Is ModelRisk the right choice for me?

Risk analysis is a diverse discipline and there are plenty of software tools on the market that the risk analyst might consider using. We do not want you to buy a ModelRisk license and later regret the purchase because it was not the right tool for you. As risk analysts ourselves, we also use a wide range of tools. The following is a brief and hopefully objective guide to help you decide if ModelRisk is what you need.

If you feel that we have missed anything, there are errors, or we haven't been sufficiently objective, please contact us and we will adjust the content of this page.

The areas covered below are:

 

Monte Carlo simulation in spreadsheet [top]

ModelRisk is primarily a Monte Carlo simulation tool for use in Excel, and as such it is up against quite some competition. The most established competitors to ModelRisk are @RISK from Palisade Corporation, and Crystal Ball from Oracle Corporation (formerly developed by Decisioneering, inc). ModelRisk is similar in approach to @RISK in that it offers user-defined functions, rather than Crystal Ball, which uses a layer above Excel.

ModelRisk is less expensive than either of these competitors, and offers an enormous range of features far beyond what the other tools are capable of - we have written 101 reasons why ModelRisk is the lead product. So, if you are trying to choose between Crystal Ball, @RISK and ModelRisk or any other Excel-based Monte Carlo simulation tool, we believe that ModelRisk is the best choice.

We wouldn't have gone to the trouble to write ModelRisk if we didn't think we could do a much better job. If you are considering changing from one of these products, or have a maintenance fee looming, then contact us and we can propose a good deal for you.

If you don't have the cash to buy a commercial Monte Carlo add-in to Excel, and are looking for a simple tool, there are a couple of products in the market that are free. The best of the bunch we know of is SimulAr, written by Professor Luciano Machain. It is a teaching tool, so it isn't sufficiently sophisticated for professional risk analysis work, but it does provide most of the analytical features that @RISK and Crystal Ball offer and is a good tool for learning the basics.

If you don't have the cash because you are a student, then we can help. Please see our academic license page for details.

 

Sharing the model results with others [top]

A common problem for risk analysts is how to present the results of their models. Static presentations like graphs and tables of simulation statistics are useful, but they don't make the most of what a risk analysis can do to help decision-makers and other reviewers. Our approach is very different from our competitors in that we offer a free ModelRisk Reader. Please send us an email to receive this free tool.

Once the reviewer has this tool installed, all the analyst has to do is send over the simulation report file and the reviewer can analyse the results, edit, save and create different reports in the same interactive way as the ModelRisk user. This has the added advantage that the results can be widely distributed without revealing the model or any proprietary information it might contain.

 

One-off risk analysis requirement [top]

If you have a one-off need to do a risk analysis, then it is probably not very efficient to buy a software product that you have to learn, as well as the potential for messing up the math and the longer time it will take to build the model. We can build your model for you, help you understand and use it, and you can purchase a short-term licence from us if you need to run the model yourself and try out different strategies.

If you only need the model for a month, then a demo copy of ModelRisk may be all you need - it's a fully working version and it doesn't have any 'Demo' watermarks to ruin your report graphs.

 

Optimizing your decisions [top]

ModelRisk incorporates OptQuest from OptTek, which is the most sophisticated simulation optimizing software on the market. OptQuest is fully integrated into ModelRisk and Excel, making it easy and intuitive to figure out the best strategy for the future based on all the uncertainty you face. The OptQuest product has a large number of built in algorithms and the ModelRisk implementation has been designed to facilitate OptQuest's ability to select the most efficient algorithm possible. If you are performing optimisation of a Monte Carlo model in a spreadsheet, then we believe ModelRisk is ideal for you.

Most of our competitors offer some sort of optimisation tool. Crystal Ball has an implementation of OptQuest, although it is not built into the spreadsheet model. Palisade Corporation has RiskOptimiser which complements its @RISK product, but this is a very inefficient genetic algorithm tool with controls that are difficult for a normal user to understand. The free Simular software even includes an optimizing feature by linking to Excel's own Solver add-in which will solve simple problems.

Spreadsheet models are by their nature 'black boxes' to an optimiser add-in. If your model is of a particular well-defined algebraic form (e.g. a generalised linear model, or some financial portfolio return) then there are plenty of specific tools on the market that will optimise far more efficiently because they can take advantage of the algebra, for example GAMS (which also uses OptQuest), or the very recent Microsoft Solver Foundation.

If you are optimizing a spreadsheet model with no uncertain elements then Excel's own Solver will often do a good job. The OptQuest tool in ModelRisk can also be used but will deal with less well-behaved models (for example, when there is a jump in the value to be optimised because of an IF function).

 

Markov Chain Monte Carlo [top]

ModelRisk has Markov chains tools, and runs Monte Carlo simulations, but it's not the same thing as MCMC, which is a Bayesian method for assessing the joint statistical uncertainty surrounding parameters of a probability model fit to data. MCMC is surprisingly simple considering its flexibility and the complexities of the problems it can analyse. The best tool is still WinBUGS, which is free, and the GNU community has developed add-on tools to make it easier to use with other software - for example, R2WinBUGS which connect 'R' and WinBUGS.

 

Project risk analysis [top]

Project risk analysis is about figuring out the uncertainty about how much a project will cost, how long it will take, and the risks that jeopardise the project and how to manage them. A natural place to start such an analysis is to build up from a project schedule model with software like Microsoft Project, or Primavera (now owned by Oracle). However, project plans built in these tools tend to be very detailed which contrasts with the more global view needed for risk analysis. Primavera does have an extra risk analysis tool.

There are also quite a few risk analysis add-ins to MS Project, but MS Project is not well set up for add-ins of this nature. We have found that spreadsheet models of some 30-50, maybe up to 100 tasks is good enough, particularly when distribution estimates of task durations and costs tend to be subjective estimates (i.e. informed 'guestimates' rather than based on precise analysis of data). Couple the spreadsheet model to a simple database of risks and you will have a good model that can be easily understood. Correlation between task durations and/or costs is very important, and worth focusing on more than replicating detailed GANTT charts.

The interaction between risks is also very important - for example that one risk may have a higher probability of occurrence if one or more others risks eventuate, which can cause a cascading effect that has ruined many a project. ModelRisk has risk event distributions, offers a wide variety of correlation tools, and is therefore suitable for this purpose.

 

Statistical analysis [top]

A great deal of statistical analysis is concerned with hypothesis testing, which is not of general concern in risk analysis. ModelRisk has most of the underlying technical functions needed for hypothesis testing, but the software isn't set up to do them automatically. Thus, if hypothesis testing is your focus, and you don't want to build up the test logic yourself, then we recommend choosing a statistical package instead.

Model fitting is a key feature of ModelRisk - fitting distributions, copulas (types of correlation structure) and time series models to data. Some of our competitors include fitting of some distributions, but do so in a less thorough manner. None include statistical uncertainty about the fitted parameters, nor fit correlation structures or time series. If this is your focus, we think that ModelRisk is well-suited to your needs.

 

Probability modelling [top]

ModelRisk has a huge range of tools to calculate probabilities. Probability calculations can be somewhat confusing, so our software offers a lot of graphical interfaces to help you understand and communicate just what the calculations are about. ModelRisk can also do very precise numerical integration and summations which are a great help in a wide variety of probability calculations. Excel offers some probability functions, but they have come under considerable criticism over the years for their inaccuracy. ModelRisk uses the best available algorithms for all its probability functions, and does not rely on any Excel probability functions.

 

Simulation of a system [top]

Monte Carlo techniques are used in simulation tools to emulate the behaviour of a complex interacting system, and then predict what would happen if a change is made to the system. For example: simulating shipping traffic through a port, and seeing the effect of widening the port entrance; or simulating a production line and seeing the effect of changing the reliability of a component of the process. Whilst we have seen spreadsheet models built for this type of problem, we don't recommend it, or ModelRisk, although the ModelRisk Developers' Kit may be of value if you want to build very fast models in C sharp, for example.

There are specialist tools for this type of problem that use Discrete Event Simulation, and have the advantage that one can build visual representations of the system being modelled (so, for example, a map of the port with ships moving around). Good tools that we have experience of are: Simul8, GoldSim and Arena. Simul8 and Arena also incorporate OptQuest for their optimisation.

More information