How I Met Your Model

Mathematical models: You hear about them in the business media all the time, but what are they? What do they do, how do they work, and how are they are created?

A mathematical model is, plain and simple, an equation, or set of equations, that represents a relationship between two or more things. Such equations are shorthand for theories about the workings of nature and society. The theory may be supported by a substantial body of evidence -- or be just a wild guess. The language of mathematics is the same in either case.

Terms such as “predictive model,” “statistical model," or “linear model” refer to specific types of mathematical models, the names reflecting the intended use, the form, or the method of deriving a particular model. These three examples are just a few of many such possibilities.

Why use mathematical models? Couldn’t the same relationships be described using words? That's possible, yet there are certain advantages to the use of equations. These include:

  • Convenience. Compared with equivalent descriptions written out in sentences, equations are brief. Mathematical symbolism has evolved specifically for the purpose of representing mathematical models; languages such as English have not.

  • Clarity. Equations convey ideas succinctly and are unambiguous. They're not subject to differing interpretations based on culture, and the symbolism of mathematics is a sort of common language used widely across the globe by people who may share no other.

  • Consistency. Because mathematical representations are unambiguous, the implications of any particular situation are clearly defined by a mathematical model (at least, the implications associated with the theory behind that particular model).

When a model is mentioned in a business setting, it’s most likely a model used to make predictions. Models are used to predict stock prices, product sales, and unemployment rates, among many other things. These predictions may or may not be accurate, but, for any given set of values for the known factors (also called independent variables or inputs) included in the model, there will be a well defined prediction (also called dependent variable, output, or result). Mathematical models are used for other purposes in business, as well, such as to describe the working mechanisms that drive a particular process.

Some mathematical models may be nothing more than one individual’s personal theory of how something works, but most have more substance behind them. Better models arise from a mix of carefully collected data from observation, building blocks based on statistical theory and other mathematical methods, and the experience of a knowledgeable researcher. And a model cannot be judged to be good until its performance has been tested using real data -- data that was not used to develop the model in the first place.

A model represents a theory about how some part of the world functions, but the world is mighty complex. So a model that has worked well for one purpose may not be adaptable for another. Or it may prove accurate in one time period, but not in the next. Alternative theories may come to light and prove to be even more accurate. Models don't represent reality, only our best understanding of it.

Meta S. Brown, Business Analytics Consultant

Meta S. Brown is a consultant, speaker, and writer who promotes the use of business analytics. A hands-on analyst who has tackled projects with up to $900 million at stake, she is a recognized expert in cutting-edge business analytics. She has conducted more than 4,000 hours of presentations about business analytics, and written guides on neural networks, quality improvement, statistical process control, and many other statistical methods. Meta's seminars have attracted thousands of attendees from across the US and Canada, from novices to professors.

Tell Me a Story: Why Data Analysts Must Be Storytellers, Too

Data alone won't make an analyst's work memorable or actionable in the eyes of a business executive. A story puts it into perspective.

It's the Data, Stupid

When it comes to acquiring the data that will feed your analytics initiative, "free" isn't always the best approach.

Re: Key Thing
  • 6/23/2013 9:01:33 AM

Oh, the magic word - assumptions!


We ciould give a little slack here, and consider that the [original] object of data mining is to empower business users to discover patterns in data without requiring them to have statistical training. Users like that wouldn't be expected to know much about assumptions behind the procedures they use. But then again, they would be expected to field test and validate their discoveries.


No matter how you slice it, some critical evaluation of claims about excellence or untility of a any model is always in order.

Re: Key Thing
  • 6/22/2013 9:07:51 PM

Meta, I like the title of this post, partcularly in light of the comment about assumptions. I am picking up more big data books in which the authors note how much our assumptions are playng out in the models being created.  We are asking our analysis to be faster, but at the same time increasingly the likelihood that we may not reviewing how we assumed the data and model in the first place. 

Re: A danger within
  • 6/21/2013 12:59:51 PM

Meta, - great summary of why it is better to use a mathematical model. Indeed it is not really a model if its not mathematical because mathematics decomposes ideas into programmable logic. Which then enables testing for practicality.

Re: A danger within
  • 6/20/2013 4:41:16 PM

That quite common, Beth. People get attached to particular models and cling to them like security blankets.

A danger within
  • 6/20/2013 4:13:23 PM

Meta, do you think there's any danger of an analyst getting too stuck on one particular type of model? Using it over and over again, even in instances where a different model would be a better choice? 

Re: Key Thing
  • 6/20/2013 1:45:53 PM

I agree!

Key Thing
  • 6/20/2013 1:21:22 PM

The key element of any mathematical model is how it was derived and what assumptions went into it. These are really important in understanding the results the model gives and gives insight as to when the model is likely to be right and when it may run into difficulty.