The Scientific Method has been proven over centuries to be the best approach to solve any type of complex problem. Business, which is a relatively young discipline, can learn from the scientific discipline. After all, if you look carefully enough, most business problems are common across most industries.
Data is the fuel of any scientific investigation, and thanks to government open-data initiatives and enterprise resource planning, data is becoming more available all the time. Additionally, computing power is getting better and better, and that means simulations can run faster and more scenarios can be considered. The Internet and cloud computing also allows us to build simulations and distribute them globally. Simulations have stopped being endeavors of small research communities to widespread collaborative exercises distributed worldwide.
With science and technology, we can do almost anything. These are our methods for using science and technology to its fullest.
A Focus on Analytics in Business Applications
We don’t know your industry as well as you do; that’s just a given. We develop the models using the best minds in your organization, debating the system and sharing the theories of the business. We don’t build a Business Laboratories model for the client – we simply give the model that has already existed in some form within the brains of the client’s best and brightest. Our role is to bring it out of those brains and document it in a clear, accurate, and comprehensive model.
The very first step in the process is the creation of a hypothesis, a restatement of the problem, often in the graphical format of data over time. Graphics succinctly capture the particular problem to be studied, and more importantly, what is not to be studied.
The hypothesis becomes a platform for the creation of a qualitative model. As the name suggests, the qualitative model doesn’t use numbers, but rather in words and pictures describes the underlying physics of the problem at hand. This addresses the cause and effect variables of the system. This is often in the form of process flow diagrams, causal diagrams, and simple graphic illustrations.
These graphic depictions tell the story behind the model, and our clients tell us that proving the process behind the model with this visualization is just as valuable as the model itself.
The qualitative model then becomes the blueprint for a quantitative model. We will use data and equations to “layer” onto the qualitative model to make it “come to life” analytically.
With the quantitative model complete, we work with your team to conduct experiments. This is the analysis stage of the project. Analysis involves changing parameters in the model to map to real world policy (replace supplier A with supplier B, relocate facility 37 to Bolivia, etc.). Many experiments can be conducted in “real time” with the project team present – some may require offline runs. In any case, all simulation runs will be captured for recommendation support.
It is during this analytical stage that our clients can run experiments on their business without breaking anything or risking any mistakes. We can make changes to the environment and test it against future scenarios where governments or competitors do different things.
Beyond the Question of the Day
Our clients eventually lose their smartest folks to retirement, and that wisdom can be lost if they have not captured that knowledge in a systematic way.
Modeling allows our clients to go beyond answering the question of the day to understand the wonderful amount of learning and knowledge transfer that comes about almost as a byproduct to solving the problem using simulation.
The newest members in the workforce learn by doing and experimentation. They are already comfortable with simulation models; after all, they have played video games! Learning via simulation model is a very comfortable way to transition the knowledge from generation to generation.
The model now becomes a learning platform. A user can learn by “playing with” the simulation, allowing them to make all the marketing, hiring, cap-ex, etc. decisions, and with a click of a button, compute the future implication. It’s like a “management flight simulator.”
Additionally, models can be integrated with existing ERP database technologies, including SAP, PeopleSoft, Oracle and others. In this way, the model always has an up-to-date view of the reality of the business, and a life of its own long after the single project. It becomes an institutional capability.
The Math of Business Decision Making
One way that Business Laboratories operates is to use mathematical methods that have been recently advanced even in the last couple of years. For example, one area of work is Monte Carlo simulations – a class of computational algorithms that rely on repeated random sampling. Monte Carlo methods are useful for modeling phenomena with significant uncertainty in inputs – for example, business risk. (We’d explain further, but we could go on for pages. We’ll just sum-up by saying that Monte Carlo algorithms are a very powerful tool, and kinda cool to us math-geeks, when used in simulations for business. You can find out more information on Wikipedia.)
Sometimes, however, the issue is not risk, but competition. In that case, simulations often call for the use of game theory. Game theory is the science of strategy and how we can find quantitative results to “softer” competitive and cooperative business decisions.
Agent-based modeling is best used in situations where you are dealing with multiple agents who have a given set of preferences, reactionary rules, decision tendencies, etc. For example, if you’re trying to model the behavior of thousands of customers. In agent-based modeling, you can throw thousands of agents into the modeled environment to see how they behave.
Nonlinear programming and optimization allow us to consider the cost of utilization of resources with non-linear costs. For example, a truck in a factory obviously costs you fuel costs, which are generally a fixed rate per mile. But you also factor in the hourly rate for the driver, the projected maintenance costs (which increase exponentially with age,) insurance, and depreciation (which increases logarithmically), and all of a sudden it can become a lot harder to determine the true cost of that truck. With mathematical modeling, we can help find the answers to these questions, and indeed, help you make many decisions, like, for example, how long to keep a truck in service before retiring it, or even which model truck is the best to buy for your needs.
Finally, and perhaps most impressively, we use neural networks – software that learns. Neural networks use networks of data, utilize performance of other nodes on the network in order to generate increasingly more accurate calculations. It’s software that gets smarter the more you use it. It’s really neat stuff.
The High Fidelity User Interface: Visualization
Visualization is a key part of all the models we do. So, for example, if we’re building a simulation of a trucking logistics company, you will see trucks in the simulation. Even if we’re building a model of problems that are less physical and tangible, we still use visualization to model these problems. We have modeled criminal justice systems, IT departments, etc. visually.
We do this because visual models can clearly show the complex relationships, parameters, and behaviors to non-technical, non-mathematical audiences (which make up the majority of business professionals and executives). When concepts are illustrated visually, the human brain can naturally recognize patterns emerging. We take advantage of this in our approach to visualization.
Additionally, visual models allow us to take advantage of what we’ve termed “data adjacency.” If you can combine the simulation that plays out along with the goal or ideal outcome, visually, you can compare it to a baseline or target goal. (For an idea of what we mean by this, consider the computer graphics for the first down market when watching any major football game, or the bar which shows the pace for the world record holder in swimming or track and field at the Olympics.)
One unexpected tool that we’ve found is that technology developed for 3D computer gaming as an engine for creating simulation visualizations. One of the programs we’ve had a great deal of success with is Second Life; using the programming interface available in the game to create business problems. As a bonus, the Second Life interface can be accessed from anywhere with an Internet connection, so it allows us to collaborate with our partners and clients much more easily than if we ran a 3D visualization service in-house.