ILOG and XM presented on optimization and business processes. Optimization in this context is about mathematical optimization aimed at decision support or Operations Research. Creating plans and schedules to optimize use of resources. Optimization helps business create best possible schedules or plans, explore their alternatives and understand trade-offs between conflicting goals. For instance, how much of each product should be produced or what trades would rebalance a portfolio at minimal cost.
Optimization takes a large array of data (demand, resources available, costs, constraints, business goals), executes a mathematical model to optimize the solution and then outputs a schedule or plan with metrics. What-if analysis and scenario planning close the loop back to the inputs and model. It can be used for:
- Long-term planning
Whether to extend a plant or open a new one? How many distribution centers?
- Short-term planning
How much to produce this week, which campaigns will have the best outcome
- Detailed scheduling
Real time or near real-time scheduling or assignment of resources
One ILOG customer is an auto manufacturer using optimization to decide what cars to produce and where. They have 2,164 car models with demand in 110 markets and manufacturing capacity in 12 plants. They produce 100,000 cars each month. They need to allocate production capacity and make supply-demand trade offs (cars share parts so a limited supply of parts must be allocated to specific models) and do this quickly so they can, for instance, change production run to reflect a new marketing campaign.
Another example was a retailer with 7,000 products, 500 suppliers and 18 distribution centers. Needed to decide how much of each product to buy from each supplier and where to deliver it to ensure timely delivery to 130 stores. They used the solution to reduce inventory and distribution costs without risking an out of stock problem. They also need to be able to do what-if analysis on stocking up on products that they know are about to rise in price.
Optimization can be used pretty broadly. ILOG, for instance, has customers like:
- Coors who optimize the loading of beer on to pallets and rail cars
- Airbus who optimize assembly scheduling
- RED Electrica for power generation scheduling to better leverage wind power
- Bundesliga for scheduling soccer games
A business process is a collection of structured activities to meet a client need such as loan approval, claims processing, order fulfillment or invoicing. They are typically focused on a single customer, a single transaction. In contrast the use of optimization is in planning and scheduling processes that cut across all your customers, all your transactions. The two kinds of process can be connected by taking a snapshot of ongoing activities, planning and scheduling that, and then exporting the solution or rules back to operations. For instance marketing campaign optimization involves getting the right offer to the right person in the right channel while constrained by products, personalization, consistency, resource constraints and personal choices. The optimal offer must then be delivered in a process context. This creates a need for a decision management process:
- Use analytics to make predictions about customers (risk, profitability)
- Feed these anlaytic models into an optimization and simulation environment to do what-if and other analysis
- Generate a plan or schedule and merge with policy and preference business rules to execute operational decisions
- Collect and analyze information about the success of this decision making and repeat
Optimization marks the top of the trend that Tom Davenport outlines in Competing on Analytics from reports to real time dashboards to forecasting to predictive analytics to optimization.
XM uses optimization in the Colombian electricity market. 65% of the power generation capacity in Colombia is hydroelectric and the market has been deregulated with many generating companies bidding to offer energy. The electricity market has Long Term elements (contracts), Short Term elements (a spot market) and a Firm Energy market (to ensure will be enough power, especially when dry weather impacts hydro power). Interconnections exist between the Colombian market and Venezuela and Ecudaor to enable interational trading. Optimization is used to manage the dispatch of different kinds of power including the hydrothermal plants, spot market calculation and even long term contracts. Optimization is used to plan tomorrow, manage today and analyze yesterday. Amazing range of uses of optimization in the management of this energy market – really a much smarter grid.