skip to main content

Optimizing Productivity Analysis For Better Factory Performance

Madhav Kidambi, Shekar Krishnaswamy and David Norman

The complexity of semiconductor manufacturing makes it difficult to optimize production for increased efficiency and output while maintaining high product quality and reducing overall costs. Yet it is imperative that semiconductor manufacturers and outsourced assembly and test (OSAT) suppliers find ways to do so, because production inefficiencies can keep them from fully capitalizing on the opportunities presented by today’s fast-changing, highly competitive markets.

One strategy is to approach the tasks of production planning, scheduling, dispatching, and reporting in an integrated, flexible, and open fashion, with the goal of optimizing production every step of the way. This means from initial enterprise-level master planning down to the factory floor dispatching of specific lots of work-in-progress (WIP) in real time.

Enabled by an integrated suite of automation software that incorporates a state-of-the-art optimization engine, such an approach would allow fabs to anticipate and better respond to fluid conditions, thereby controlling the speed of WIP processing and reducing waste. It could help manufacturers overcome the impediments to higher productivity they face every day, including:

  • Inability to look at factory resources on a holistic basis.
  • Disruption, delays and cost associated with having to analyze and validate changes to dispatching rules and scheduling policies directly on the factory floor, in response to changing conditions.
  • Inability to simulate dispatching rules with su˜fficient accuracy to build realistic “what-if” models that reduce bottlenecks, increase overall throughput, and ensure that cycle time commitments are met.
  • Lack of cohesion resulting from the di˜ffculty of coordinating the efforts of engineers who work separately in planning, dispatching and shop floor “silos,” making it di˜fficult to achieve faster and better operational decision-making and reporting.

Manufacturers often see the need to optimize production, but cannot do so e˜fficiently for various reasons. For example, their systems may not provide good, real-time data from the fab (as opposed to data used in spreadsheet planning). Or there may be a lack of integration between optimization and execution mechanisms, combined with an inability to automatically initiate the right actions based on certain critical events and timing.

Other factors may include the lack of an integrated way to visualize inputs and outputs and limited diagnostic information to help trace why and how decisions are being made.


To begin to address these issues, it is useful to think of production planning, scheduling, dispatching, and reporting as a hierarchy of functions, organized by specific tasks and requirements and with different time frames (see figure 1).

Figure 1. Looking at production planning, scheduling, and dispatching as an integrated hierarchy of functions, as illustrated above, optimizes the use of production resources every step of the way, from enterprise-level master planning to real-time dispatching.

The master planning function deals with enterprise-level planning issues. Master planning is generally the responsibility of a central planning organization whose main goal is to interface with customers and address their overall needs. Its scope includes overall sourcing strategies and capacity analyses to decide which fab(s) will be used to produce parts for a customer. The master planning group typically looks 1–18 months ahead.

Factory planning, meanwhile, performs capacity analyses for tool groups with the goal of delivering product by the due date while balancing factory constraints. Tasks in the purview of this group include decisions to set up tools in certain ways, whether to bring on more personnel for given work shifts, and so forth. These planners typically look 1–4 months ahead.

The scheduling group makes sure that key tools and tool groups are available and WIP is balanced so that they can meet the factory objectives. Another key goal is how to load the products across the tool group to achieve factory planning objectives (e.g., to meet lot due dates, cycle time targets, setup time goals, etc.). The time frame here is generally 12–24 hours ahead, essentially a work shift or two.

Finally, the dispatching group is real time-oriented, with the goal of executing tool schedules to meet fabrication process requirements and maximize production throughput.


Most manufacturers use a disparate mix of software applications for planning across the hierarchy. These tools may include open source solutions, spreadsheets, commercial products, and homegrown applications.

However, these heuristics-based tools trade quality for speed. That is, while they may produce results quickly in response to changes on the factory floor, those rules may not represent the best of all possible solutions.

For example, a new dispatching rule that is adequate for a given process step may be so narrowly written that it precludes opportunities to balance production down the line, as can happen when manufacturers inadvertently create downstream bottlenecks by trying to push more WIP through a given set of tools.

Although optimization algorithms may be used to improve the performance of these heuristics-based programs, they don’t offer a comprehensive infrastructure for developing and maintaining solutions. Also, these systems require additional databases and applications to be supported. For example, additional custom code is required to prepare input data, set up and run optimization problems, and validate and post-process optimization results.

By contrast, an integrated automation framework can offer users the biggest potential productivity gains (see figure 2). Applied Productivity Family (APF) software can be used to build a framework to optimize planning across the hierarchy. It encompasses production planning, dispatching, automation, and scheduling software for equipment and process-control systems.

Figure 2. An integrated automation framework is necessary to achieve optimally efficient advanced production systems in fabs and OSAT packaging operations. (Source: Production Planning & Control.[1])

A fast, robust mathematical optimization program—referred to here as a “solver”—works with the APF software to holistically analyze and implement production requirements. With it, manufacturers can simulate, select, schedule, and dispatch production resources and WIP with optimal productivity and efficiency in real time.

The APF software currently supports the CPLEX and Gurobi commercial solvers as well as the COIN open source solver. This gives users a unique, fully integrated software framework (see figure 3) that is scalable and flexible, and enables them to protect their IP.

Figure 3. These screens are examples of the integrated interface between Applied Materials APF automation software and one of the mathematics-based optimization programs, or solvers, that can be incorporated with it for factory planning, scheduling, dispatching, and reporting. The interface allows users to set variables, constraints, and objective functions as needed.


The productivity advantages this framework offers can be illustrated by a hypothetical look at how the master planning function might be conducted at an OSAT company.

Assume that the OSAT receives wafers from a particular semiconductor manufacturer every week for packaging, testing and assembly, along with specific, detailed requests for the production of multiple lots needed over each of the next 10 weeks.

The OSAT must respond to the customer by specifying what it can actually produce week-by-week within that time frame and committing to do it. With manual systems it takes a few days to formulate this response and commitment, but an automation framework is expected to reduce that response time to a few hours.

The commitment would have to consider not only the customer’s specific requests (generally for multiple lots of product per week), but also additional work for that customer and other customers that may be in process or anticipated in the same time frame.

It is also possible that the customer may change the production request. If that happens, the OSAT would have to quickly determine whether it can meet the new requirements.

Taking into account all of the variables involved in making production commitments, modeling and verifying optimum production scenarios, and then actually carrying them out in a dynamic environment is a formidable task.

Conceptually (see figure 4), a master planner would follow these steps:

  1. Initiate the APF system.
  2. Specify the data to be used.
  3. Configure master planning scenarios to be run.
  4. Initiate plan execution.

Figure 4. By incorporating a solver, APF software enhances master planning accuracy.

The planner would consider higher-level defined objectives such as customer end-product demand, process and device mix, wafer/die schedule, suggested lot release dates, and weekly commitment plans.

Planning inputs would include constraints and other information such as customer orders in the queue, upstream WIP equipment availability, potential capacity shortages, long cycle times, wafer costs, materials requirements, arrivals and inventories, line imbalances, and high levels of variability. All assembly and test operations would be considered, including load boards and handling equipment in the testing area.

The optimization model operates according to predefined logic and “what-if” scenarios executed by the optimization solver, based on automation workflows defined in Applied’s APF framework. The solver minimizes lack of responsiveness to the customer request, and gives a reason for any unsatisfied demands, such as wafer lots that didn’t arrive on time or tools that were over capacity.

As a result, the OSAT may realize specific improvements. These can potentially include better dispatch rules and scheduling, improved bottleneck management, greater or different use of automation systems, and overall higher efficiency and productivity.


In the manufacturing, assembly, testing, and packaging of semiconductors, meeting customer needs for quality, delivery, and cost has never been easy. But today’s complex technology and demanding economics make it more difficult than ever before. Now more than ever, factories and tools alike must be planned, scheduled, and used as efficiently as possible. New solver-based optimization capabilities integrated with already powerful planning, scheduling, dispatching, and reporting automation software will help make this possible from the enterprise level down to the factory floor.

For additional information, contact

[1] Martin Rudberg and Jim Thulin, “Centralised Supply Chain Master Planning Employing Advanced Planning Systems,” Production Planning & Control, Vol. 20, No. 2, March 2009, pp. 158–167.