OR/MS Today - August 2000 - Was it Something I Said?

August 2000

Was it Something I Said?

The Magic is in the Model, Not the Math

By Vijay Mehrotra

In my last column I wrote that I had worked on the development of a sophisticated scheduling solution. Promoted by a skilled professional sales force, the product was a huge success with the targeted customer population when it made it to market in 1999. "Once we started working with these customers," I noted, "we discovered something that totally stunned us: More than 60 percent of these organizations were not using the Optimal Scheduling functionality that was the primary mathematical capability of this software. Yet, to this day product sales continue to escalate, and customer success stories keep pouring in."

This excerpt raises the following questions:

What kind of success were these customers having without the Optimal Scheduling?
Why aren't they using Optimal Scheduling, anyway?

The software product provides an infrastructure for forecasting traffic and scheduling agents within a call center. The planning system is built around an optimization engine, while also drawing on time series, queueing theory and simulation. Its user interface uses the terminology of the call center, and inputs and outputs are stored in a relational database.

Our design assumption was that customers would generate weekly forecasts for each call queue and build optimal schedules. However, few call centers actually have a culture that supports changing every agent's schedule on a weekly basis. Therefore, relatively few customers use the software as we had originally envisioned.

So how does the software provide business value to these customers?

Automation. With or without data, software or optimization, every call center must create schedules for its agents. This can be an intensely time-consuming activity. Therefore, many customers bought our software because it enables them to do exactly what they were already doing — on paper or in spreadsheets — with less effort and in less time. "Major labor savings," promised the salespeople.

When I first heard this, I was furious! Pearls to the swine! Had we created nothing better than an elegant guillotine, enabling customers to cut headcount by automating existing functions? Instead, we found that we were helping perpetually overworked analysts automate the most tedious part of their work. One customer put it this way: "Two of us used to spend a full day every week building our forecasts, and nobody ever believed them anyway. Now, we spend a total of 15 minutes per week on forecasting, and everybody agrees on the inputs and results."

Even without optimization, customers now have the time and energy to use their unique operational knowledge to find and implement improvements.

Tracking system. Information needed to produce optimal schedules comes from such divisions as telecommunications, human resources, finance and management. Some detailed data (most notably call traffic and system performance) is stored in existing systems with limited historical duration and access. Other information (agent attendance, training plans) is maintained informally, inconsistently, irregularly or not at all.

Our planning software provided a structured database and a friendly user interface for entering and storing data. In addition, the process of building the system created the need for different people to engage in discussion about key policies and objectives that need to be quantified as model inputs.

Visibility into expected system performance. Call centers are increasingly complex stochastic systems, managed by people who are not familiar with statistical variation. Historically, they have lacked the tools or methodology to quantify the connection between staffing levels, call forecasts and waiting times. Through our software, analysts enter core operational data and quickly get a handle on which queues are likely be understaffed at which times, giving them a fighting chance to head off disaster.

Enabler for culture changes. Most significantly, we found a chicken-and-egg relationship with optimal schedules. Few customers disputed the potential benefits of optimization. However, their culture was not set up for new schedules to be created on a regular basis. The story was usually something like this:

Without automation, it had taken a huge amount of time to create a schedule.
Therefore, schedules were either created infrequently (once a quarter, once a year), or worse yet, each agent was assigned a permanent schedule when first hired.
Consequently, agents are used to having "fixed" schedules and would resent being told when they should come in to work.

However, once our system was implemented, many customers were able to establish new business processes for building or updating schedules on a more regular basis. More commonly, customers began to tell new employees to expect new schedules every week or two based on the forecasted call traffic.

What general results can we take away from all of this? Most significant is what I'll call Vijay's Second Law: "The magic is in the model, not in the math."

We set out to build a framework for optimal scheduling, but in the process we built a representative model of how different operational variables interact within the call center along with the motivation and means for storing the key data elements. This framework proved to be more accessible, and more valuable, to our customers than we had ever imagined.

Vijay Mehrotra is the CEO of Onward Inc., in Mountain View, Calif. He can be reached via e-mail at vijay@onward-net.com.

Table of Contents

OR/MS Today Home Page

Lionheart Publishing, Inc.
506 Roswell Street, Suite 220, Marietta, GA 30060, USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com
URL: http://www.lionhrtpub.com