Innovation + Smart Planning = Happy Customers

HP Labs’ Revenue Coverage Optimization tool breaks new ground in operations research, improves the customer experience and saves HP millions of dollars in the process

By Simon Firth, April 2009

Principal Scientist Julie Ward

Principal Scientist Julie Ward
Photo by Eric Anderson

Genuine ‘Eureka!’ moments are few and far between, even in world-class research labs. But not so long ago HP Labs researcher Julie Ward got to witness one firsthand.

Ward and her colleague Bin Zhang were both attending a meeting when Zhang started thinking about a tough problem he and Ward had been trying to solve in applied operations research.

“Bin was supposed to be paying attention to something else,” Wards recalls. “But he started writing, frantically turning pages. Basically, we watched that innovation unfold.”

Seeing what Zhang was onto, Shailendra Jain, manager of the research team, kicked Zhang out of the room so he could fully work out his solution. It turned out, says Ward, to “constitute a significant advance in operations research.”

Not only that, Zhang’s insight helped Ward, Jain and Zhang refine a software tool – called Revenue Coverage Optimization (RCO) – they’d been developing for HP’s own businesses. It’s a tool that’s since allowed the company to offer customers a significantly improved service and at the same time save HP hundreds of millions of dollars in improved efficiencies.

In recognition of these achievements - both theoretical and practical - the HP Labs and HP business teams responsible for implementing the RCO tool won the 2009 Franz Edelman Award, the world's leading prize for excellence in operations research practice.

One shot solution

“Once you see it, it’s not a hard problem,” insists Bin Zhang of his theoretical breakthrough. “I was really just lucky to see things in that particular way.”

Zhang was wrestling with a well-known problem in operations research – how to find the maximum flow through a network of nodes connected by links that have limited capacity. In particular, he was interested in how to solve a series of maximum flow problems in the same network, each with different link capacities.

Existing maximum flow algorithms did this by solving a series of related, but distinct, max flow problems. Zhang’s insight was to box all the maximum flow points together to radically simplify the problem in a technique he calls flow balancing. The upshot is that you can solve all the problems in one shot.

An added bonus, says Zhang, is that “the algorithm’s code becomes so easy to implement. I just wrote it up in one night and the next morning I had it running.”

Coverage optimization produces tangible business results

Thanks to flow balancing, the RCO tool now ran five times faster. That, in turn, helped the HP Labs team apply it with great success at several HP business units.

First to use it was HP’s Personal Systems Group, at whose request Jain’s HP Labs team initially began working on revenue optimization. The maximum flow problem associated with PSG’s application of RCO is very large, with hundreds of thousands of nodes and millions of arcs in the underlying network, so having an efficient algorithm is critical.

With the RCO tool says Ward, “we were able to offer PSG a new metric that hadn’t been considered before: what we called ‘coverage.’ Coverage accounts for the importance of products that generate revenue on their own, as well as those that enable the sales of other products.”

Figuring coverage into the equation allowed one division of PSG to establish that roughly 20% of its products, if optimally selected, could completely fulfill 80-85% of all customer orders. As a result, the group was able to significantly decrease the time it took to serve their customers, as well as make order fulfillment times reliably consistent.

PSG now runs the RCO tool as a part of its regular planning cycle, says PSG special project manager Holger Mishal. “We call it the Recommended Offering Program,” he says, “It’s really helped us create predictable and therefore very competitive order cycle times.”

RCO is also integrated into a second program at PSG, called the Global Product Offering Program, which offers HP’s global customers a set of products that they can buy in every country in which they do business.

“We’ve used the tool to analyze all orders from enterprise customers worldwide,” says Mishal. “It allows us to accurately determine which 20 to 30 percent of the features and configurations we offer satisfy 80 percent of the orders.”

The result, he says, is a global product set much better attuned to customer needs. Before RCO was applied to the process, Mishal notes, “only around 20% of customers used the process. Now, the adoption rate stands at 83%.”

Product portfolio rationalization

RCO has also helped HP’s Business Critical Systems (BCS) unit in the company’s Enterprise Servers and Storage group radically rationalize their product portfolio.

Since late 2004, the unit has used RCO to eliminate 3,313 products from its portfolio.

That resulted in over $11M in reduced inventory and planning cost, says BCS Supply Chain BOP Cookie Padovani. But the real value, she says, “is in how we’ve changed the customer experience by cleaning the portfolio. What we’re able to offer now is much simpler and much easier to use.”

BCS has even used RCO to help design their new product offerings based on order histories from prior generations of products – allowing the unit to offer fewer products that are at the same time more likely to satisfy their customers’ needs.

Extending the solution

Part of RCO’s value as a tool is that it’s very flexible, says HP Labs' Ward. Indeed, Ward and her team are now actively working with other HP business groups to deploy the program.

Other Labs colleagues, led by HP Senior Fellow Robert Tarjan, have also been generalizing Bin Zhang’s insight – technically known as a parametric bipartite maximum flow algorithm – to non-parametric max flow problems in general graphs, making an even more fundamental contribution to the field.

As a piece of path-breaking applied research, though, RCO has already more than paid its way. Each time it’s been deployed, RCO has helped create product offerings better attuned to customer needs, resulting in better customer satisfaction and improved sales for HP. On top of that, PSG alone estimates over $300 million dollars in savings from using the tool.

“It’s not that often that you have such an important algorithmic contribution,” Ward says, “that at the same time has such immediate practical value.”

The Franz Edelman Award is awarded by the Institute for Operations Research and the Management Sciences (INFORMS). The 2009 winning team, HP, was announced at the INFORMS Conference on O.R. Practice in Phoenix, AZ on April 27th.

Getting to know Bin Zhang

Researcher Bin Zhang

Principal Scientist Bin Zhang
Photo by Eric Anderson

Bin, how long have you been at HP Labs?
I’ve been here since 1998.

What's your area of research?
I'm in computing. Mostly, I'm interested in data mining - how can you handle large data sets and do computation on them fast.

How did you get interested in computer science?
I was interested in assembling radios and TVs. Later I was curious about logical circuits. I ended up majoring in mathematics and then computer science.

What’s the most rewarding part of your job?
It’s when you have an idea you just want to work on without stopping. That’s the most fun part. You forget about everything else.