Mistakes are expensive.  Everyone wants a model to help them avoid mistakes and repeat successes.

We want a good business model that provides a framework for success.  If the business model works and we stick with it, we will make money.

In the chain store business, we want to make good site selection decisions.  Avoid bad real estate; pay the right price for good real estate.  We want a sales forecasting model that will help us estimate the top line number to plug into our pro forma operating model for a store (which is based on our business model, of course).

Predicting the future with a model is a specialized skill.  There are many smart people in business who can invent, adapt, and execute against business models but don’t have a clue about sales forecasting models.  That’s because people who build predictive models have training and experience in statistics, data sources, and are typically subject matter experts in their industries.

We listen to the weather forecast and make plans based on it, but most of us don’t build meteorological models.  We listen to economists and politicians talk about government models, but we don’t know how they compute them.  The technical skills of modelers create a clear distinction between the “haves” and the “have nots;” those who make the models and those who consume their output.

I prefer the term “analytics” to describe the use of data and metrics to support decision-making.  You don’t have to be a meteorologist to see if it’s cloudy or cold outside or to predict rain when a storm that’s moving east is west of you.  You don’t have to be an economist to recognize that high foreclosure rates in your state mean that the economy is soft in your area.  You don’t need to be a statistician to see that 5,000 people within 20 miles of a proposed sporting goods store constitutes a weak trade area.

There is clearly a place for models and modelers.  There are many market dynamics that can be simulated with statistical models that help identify good and bad locations.  What we need is a way to “democratize” the use of data and analysis in decision-making by putting tools in the hands of the decision-makers that make use of their knowledge and experience DIRECTLY.

We can start by referring to store location research as “analytics” rather than “modeling.”  Here is a brief list of the most frequently used “analytics” in real estate planning and site selection:

1. Look at a map that is shaded by population density or income to find good areas for expansion.
2. Run a demographic report on a proposed location and look at the values of the key variables at different distances from the site, e.g. 1 mile, 3 miles, 5 minutes, or a hand-drawn polygon based on an expert’s knowledge of the market.  Compare the numbers for the proposed location to “benchmarks” used in screening sites (e.g. 30,000 people within 3 miles, median income $75K).
3. Compare the demographics of certain trade area sizes for a proposed location to those of existing stores and find the ones that are most similar.  Use google Earth or Bing Maps to compare the similar existing stores to the proposed site to estimate sales (analog method).

These techniques would not be considered “modeling” by most people, but they are definitely “analytics.”  Few people would say that these techniques are inferior to predictive models per se; however, they rely more heavily on the judgment and experience of the people using the output.  An enlightened organization might use a both a predictive model AND these other analytics as they select and optimize markets and evaluate new sites.

According to Russell Ackoff, a systems theorist and professor of organizational change, the content of the human mind can be classified into five categories (my examples added in italics; see http://www.systems-thinking.org/dikw/dikw.htm):

1. Data: symbols – customer transactions, business locations, demographics, roads, traffic counts
2. Information: data that are processed to be useful; provides answers to “who”, “what”, “where”, and “when” questions – shaded maps, trade area summaries, distributions
3. Knowledge: application of data and information; answers “how” questions – trade area size rules, analog comparisons, local market trends from brokers, predictive models
4. Understanding: appreciation of “why” – expert critical thinking and recommendations, investment committee meetings
5. Wisdom: evaluated understanding – consensus opinions, dissenting opinions

Our goal is to always apply “Wisdom” to real estate planning and site selection decisions.  Where do “models” fit into this schema?  A good model is built using a process based on all five of the categories.  However, after it’s built, it really only functions in category 3 (Knowledge) and must be combined with steps 4-5 to become “Wisdom.”

Where do “analytics” that are not predictive models fit in this schema?  They are built and used in the same way as predictive models.  The commonly used techniques described in the list above (maps, reports, analogs) are “information” and must also be processed through steps 4-5 to become “Wisdom.”

There is something mysteriously attractive about “push button” predictive models.  Many companies focus on “getting a model” before they have even cleaned their data, put decision-making processes in place, or installed a basic real estate information system that generates reliable maps, demographics, and lists of nearby existing stores and competitors.  Sometimes it looks kind of silly, like making someone wait while you enter their phone number into your smart phone instead of writing it down.

There is a logical sequence to building better decision-making systems.  The path to Wisdom starts with data:  the right data and getting the data right.  A highly reliable system (recently called “Unified Market Knowledge System,” see http://www.tradeareasystems.com/eblast/umks.pdf) will deliver better real estate planning and site selection decisions long before a “push button” predictive model.  After the base analytics are in place, building a predictive model may be a great idea.