FICO Xpress Optimization Delivers up to 5.7x Faster Mixed Integer Programming (MIP) Since 2020

FICO Xpress Optimization’s Capabilities

FICO® Xpress Optimization is the leader in commercial optimization, enabling organizations and business users to solve large-scale optimization problems quickly.

FICO Xpress Solver is the technology component supporting all of FICO's optimization solver technologies including:

• Xpress Optimizer: the core engine comprising of the linear programming (LP), mixed integer programming (MIP), (MI)QP, (MI)QCQP, and (MI)SOCP solvers. It also provides core services, like the solver tuner, to other components.
• Xpress Nonlinear: provides successive linear programming and an interior point solver for nonlinear problems.
• Xpress Global: provides an engine for solving nonconvex mixed integer nonlinear (MINLP)  problems to global optimality
• Xpress Kalis: provides a constraint programming (CP) solver. Combined with the optimizer, it can make use of MIP relaxations of the CP problem.

Understanding Mixed Integer Programming Mixed integer programming (MIP) is a powerful mathematical optimization approach that extends linear programming by allowing some or all of the decision variables to be integer values. This blend of discrete and continuous variables enables the modeling of complex real-world decision problems (such as scheduling, facility location, supply chain network design, and resource allocation) where certain choices must be made in whole units or binary terms.   Consider the following conceptual, but real-world, example. A consumer-packaged goods company is testing three new product varieties (A, B and C), with plans to launch only two of the varieties to market. During the market research phase, the company surveys four individuals, each indicating which of the three product varieties they prefer (1 = prefer, 0 = do not prefer).   The company’s objective is to identify which two product varieties will appeal to, or reach, the most individuals.      Now consider a true, commercial-sized problem, where evaluating every combination is intractable. The company is now evaluating 60 new product varieties with the goal of launching 18 varieties to market, and surveys 25,000 individuals who indicate which of the 60 product varieties they prefer.   This commercial-sized example results in over 925 trillion (> 925,000,000,000,000) eighteen-product combinations. The problem is the same; the objective is the same, but it’s now on a scale that requires a high-performing commercial solver like FICO Xpress to solve.   Mixed integer programming models are capable of capturing intricate business rules and logical relationships, making them crucial in prescriptive analytics and decision optimization initiatives. As the number of integer variables grows, the complexity of the problem increases significantly and advanced solver technologies are essential for scalability and efficiency: this is where optimization software like FICO® Xpress comes in. This is the FICO Xpress difference.

FICO Xpress New Release and Improved Mixed Integer Programming Solver Performance

With its latest 9.8 release, FICO Xpress continues to redefine the limits of what’s possible. By incorporating a GPU-accelerated implementation of the PDHG algorithm, users can experience speed accelerations up to 50x on large linear programming problems.

Since 2020, FICO Xpress’s performance on its core algorithms have experienced double-, triple-, and quadruple-digit percentage speedups across the board. These performance gains allow users to solve larger, more complex optimization problems faster today than ever before.

Dinakar Gade, Director of Product Management for FICO® Xpress, puts these performance gains into context by noting, “FICO Xpress is used for many flights you take, many of the packages that arrives on your doorstep, and even phone calls you make relies on commercial optimization software to keep operations running smoothly. As these logistical challenges continue growing in size and complexity, optimization software must continue improving to meet the demands of tomorrow.”

Year-over-year comparisons (version 9.8 vs. 9.5) follow the same trend. The mixed integer programming (MIP) solver is 14% faster overall and 24% faster on models taking at least 100 seconds to solve. The global mixed integer nonlinear (MINLP) solver is 68% faster overall and 5.3x faster for models taking at least 100 seconds to solve.

How FICO Xpress Customers Benefit From Our Performance Improvements

Our newer customers appreciate continual double-digit+ performance gains, and our longer-tenured customers have come to expect them. Even those early in their adoption journey, those who are just now beginning to incorporate optimization into their broader AI and ML workflows, recognize that as their problems get larger and their use cases expand, FICO® Xpress will be there to meet their needs.

One such customer is PepsiCo, a global food and beverage giant needed to optimize complex production and staffing decisions across its global manufacturing network to keep pace with changing demand. PepsiCo has relied on FICO® Xpress Optimization since 2009 to tackle some of its most complex planning challenges. By using Xpress, PepsiCo has cut daily production model solve times by 30% and has achieved an instant 15% reduction in solve time over the previous release by upgrading to the newest Xpress version. 

Lirthik Durai, principal optimization scientist at PepsiCo, stated, “Each new release of Xpress delivers faster solve times on the same models and problem sizes, with no changes required on our side. Those gains have real operational impact by letting us run more scenarios, make better decisions, and keep production on schedule. The support from FICO has also been exceptional. Their experts worked side by side with our team to fine-tune the solver and achieve major performance gains. That collaboration has built real trust and has been a big factor in why we continue to choose Xpress year after year.” 

Download the case study to learn more about the PepsiCo and Xpress partnership.

See How FICO^® Xpress Can Deliver the Optimization Performance Your Organization Requires

• Try FICO® Xpress 9.8 for yourself free for 60 days by contacting SharbelHanna@fico.com or visiting the FICO® Xpress Licensing page to get a free evaluation from Xpress experts
• Access the FICO Xpress Solver product sheet for deeper insight into its capabilities
• Read more about FICO Xpress’s performance compared to AlphaEvolve
• Download the Shell case study to see how Xpress helps optimize control processes