Strategic decision-making tools

Continental Airlines needed to re-plan its flights because of a snowstorm, and while the company's operating system could unmistakably pinpoint where thousands of planes and flight attendants are now, it couldn't tell managers how to make flights The plan is to get back to normal as soon as possible; when advertisers call NBC, salespeople only know which time slots are available, but they don't know how to provide the best mix of time slots for the advertiser's target customers.

At this time, Continental Airlines and NBC have used operations research (OR), and the most common complex problems in management can be used to find the best solutions.

Enterprises expect information technology to connect various business activities and invest in it. "Enterprise Resource Planning" (ERP), "Customer Relationship Management" (CRM) and "Supply Chain Systems" (SCM) all provide up-to-date data for business operations, but whether decisions are made based on this data remains to be further analyzed. Especially in the era of increasingly competitive meager profits, entrepreneurs need the most ideal results.

Operations Research Helps Decision-Making

Ten years ago, most companies were unable to make optimal decisions. The scarcity of information and the high cost of computation made it only possible for large companies to try. Now, high-quality data is more and more readily available, optimal decision-making is commonplace, and operations research shines brightly.

Ops research is the science of decision-making. The methods include identifying business problems, providing possible solutions to the problems, and modeling these variables with various forms of applied mathematical software. Specifically, operations research experts collect relevant data to build and optimize models, and finally evaluate the results, which provide management experts with actionable courses of action.

Operations research can help address the following questions:

● How to arrange shipments to meet customer needs and maximize driver usage?

●How to allocate marketing budget to maximize customer response rate?

●Where can a warehouse be set up to minimize shipping costs?

●What will be the impact on the speed of the machine during the production process?

Faced with these wide-ranging business problems, an unwieldy manager might say, "Let the computer give the answer!" Unfortunately, straight-forward computational methods can't do anything about it. Take the warehouse problem as an example. Suppose a company has 100 possible storage locations and wants to determine the best 10. With the help of a computer, you spend an hour evaluating a storage solution and two hours evaluating all the possibilities. That's okay, but if you increase the selection to 15, the computation time increases significantly. If there are 200 possible storage locations, the computing time will increase by a factor of 1000, and no amount of computing power can solve this problem.

Pushing the boundaries of ERP

Operations research integrates mathematics and other scientific methods to model real-world problems and propose solutions. Fifty years ago, linear programming was a revolutionary achievement in operations research, helping planners find solutions from thousands of possibilities and constraints. Although we have known for decades that all system constraints and objectives in linear programming can establish linear equations, it was only 15 years ago that we developed effective execution software.

Other methods of optimization either promote linear programming, or dismantle the basis of the linear model in a completely different way, "constraint programming" being an example. This method can handle more complex decisions, but the fly in the ointment is that it proposes fewer decision-making options.

More and better software brings operations research to more enterprises in the form of batch processing. Microsoft's EXCEL has the ability to optimize solutions. Other tools include analytics, decision support, ERP and financial planning software. Manufacturers are the initial users of optimization technology. They design production platforms and production lines to ensure that customers can get products in the shortest possible time.

For example, a mid-sized manufacturer spent millions of dollars on an ERP system that reduced lost orders, shortages of raw materials, and more. But the factory system couldn't keep up, and the delivery cycle was four weeks long, and the situation didn't improve after the company added new automation equipment.

New and old machines in the workshop have their own operating characteristics. Every morning, the ERP draws out the production list for the day. Experienced schedulers will distribute production tasks across machines and arrange them in an order that minimizes the number of changeovers.

A key question is whether the new machine will handle high-volume or low-volume production. The operators believed that the automatic changeover process of the new machine had shortened the time between runs, so they put small batches on the new machine. This practice increases the number of automatic changeovers and reduces the burden on the operator. Another approach is to use fast new machines for high-volume production to reduce the impact of machine acceleration and deceleration on the product during changeovers. So, which method is more efficient?

ERP cannot answer such questions, it only provides the most basic planning analysis. Instead, an operations research model that incorporates key metrics such as the number of transitions, the impact of acceleration/deceleration, and the effect of mixing work can address the issue. The calibrated model was used to measure system performance, then optimized over several days of data, and the result was clear: the operator was wrong. It was more cost-effective to have the new machines do high-volume production, and the system increased production capacity by about 20 percent. With this optimization tool, lead times are greatly shortened.

Airlines have also benefited a lot. Continental worked with Caleb Technologies to develop a "disaster-recovery model" that asked questions such as: How can Continental resume normal operations as quickly as possible if Chicago's O'Hare Airport is forced to close for a day due to a snowstorm ? Problems like this are often difficult to solve because the scale of the problem is so large. For Continental, the contingency plan must account for 1,400 flights per day, 5,000 pilots, 9,000 flight attendants, complex air traffic regulations and contracts with unions, which are by no means a spreadsheet-optimized solution.

The decision-making system allows Continental to deal with severe weather more quickly and cost-effectively, while minimizing the negative impact on passengers. This system can easily save millions of dollars a year, while also improving responsiveness to customer requests in difficult situations.

Wide range of applications

Operations research can also be used effectively in sales departments. NBC faced a variety of situations when it came to planning sales for its advertisers. A sales plan requires a certain balance between channel resources and customer needs. In addition, NBC also hopes that the pricing of ad slots will maximize revenue. According to the original method, when a salesperson customizes a sales plan, they often encounter such a situation: some time periods of TV programs are full of advertisements, while other time periods are empty. When the price needs to be adjusted, most of the program plans They are all set to die, and a lot of income is lost as a result.

This is where operations research comes in handy. The model consists of two main aspects: first, to develop an advertising strategy that meets both budget requirements and the needs of each advertiser in all ad slots; second, to identify possible opportunities for price changes in a timely manner and make them Affect advertising campaigns. If a certain ad time slot is too expensive, lower the price; conversely, if a certain time slot is particularly popular, increase the price in order to get more money from the advertiser.

Research results show that this program better meets advertiser needs while selling as many time slots as possible at higher prices. In the four years since implementing the new system, NBC has increased its revenue by $200 million, and salespeople no longer have to struggle to pull together all of their sales plans, which greatly increases their productivity. On top of that, NBC is pleased to see advertiser satisfaction also rising.

The biggest trend in operations research over the past few years has been the widespread adoption of optimization techniques in various business areas. In particular, optimization software is integrated into various spreadsheet software packages and supply chain management systems. Operations research models are now widely used to plan medical examinations, various sports groups, agricultural harvesting, and other fields, and the benefits of operations research are everywhere.

Another trend is the emphasis on flexibility. The example of Continental Airlines shows that the airline and other departments create models that can adapt to changing requirements.

Optimization also allows people to better deal with uncertainty -- something that's not uncommon in business. Traditional operations research methods assume that the data is known and unchanged. Methods such as scenario optimization and stochastic programming reduce the requirements for assumptions. The use of various techniques enables models to operate under uncertain conditions. Below is more optional value.

The need for business optimization has never been more urgent. The pursuit of operational excellence in business has led companies of all types to apply operations research to improve business processes, production systems and collaborative efforts. Without operations research, information technology can also provide a range of data, but you don't know how to apply them.

Often company management will organize an internal think tank or mobilize the whole company to solve certain problems, such as production or supply chain management.

Continental's Operations Research Specialist is part of the IT department, reporting to Chief Information Officer Janet Wejman. Their efforts over the past few years have freed the company from dysfunctional operations. "Every time a company deviates from the normal shipping trajectory, we say it's an abnormal state, such as the flooding that forced the closure of the Houston airport a few years ago, and the 9/11 terrorist attacks," Wejman explained. "If we can learn from these We can save a lot of money by shortening even a few minutes in an abnormal situation." It is the optimization technology that has helped Continental Airlines save a lot of time in these abnormal operations.

Continental's IT department developed three key optimization application areas to plan for day-to-day business and deal with unplanned disruptions. The three application areas are: crew solving, manpower solving and operations solving. Developers, operations research specialists, Wejman and flight management under Debbie McCoy worked together to design these systems and put them into service. "We start by identifying problems that can be solved with software, and then let our intellectual property shine."

The result is not only operational efficiency, cost savings, but also morale. The system sends flight plans to flight attendants through the Internet and updates them at any time according to the situation. "Employees like to keep up with changes," says Wejman.

This text is reproduced with permission from Optimize magazine, published by CMP Media LLC. The author registered copyright in 2003. Translated by Hong Jing. Michael Trick is Professor of Operations Research at the Institute of Industrial Management at Carnegie Mellon University.