|
Contact Information: |
 |
Frontline's History
This page highlights just some of the milestones in Frontline Systems' history, when Frontline's Solver technology was widely adopted by users and chosen by other, larger software companies, often over other developers' technology and even over internally developed software -- and several occasions when Frontline advanced the state of the art in optimization and simulation technology itself.
What-If Solver
In 1990, Frontline introduced What-If Solver, an add-in for Lotus 1-2-3 and Symphony which was the precursor of the modern spreadsheet solvers. What-If Solver solved both linear and nonlinear problems, at speeds much greater than competitors such as the solver in Lotus 1-2-3/G for OS/2, which appeared at about the same time. What-If Solver won the Editor's Choice in its category in PC Magazine's 1991 roundup of spreadsheet add-ins. In 1992, Frontline developed the Student Edition of What-If Solver, distributed by Addison-Wesley Publishing Company to colleges and universities for use in MBA programs.
The Excel Solver
The introduction of Lotus's spreadsheet solver in 1-2-3/G motivated the other spreadsheet vendors to develop or acquire solvers of their own. In 1990 -- well before the launch of Windows 3.0 -- Frontline won a competition among third-party Solver developers to create a Solver on an OEM basis for Microsoft Excel 3.0. In 1991, Windows 3.0 and Excel 3.0 with its new Solver were both introduced, beginning a period of rapid adoption and growth for the Microsoft products. Frontline Systems has continued to work with Microsoft on the Excel Solver, through and including its latest releases in the 2020s.
Quattro Pro Solver
Borland International originally developed its own linear programming Solver, called the Optimizer, for Quattro Pro 2.0 for DOS. But in 1991, Frontline won another OEM-level competition to develop a new Solver for Borland, simultaneously for Quattro Pro 3.0 for DOS and for the new Quattro Pro for Windows. This Solver appeared in both Quattro Pro versions and replaced the former Borland Solver. In 1992 and 1993, Frontline developed enhanced versions of these Solvers for new Quattro Pro releases. Ownership of Quattro Pro was transferred from Borland to Novell, Inc., which contracted with Frontline to develop further Solver enhancements for the next Quattro Pro version. This latest Solver is part of Quattro Pro as marketed by Corel Corp. in the WordPerfect Suite.
Lotus 1-2-3 Solver
In 1-2-3 for Windows, Lotus carried forward the Solver from 1-2-3/G through 1-2-3 Release 5, the last version for 16-bit Windows. In planning for a 32-bit Windows version of 1-2-3, Lotus recognized the merits of Frontline's (then) five-year program of continuous improvement in spreadsheet solver capabilities, and asked Frontline in early 1996 to develop a new and better Solver for 1-2-3 97 Edition. Frontline's new Solver replaced the earlier Lotus-developed Solver, and was carried forward in subsequent years through 1-2-3 Millennium Edition. It provides capabilities on a par with Frontline's Solvers for Excel and Quattro Pro.
Mathcad Solver
In 1998, MathSoft, Inc., developers of the popular Mathcad product which provides visual display, editing and computational power for technical mathematics users in Windows, surveyed alternatives and then asked Frontline to develop a new and better Solver for Mathcad, with constrained optimization capabilities for linear and nonlinear problems. Frontline's Solver technology provided the new optimization capabilities built into Mathcad 8.0, plus an add-in product marketed by MathSoft called the Mathcad Expert Solver. This Solver has been carried forward to MathSoft's most recent product releases.
Evolutionary Solver
In 1997 and 1998, other vendors developed "Solver" products for Microsoft Excel that were based on genetic algorithms. Although these products were hundreds of times slower than Frontline's Solvers which used classical optimization methods, they had the advantage of handling arbitrary Excel functions, such as IF and CHOOSE, and of getting closer to (but not necessarily finding) a global optimum than classical methods.
In response, Frontline introduced the Premium Solver V3.5 in 1999, which included both fast Solvers based on classical methods and a new, state-of-the-art Evolutionary Solver based on genetic algorithms (that by itself outperformed competitive GA-based Solvers). This gave users "the best of both worlds," since they could define their models once, even using the standard Excel Solver, then apply all of these methods to find the best solution. The Evolutionary Solver has proven extremely popular with users. A special version is bundled with more than twenty textbooks, several of which devote extensive coverage to its methods.
Large-Scale Solvers
As the types and sizes of optimization problems that users wanted to solve grew ever more ambitious, Frontline responded to the challenge by introducing the Premium Solver Platform in 2000. The Platform included major enhancements to the Simplex, GRG and Evolutionary Solvers in the Premium Solver, and also accepted "plug-in" Solver Engines employing specialized methods for large-scale problems. By 2004, the Premium Solver family had grown to 12 Solver Engines, solving problems ranging from a few thousand to over a million variables and constraints. Most of the best known algorithms and methods for linear, quadratic, conic, convex, smooth nonlinear, and non-smooth optimization are now available in Excel, thanks to the Premium Solver Platform.
Interval Global Solver
In 2002, Frontline broke new ground by launching the Premium Solver Platform Version 5 with a new Interval Global Solver -- publishing new research results in academic journals in the process. The Interval Global Solver is the first commercially supported implementation of a range of new methods (interval Branch & Bound, interval Newton method, linear enclosures and the like) that can reliably find solutions to global optimization problems, and all real solutions of a system of nonlinear equations -- a feat that was widely regarded as 'impossible' twenty years earlier, but is now routine for Excel users.
Polymorphic Spreadsheet Interpreter
To make the Interval Global Solver possible, Frontline had to take the radical step of replacing -- and extending -- the core of Microsoft Excel with a new and more powerful "recalculator" -- the Polymorphic Spreadsheet Interpreter (PSI Technology) in the Premium Solver Platform V5. This enables the Platform to evaluate Excel models in many different ways.
In addition to computing "regular number" values for Excel formulas, the Platform can compute intervals such as [1, 2] at every stage of a formula. It can also compute derivatives (gradients and Hessians) for Excel formulas involving thousands of variables, through a process known in the technical literature as automatic differentiation. And it can perform 'diagnostic evaluation' of Excel formulas, to find likely user errors involving poor scaling or non-smooth functions.
This powerful Polymorphic Spreadsheet Interpreter is an 'enabler' not only for the Interval Global Solver, but for all of the Solver Engines supported by the Platform. For example, smooth nonlinear models were solved seven times faster on average, by all of the nonlinear Solvers in the Premium Solver Platform V5.
Convex and Conic Optimization
In 2004, Frontline again broke new ground by launching the Premium Solver Platform Version 6 with a new automatic test for convex models, and a new SOCP Barrier Solver for conic optimization. The automatic convexity test, so new that it was first presented by Frontline at research conferences in the summer of 2004, is not available in any other commercial software -- though it addresses a problem so important that it is sure to be imitated. The SOCP Barrier Solver is the first commercial product of its kind in spreadsheets, and one of the first and only conic optimizers available in any commercial software.
Second order cone programming (SOCP), the natural extension of linear programming (LP) and quadratic programming (QP), has important applications in finance/investment and engineering design, especially in dealing with uncertainty and risk. Convex optimization -- which includes conic optimization -- is the heart of the whole subject of solvers and optimization; the difference between convexity and non-convexity has been called "the great watershed in optimization."
High-Speed Monte Carlo Simulation
Since Microsoft Excel is by far the most-used tool for building business planning models, it is a natural "host" for tools that enable users to "stress-test" the uncertainties in their models through Monte Carlo simulation -- a process where many trials (spreadsheet 'what-if' tests) are automatically performed, sampling values from user-specified probability distributions. While highly useful, simulation in Excel has been slow, because of the need to perform thousands of recalculations of potentially large spreadsheet models. In 2005, Frontline broke new ground in this domain, creating an extension to its Polymorphic Spreadsheet Interpreter (PSI Technology) that was capable of performing Monte Carlo simulations up to 100 times faster than previous Excel-based approaches, using innovative "vectorized evaluation" that computed values for all Monte Carlo trials simultaneously.
Frontline licensed PSI Technology to Decisioneering Inc., a leading vendor of Monte Carlo simulation software, to provide the Extreme Speed feature of Crystal Ball Professional 7.1, their flagship product for Excel. Users and industry experts have applauded the results. According to David Vose, Partner in Vose Consulting and a recognized authority on simulation modeling, "Extreme Speed is a welcome relief to all of us who have large simulation models."
Interactive Simulation in Excel
In 2006, Frontline further leveraged PSI Technology to create Risk Solver Engine for Excel -- a new tool that performs an instant Monte Carlo simulation whenever a user changes a number on a spreadsheet.
"Risk Solver Engine provides the easy to use, lightning-fast Interactive Simulation in Excel that I've envisioned," said Dr. Sam Savage, consulting professor at Stanford University, and a noted authority on Monte Carlo simulation. "It has the potential to do for uncertainty what the spreadsheet did for numbers."
Risk Solver Engine was also the first software product to support two new ideas of Probability Management in large enterprises -- with direct support for Stochastic Libraries (SIPs and SLURPs) of pre-generated Monte Carlo trials, and Certified Distributions, created or reviewed by experts and made available in easy-to-use form to end users building risk analysis models.
High-Speed Simulation Optimization
Once they've modeled uncertainty, users want to make choices for the factors that are under their control, to achieve best results over the range of future scenarios explored by Monte Carlo simulation. This is an optimization problem, and other vendors of simulation software have sought to address it by providing "simulation optimization" tools in their packages. But these tools have been exceedingly slow, and capable of finding solutions only for very limited-size models.
In late 2006, Frontline introduced the Premium Solver Platform V7.0, and in early 2007, Frontline introduced Risk Solver V7.0 for Excel -- its own powerful, interactive simulation software with a full GUI, charts and graphs. Premium Solver Platform and Risk Solver work together to empower users to solve sizable simulation optimization problems. The potent combination of the Premium Solver Platform and its multiple Solver Engines with Risk Solver is frequently ten to a hundred times faster for simulation optimization than other products for Excel.
Robust Optimization and More
Simulation optimization is just one way to solve optimization models with uncertainty. In October 2008, Frontline introduced Risk Solver Platform -- the first commercial product to implement new methods of robust optimization, as well as classical methods for stochastic programming. Users can now define a model with uncertainty once, in a simple form upward compatible from the Excel Solver, analyze its structure automatically, and solve it using any applicable method.
For linear programming problems with uncertainty -- very common in practice -- the Platform's new solution methods can be as much as a thousand times faster than the simulation optimization methods in other products for Excel. Even more important, users can now define models that include recourse decisions (made on a "wait and see" basis, once the uncertainty is resolved), and solve them using robust optimization and stochastic programming methods. Such models -- which often yield superior solutions with far better objective values -- can't even be defined in other products for Excel, let alone solved.
Multi-Core Optimization and Simulation
In 2009, Frontline was again first to offer a wide range of parallelized algorithms, exploiting modern multi-core processors to solve challenging nonlinear optimization, non-smooth optimization, and simulation optimization problems faster than ever, in Risk Solver Platform V9.5 for Excel, and in Solver Platform SDK V9.0 for application developers working in a programming language. Frontline has not only parallelized algorithms such as Multistart methods for global optimization and Branch & Bound for mixed-integer nonlinear problems -- it has embedded parallelism deeply into its Polymorphic Spreadsheet Interpreter, which exploits multiple cores to compute derivatives (gradients and Hessians) for Excel formulas in parallel -- greatly speeding nonlinear optimization.
Solving Excel Models Outside Excel
In 2010, Frontline took another major step, to meet customers' needs to deploy their Excel-based optimization and simulation models to remote users: By adding XML parsing facilities and embedding its Polymorphic Spreadsheet Interpreter in Solver Platform SDK V10.0, Frontline made it possible to load Excel workbooks and solve them anywhere -- on the desktop, on Intranet or Web servers, or "in the cloud" -- even on Linux. The SDK solves conventional optimization models, Monte Carlo simulation models, and even stochastic and robust optimization models "at runtime." This gives customers unprecedented flexibility to develop and deploy applications.
Making Technology Easier to Use
In 2011, Frontline focused on making its advanced technology even more accessible to business analysts without a formal background in optimization or simulation. Guided Mode, first introduced in 2010 and enhanced in 2011, takes users step-by-step through the PSI Interpreter's automated model diagnosis and transformation, automatic Solver Engine selection, and solving of the problem. A new Constraint Wizard helps new optimization users define constraints, while a new Distribution Wizard helps new simulation users select probability distributions, automatically analyzing their formulas and data behind the scenes. And work begun on the Evolutionary Solver came to fruition in 2012.
GPUs, Web Services and More
In 2012, the Evolutionary Solver was extended to include both a full set of genetic algorithm based methods, and a full set of methods from tabu search and scatter search -- while also using classical linear, nonlinear and derivative-free optimization methods. Frontline also pioneered the use of GPU (graphics processing unit) technology to achieve massive parallelism in the solution of challenging non-smooth, non-convex problems in the Evolutionary Solver. Risk Solver Platform V12.0 was enhanced with the stochastic decomposition algorithm, used to solve the largest stochastic linear problems ever solved, on any platform. And Solver SDK Platform V12.0 was enhanced with a Web Service API and the ability to "solve models over the wire" -- with cloud-based servers, and clients from Excel to web browsers and mobile phones.
Solver for Office 365 and SharePoint
With Office 2013 and SharePoint 2013, Microsoft introduced a new way to create "Apps" based on Web technologies, that can run in both desktop Office and its "Web App" counterpart. Frontline was again first to introduce an "analytics" App, the Solver App for Excel 2013, Office 365 and SharePoint 2013, which appeared in the Office Store in July 2013. The free Solver App provides the same solving power offered by Solver in Excel, which Frontline had developed for Microsoft more than 20 years earlier -- but it can be used in the Excel Web App on a tablet or phone, or in any Web browser. Models in the Solver App are solved by Frontline-operated servers running on Windows Azure, leveraging Solver SDK Platform's Web Service API.
Multi-Dimensional Modeling, Google Sheets
In December 2013, Frontline introduced its Solver Platform V2014 product line, again leveraging PSI Technology to offer a powerful new set of tools to build multi-dimensional models in Excel -- giving spreadsheet modelers capabilities previously available only in specialized modeling languages or high-end business intelligence tools. Users can create multi-dimensional cubes, based on either PivotTables or regular cell ranges, use Excel formulas to perform multi-dimensional calculations, and show calculated results in either PivotTables or regular cell ranges. What-if and sensitivity analysis, optimization models, and simulation models can all take advantage of multi-dimensional modeling.
In Fall 2014, Frontline introduced three analytic tools as free add-ons for Google Sheets, the popular web spreadsheet: Solver for optimization, Risk Solver for simulation and risk analysis, and XLMiner Analysis ToolPak for forecasting and basic statistics. Also new were Risk Solver, XLMiner Analysis ToolPak, and XLMiner Data Viz apps for Excel Online.
2015: RASON® Modeling Language in JSON
In 2015, Frontline was first to introduce a general-purpose REST API for analytic models. Unlike other APIs, RASON® -- RESTful Analytic Solver™ Object Notation -- is a full modeling language for optimization, simulation/risk analysis, and (soon) data mining, embedded in JSON (JavaScript Object Notation). RASON software makes it 'radically simple' to use and solve analytic models in Web and mobile applications. And Frontline engineered a way to automatically translate many common Excel-based optimization and simulation models into RASON form, instantly turning them into web and mobile apps.
2016: XLMiner SDK, AnalyticSolver.com
In Spring 2016, Frontline introduced XLMiner SDK, a complete Software Development Kit for data mining and machine learning, with a high-level Object-oriented API in C++, C#, Java, R and Python. In Fall 2016, Frontline was first to introduce a complete, cloud-based, "point and click" end user platform for advanced analytics that interoperates with an equivalent desktop software suite, accepts Excel workbook models, and empowers business analysts to create and solve their own models, without any programming -- using Frontline's own cloud-based RASON server as its computational engine. By year-end, AnalyticSolver.com had over 30,000 user accounts.
2017: Analytic Solver: Models in Power BI, Tableau
In Spring 2017, Frontline introduced Analytic Solver®, its comprehensive Excel add-in for analytics, with all the optimization power of Premium Solver®, all the simulation power of Risk Solver®, and all the data mining and machine learning power of XLMiner®. Fall 2017 was an exciting time with the launch of capabilities to create live models in the form of Power BI "custom visuals" and Tableau "dashboard extensions" -- culminating in a "main stage demo" for 14,000 enthusiasts at the Tableau User Conference in Las Vegas. New versions of Solver SDK and XLMiner SDK brought new analytics power to professional developers.
2018: Data Science Workflows, Solver Academy
In Spring 2018, Frontline enabled Analytic Solver users working in Excel to easily create multi-stage data science workflows, using either “drag and drop” icons connected via arrows on a visual canvas, or a "record macro" feature for Excel-based actions -- bringing the power of more expensive tools (claiming to "go beyond Excel") directly to Excel users. In the same year, Frontline introduced Solver.Academy, a new educational platform with online interactive courses about optimization, simulation/risk analysis, data mining and machine learning.
2019: Analytic Solver Cloud, Business Rules
In May 2019, Frontline was first to introduce a "modern Office add-in" version of its leading Excel product, called Analytic Solver® Cloud version -- rewritten from the ground up to use web hosting, HTML and JavaScript API to interact with Excel. Analytic Solver and Analytic Solver Data Mining were two of the very first Office add-ins for analytics to be published in Microsoft AppSource, where any Excel user -- using Windows, Macintosh, or Excel for the Web -- can easily "insert" and use them with a few mouse clicks. In the same year, Frontline entered the "business rules" market, with tools for building rule-based decision tables, compatible with the popular DMN standard, in both Analytic Solver and RASON.
2020: Broad Support for Power Platform
In early 2020, Frontline introduced RASON® Decision Services, a major enhancement of its Azure-hosted analytics REST API, several years in the making, with broad support for Microsoft's Power Platform -- including Power BI, Power Apps, Power Automate (formerly Microsoft Flow), and Dynamics 365. This release also introduced multi-stage analytics decision workflows -- going well beyond data science workflows to include optimization, simulation, business rules and more -- and rich data connectivity support for the Common Data Service, CData Cloud Hub, OneDrive for Business, and OData endpoints. In late 2020, Frontline introduced a major enhancement of Analytic Solver for Excel and a new RASON release that unified these two offerings -- enabling, for the first time, point-and-click deployment of Excel-based analytic models to the RASON cloud, with model versioning and management.
2021: Collaboration and Cloud Security
In January 2021, Frontline introduced another major enhancement of RASON Decision Services to enable users to ‘publish’, manage and govern analytic models as RESTful decision services, easily update them with new data and solve them, with exceptional security, privacy, and new collaboration and governance features, for firms using Microsoft Azure with teams of analytic model-builders. In the same year, Frontline introduce a new version of Analytic Solver for Excel with support for custom functions that users can define either with Excel's new LAMBDA and LET functions, or with DMN 1.3 compatible “box functions”.
And There's More To Come...
Frontline has exciting plans for further, industry-leading capabilities for predictive analytics, risk analysis and optimal decision-making under uncertainty, using quantitative models inside and outside Excel. Check our Blog and our Press Releases page for the latest news.