the records end up being used. In the matched groups test, the “standard” group has 69 percent usage – but the test group had 93 percent.

And what about unwanted calls? The new method substantially reduces the number of call attempts per hour per representative, and even the number of completed calls per hour per representative. In the matched groups test, the test group averaged 4.31 completed calls per representative per hour, versus 5.54 for the other group; but the test group had 3.54 effective calls per hour, that is, calls that result in either a buying decision or a no-buy decision, versus 3.34 for the other group. (Completed calls also include connections to people who do not stay on the phone long enough to have the intended conversation, plus call sequences that reach the set limit on number of call attempts.)

Not surprisingly, the representatives in the test group had longer talk times, averaging three minutes and 19 seconds as compared with two minutes and 59 seconds for the other group. Conversations with the right party, when the person is receptive, do tend to last longer.

Also quite noteworthy is the abandoned call rate, the proportion of call attempts that result in the system hanging up on the called party because the dialing went too fast and there was no representative available to talk when the called party answered. The new method cut this rate almost in half, from 1.9 percent of attempts for the other group to 1.0 percent for the test group. The higher proportions of answers and long connects with the new targeting approach makes the system more stable and enables less “aggressive” dialing to keep representatives busy. The test group had an average of 24.38 dialing attempts per representative per hour, versus 30.85 for the other group. In short, Smart-Call reduces the call center’s costs and the number of unwanted calls it generates.

Of greatest interest to the call center and its clients, of course, is the sales rate. In the matched groups test, the test group had 0.24 sales per representative per hour, versus 0.18 for the other group. This is about a one-third increase, which clients can easily be persuaded to find admirable.

In summary, Smart-Call breaks the traditional focus on dialer performance. Instead, it concentrates, rightly, on improving the called parties’ experience, taking into account their demonstrated preferences about when they would prefer to be called. In fact, we are not at all convinced that the dialing system 2Contact used is all that good – but the tests and campaign experience using Smart-Call seem to indicate that the performance of the dialing system doesn’t matter all that much. Getting the timing right is much more important. (It should be noted that the numbers reported here have not been independently verified.)

If the success of this approach means that eventually more call centers will be calling you, and if that bothers you, we sympathize. As we said at the beginning, the call and contact centers industry continues to thrive, although perhaps not growing as fast as it did earlier, because, despite all the complaints, these calls do work. On the other hand, if contact centers use this new technology, they seem quite likely to bother you less than they do now, even while making more calls. They can achieve this by nearly eliminating nuisance calls in which you answer and they hang up because they have no one to talk to you, and by recognizing when you’re more likely to be receptive. At least they will know not to keep calling you at dinnertime.

Douglas A. Samuelson (samuelsondoug@yahoo.com), D.Sc. in operations research, is president and chief scientist of InfoLogix, Inc., a small R&D and consulting company in Annandale, Va. He is a contributing editor of OR/MS Today and Analytics magazines and a longtime member of INFORMS.

Olcay Yucekaya (oyucekaya@2contact.com) is a data scientist at 2Contact who also works with major companies outside the scope of 2Contact. He holds a master’s degree in marketing management from Tilburg University in Tilburg, the Netherlands. He gratefully acknowledges the assistance of Professor George Knox at Tilburg University in developing Smart-Call.

References

  1. Douglas A. Samuelson, “System for Regulating Arrivals of Customers to Servers,” U. S. Patent No. 4,858,120, Aug. 15, 1989.
  2. Douglas A. Samuelson, “Predictive Dialing for Outbound Telephone Call Centers,” Interfaces, September-October 1999.
  3. Michael Kaiser-Nyman, D. Samuelson and B. Swieskowski, “Predictive Dialing System Using Simulation,” U.S. Patent No. 9,088,650, July 21, 2015.

Analytics data science news articles

Related Posts

  • 38
    Most companies have massive databases of historical sales data, but few firms invest the money and staff time to mine the intelligence hidden in those databases. It seems that everyone has sales data, but almost no one does a good job of analyzing that data. The purpose of this article…
    Tags: sales, data
  • 37
    It’s long been popular to talk about customer interaction data such as clickstream, social activity, inbound email and call center verbatims as “unstructured data.” Wikipedia says of the term that it “…refers to information that either does not have a pre-defined data model or is not organized in a pre-defined…
    Tags: data
  • 36
    With the rise of big data – and the processes and tools related to utilizing and managing large data sets – organizations are recognizing the value of data as a critical business asset to identify trends, patterns and preferences to drive improved customer experiences and competitive advantage. The problem is,…
    Tags: data
  • 35
    Many organizations have noticed that the data they own and how they use it can make them different than others to innovate, to compete better and to stay in business. That’s why organizations try to collect and process as much data as possible, transform it into meaningful information with data-driven…
    Tags: data
  • 32
    I have, like many others, been caught up in the Hamilton craze. Duels are a central feature of the Broadway show, with three prominently featured, two of which were fatal. Since neither I nor Analytics editor Peter Horner have been able to procure tickets via the normal means, we will…
    Tags: number, percent, data


Headlines

Report: One in five cloud-based user accounts may be fake

According to the Q2 2018 DataVisor Fraud Index Report, more than one in five user accounts set up through cloud service providers may be fraudulent. The report, based on information gathered between April and June, analyzes 1.1 billion active user accounts, 1.5 million email domains, 231,000 device types and 562 cloud hosting providers and data centers, among other indicators. Read more →

When managers respond to online critics, more negative reviews ensue

A new study in the INFORMS journal Marketing Science found that when managers respond to online reviews it’s possible that those responses could actually stimulate additional reviewing activity and an increased number of negative reviews. The study, “Channels of Impact: User Reviews When Quality is Dynamic and Managers Respond,” is authored by Judith Chevalier of the Yale School of Management and NBER, Yaniv Dover of the Hebrew University of Jerusalem and Dina Mayzlin of the Marshal School of Business at the University of Southern California. Read more →

IE student designs software to optimize snow removal at Penn State

It is well known among the State College and Penn State communities that it takes a lot for university officials to shut the campus down after a major snowfall. In fact, since 2010, the University Park campus has been shut down just three full days due to snowfall. Much to the chagrin of students – and faculty and staff – the snow day at Penn State may just have become even more elusive, thanks to software developed by recent industrial engineering graduate Achal Goel. Read more →

UPCOMING ANALYTICS EVENTS

INFORMS-SPONSORED EVENTS

Winter Simulation Conference
Dec. 9-12, 2018, Gothenburg, Sweden

INFORMS Computing Society Conference
Jan. 6-8, 2019; Knoxville, Tenn.

INFORMS Conference on Business Analytics & Operations Research
April 14-16, 2019; Austin, Texas

INFORMS International Conference
June 9-12, 2019; Cancun, Mexico

INFORMS Marketing Science Conference
June 20-22; Rome, Italy

INFORMS Applied Probability Conference
July 2-4, 2019; Brisbane, Australia

INFORMS Healthcare Conference
July 27-29, 2019; Boston, Mass.

2019 INFORMS Annual Meeting
Oct. 20-23, 2019; Seattle, Wash.

Winter Simulation Conference
Dec. 8-11, 2019: National Harbor, Md.

OTHER EVENTS

Applied AI & Machine Learning | Comprehensive
Dec. 3, 2018 (live online)


Advancing the Analytics-Driven Organization
Jan. 28–31, 2019, 1 p.m.– 5 p.m. (live online)

CAP® EXAM SCHEDULE

CAP® Exam computer-based testing sites are available in 700 locations worldwide. Take the exam close to home and on your schedule:


 
For more information, go to 
https://www.certifiedanalytics.org.