An Interview with Manoj Chari, R & D Director of Operations Research
The SAS/OR® product has been a key member of the analytical offerings from SAS for years, and more recently,
it's undergone a significant "growth spurt" to stay current with recent methodological developments.
Optimization techniques now play a major role in other SAS products as well, such as SAS® Marketing
Optimization and the SAS® Revenue Optimization Suite.
Manoj Chari is SAS Director of the Operations Research department. After graduating with a PhD in
Operations Research from the University of North Carolina at Chapel Hill, he spent eleven years
teaching mathematics and performing research at the University of Waterloo and Louisiana State
University. He returned to North Carolina to take a position at SAS.
You realize that our division is becoming a haven for former math professors. What do we have now, six?
What got you away from mathematics and back to your operations research roots?
I did a part-time gig as a consultant for a friend, designing stochastic optimization models for natural gas storage.
It was a chance to take a break from the abstract research I was doing for some practical work, including hands-on
programming, and I loved it. When the word got around, my graduate school colleague Gehan Corea called me about
working for SAS and designing optimization algorithms for some of our new solution projects. My first job was
designing the algorithms used in the SAS
® Marketing Optimization software. I'm still very proud of that.
I remember Lagrange multipliers, and I studied stochastic processes as part of my statistical training. But I
never took an actual OR course. What's a good working definition of operations research for our readers who
are more versed in statistics and econometrics?
Operations research offers a variety of tools that enable a structured way of understanding and improving the
performance of systems through mathematical (and other logical and analytical) models. When the model's
parameters and their relationships are quantifiable as closed-form mathematical expressions, an application
of optimization might be appropriate. Systems with more complex relationships, especially those with dynamic
and uncertain elements, are best studied with a simulation approach. Fundamentally, all OR techniques support
quantitative decision making and usually result in a prescription of a certain course of actions over others.
So, what are some good examples of the use of SAS/OR techniques?
Determining the flow of goods from suppliers to customers through a distribution network so as to minimize
costs; scheduling and monitoring a large and complex project—for example, assembling a rocket system or a
ship from components-that consists of an enormous number of activities with precedence relations and resource
sharing; modeling the dynamics (movements of patients, staff and other resources) in the surgery unit of a
hospital—the prospective list of applications is virtually endless.
When I started at SAS, the OR group had three people. Now it has close to thirty. What accounts for this type of growth?
There has been greater recognition of the value of operations research for improving all aspects of business
decisions. There has also been tremendous progress in computational capabilities due to dramatic improvements
in algorithmic techniques as well as an increase in raw computing power. SAS was also interested in developing
solutions that relied on optimization. So, we invested in a huge effort to completely update our optimization
software and begin work on these new solutions, too. We now develop SAS/OR software, design algorithms for solutions
software, and serve as a resource for optimization and OR at SAS in general.
Is it true that the OR group at SAS might be one of the largest, if not the largest, groups of OR
specialists anywhere in the world?
That's right. Most of my staff have PhDs in operations research, but we also have people with PhDs
in applied mathematics and engineering who focused on operations research techniques in their training.
We are quite active professionally. Our staff members are often invited to give talks at the INFORMS (Institute for
Operations Research and the Management Sciences) meetings and at other major conferences such as the
International Symposium on Mathematical Programming.
So what are the major outcomes of the SAS/OR updating?
We wrote the OPTMODEL procedure, which is the modeling front end to a completely new family of optimization
solvers covering linear, mixed-integer linear, and nonlinear optimization. PROC OPTMODEL is a full-fledged
interactive algebraic modeling language that features flexible input and output, including the ability to
read from and create any number of data sets. It's integrated with the SAS programming environment while
allowing direct access to all new SAS/OR solvers. Our lineup also includes other new procedures for
genetic algorithms and constraint logic programming, as well as new software for simulation, SAS
® Simulation Studio.
So is operations research the new frontier?
Well, I don't know if I would call it that, but certainly having knowledge of the value of operations
research is important in nearly any organization. In some sense, once you've assessed your data with
statistical and econometric techniques, it's a natural next step to want to understand, optimize, or
simulate the performance of your systems. Operations research is essential here. I'd encourage all
analytical groups to become more aware of operations research techniques—especially as we see the definition
of operations research expand to embrace more and more analytical methods.
So, are more people using SAS/OR Software these days?
Yes, we had sizeable new sales growth in 2008, and, interestingly, we saw rising use
of SAS/OR software at sites where they've had it for years, but perhaps have used it sparingly.
What's the OR Center of Excellence?
We were finding ourselves involved in a lot of pre-sales and consulting support situations, helping
potential customers formulate their problems to show them how operations research in general and SAS/OR
software specifically could help them. We decided to formalize the activity, so we'll have some staff
members dedicated to this sort of activity. It might serve as an incubation group of sorts, in which we
might grow new optimization solutions as we become familiar with some common and recurring problems
facing our customers.
Is there a guiding principle to creating models in the consulting work?
I always think of something Einstein once said-that models should be as simple as possible but no simpler.
You strive for simplicity, but it can't be so simple as to be useless. There's a real art to creating
models that balance tractability and usefulness.
Your group has a lot on its plate!
That's right. Not only are we writing code, we're designing algorithms for other groups and working
with both internal and external customers. So, we recruit people who can excel at all of these activities. It's very challenging but very exciting.
What's the best thing about working at SAS?
I get to work with extremely smart and dedicated people every day.
You stole that from me!
No way! We work in different buildings! But seriously, that's what makes me want to come to work every day. I love working with my group.