A New Era in Optimization! Breakthrough Product Optimizes Models with Uncertain Factors!

Want to get the best out of a risky situation? Look no further than RISKOptimizer, revolutionary new DecisionTools product. RISKOptimizer combines the advanced Genetic Algorithms of Evolver with the power of @RISK's Monte Carlo simulation engine to optimize models that include uncertain, "stochastic" factors. No other package available can provide the solving power of RISKOptimizer!

Simulation with Optimization!
RISKOptimizer is the simulation optimization add-in for Microsoft Excel. RISKOptimizer combines the Monte Carlo simulation technology of @RISK, Palisade's risk analysis add-in, and the genetic algorithm optimization technology of Evolver to allow the optimization of Excel spreadsheet models that contain uncertain values. Take any optimization problem and replace uncertain values with @RISK functions that represent a range of possible values. RISKOptimizer runs an optimization of simulations, finding the combination of adjustable cells that provides the best simulation results! RISKOptimizer finds solutions quickly and is easy-to-use. Anyone familiar with @RISK and Evolver will jump right into RISKOptimizer with ease.

The RISKOptimizer Advantage
RISKOptimizer combines the advanced genetic algorithms of Evolver with the power of @RISK's Monte Carlo simulation engine to give you a revolutionary system for optimization under uncertainty! No other package available can provide the solving power of RISKOptimizer. Replace uncertain values in your optimization problems with distribution functions, or select values to optimize in an @RISK model. RISKOptimizer runs a fast, efficient optimization of simulations, finding the best combination of parameters to maximize or minimize any value in your model.

RISKOptimizer was developed by the team that brought you @RISK and Evolver. Instead of tying the two separate products together with a clumsy interface, we have integrated them into a unique new product. The result is a package that combines the best aspects of @RISK and Evolver -- fast simulations, accurate optimization, intuitive user interface - and is as fast and easy to use as either package!

Why RISKOptimizer?
Maybe you already use @RISK for Risk Analysis and Solver or Evolver for your optimization problems. Why would you need RISKOptimizer? Put simply, RISKOptimizer handles problems that no other program -- risk analysis or optimization -- can solve. In fact, RISKOptimizer can solve problems you didn't even know you could solve. Let's look at two examples.

Optimization Add Simulation!
Standard optimization programs are good at finding the best combination of values to maximize or minimize the outcome of a spreadsheet model given certain constraints. However, these programs are not set up to handle uncertainty. Add any "uncontrolled" uncertainty and traditional optimizers just can't move to an optimal solution. They get lost in the permutations of possible values and changing solutions.

Here's a simple example: Suppose you have several factories and want to find the best locations to manufacture different products to meet demand in nearby cities. You want to maximize profits and minimize shipping costs.

This is a straightforward optimization problem where you want to assign manufacturing volume, by product, to different factories. You begin to set up your model...but then realize that several key factors -- shipping costs, demand, manufacturing costs, etc. -- are all uncertain. There's no "best" combination of factors; these are all factors outside of your control.

Standard optimization programs can't handle uncontrolled factors. Traditionally you would have had to guess at the uncertain factors and hope for the best. But simulation is designed to handle uncertainty, and RISKOptimizer, with built-in simulation, can handle this type of problem easily!

Simulation... Add Optimization
Simulation programs such as @RISK use Monte Carlo simulation to account for the uncertainty in models and determine the probability of various outcomes occurring. But Monte Carlo simulation does not deal with decision variables whose values you can set. It handles random, uncertain values at a single state of those decision variables.

Suppose you are developing a new product and want to determine whether or not this venture will pay off in the long run. You build a standard spreadsheet model to calculate the profit for this venture. You decide to use @RISK to run a Monte Carlo simulation and start replacing uncertain values with @RISK functions that represent a range of possible values.

Then you realize that some of your assumptions are based on using specific vendors and production methods to construct components of your product. There may be other vendors and methods available to you that could save money. It's also possible that some production methods may make shipping costs unattractive. With @RISK alone, you could run multiple simulations and compare results -- but did you try every possible combination of inputs? To do that, you would need optimization. But in this case, optimization would need to be combined with simulation, because the results you are optimizing are simulation results!

Why Not Use @RISK and Solver or Evolver?
You may be asking, "Why can't I just use @RISK and Evolver together, or use @RISK with Excel's built-in Solver?" You can't because Solver and Evolver are designed to handle models that can use a simple spreadsheet recalculation to generate a result. For optimization under uncertainty, you need an optimizer that can minimize or maximize simulation results. A full simulation needs to be run for a possible set of values for your decision variables. That's a tough job, and that's why you need RISKOptimizer!

How RISKOptimizer Works
RISKOptimizer uses the combined power of @RISK and Evolver to solve optimization problems under uncertainty. Probability distributions from @RISK are used to model the uncertainty present in your spreadsheet, just as they are in @RISK. During an optimization, RISKOptimizer generates a number of trial solutions and uses Genetic Algorithms to continually improve results of each trial. However, unlike Evolver, each trial is an @RISK simulation! During the simulation for each trial, probability distribution functions are sampled and a new value for the target cell is generated - over and over again. At the end of a simulation, the result for the trial is the statistic that you wish to minimize or maximize for the distribution of the target cell (mean, standard deviation, etc.). For each new trial solution, another simulation is run and another value for the target statistic is generated. The result is the trial solution that provides the best answer to your problem!

RISKOptimizer doesn't replace @RISK or Evolver but works with them to solve optimization problems that include uncertainty. And that's a wide variety of problems! Just think of the problems where you need to find optimal solutions. Then, in those problems, identify values which are uncertain. In the past, you might have just guessed at values for these uncertain factors -- greatly diminishing the validity of your results. Now you can stop guessing and use RISKOptimizer to generate robust, optimal solutions!

Using RISKOptimizer
RISKOptimizer uses components of both Evolver and @RISK to set up an optimization problem:

Step 1: Set up the optimization problem (Evolver)
Step 2: Add uncertainty to the model (@RISK)
Step 3:Run an optimization (Evolver and @RISK)

Performing these steps is as easy as working with @RISK and Evolver.

Step 1: Setting up the Optimization
There are five main steps to set up an optimization in RISKOptimizer:

1) Specify the Target Statistic
Since RISKOptimizer is optimizing simulation results, the value you are trying to maximize or minimize must be a simulation statistic, such as the Mean, Standard Deviation, Variance, etc. The target cell statistic is selected from the drop down list next to the cell reference (see right). You can also select a specific percentile or target value. (For example, Percentile(.99) would select to minimize or maximize the 99th percentile of the target cell, or Target(100000) would select to minimize or maximize the probability of a value <=100000 occurring.) These statistics can also be specified in the model using new functions (see below).

2) Select Adjustable Cells
The adjustable cells contain the values that RISKOptimizer will change in order to optimize the target statistic. Multiple groups of adjustable cells may be specified, and each group may contain multiple cell ranges.

RISKOptimizer uses six different solving methods to find the optimal combination of adjustable cells. Different methods are used to solve different types of problems. The six methods are:

• Recipe - a set of variables which can change independently.
• Grouping - a collection of elements to be placed into groups.
• Order - an ordered list of elements.
• Budget - recipe algorithm, but total is kept constant.
• Project - order algorithm, but some elements precede others.
• Schedule - group algorithm, but assign elements to blocks of time while meeting constraints.

3) Define Constraints
Constraints (both hard and soft) which must be satisfied during the optimization can be specified in RISKOptimizer. Hard constraints may be evaluated each iteration of a simulation run for a trial solution (an iteration constraint), or at the end of the simulation run for a trial solution (a simulation constraint). Solutions that fail to meet hard constraints are discarded. Soft constraints -- constraints that cause a penalty to be applied if they are not met -- are calculated at the end of a simulation.

4) Set Stopping Conditions
The stopping conditions available in RISKOptimizer include settings for halting the optimization and settings for halting each simulation run during the optimization. These conditions include:

Optimization Stopping Conditions


• Number of simulations
• Time elapsed
• Change in last n simulations less than x%
• Excel formula evaluates to TRUE

Simulation Stopping Conditions


• Number of iterations
• Convergence Monitoring to stop simulating when results are stable
• "Projected Convergence" to stop simulating based on previous trials
• Specifying a Tolerance for convergence or using automatic Tolerance adjustment

5) Set Other Options
You can specify the sampling method used during the simulations (Monte Carlo or Latin Hypercube), as well as what values are displayed during a normal spreadsheet recalc (Expected Value, True Expected Value, or a Monte Carlo random value). RISKOptimizer allows macros to be run at various times during the optimization and simulation process.

Step 2: Adding Uncertainty to the Models
Uncertainty is represented in RISKOptimizer (as well as @RISK) by probability distribution functions. Any value in a spreadsheet cell or formula can be replaced by a probability distribution that specifies the range and probability of possible values that could occur.

Models in RISKOptimizer can use any of the 38 probability distribution functions available in @RISK to define uncertainty. A value of 10 in a spreadsheet cell, for example, can be replaced with the @RISK function =RiskNormal(10,2). This would specify that the possible values for the cell are described by a probability distribution with a mean of 10 and a standard deviation of 2. As in @RISK, these functions behave as standard Excel functions. Correlations between probability distribution functions can be specified with any of the @RISK correlation functions including RiskDepC, RiskIndepC and RiskCorrmat.

New @RISK Functions
A set of new @RISK functions are available in RISKOptimizer that return statistics on simulation results directly to the spreadsheet. These functions include: