BUSINESS

Home > Business > Ant farms, analytics, and the predictive organization

---

Ant farms, analytics, and the predictive organization

July 9, 2008

Who will win the World Series? How many iPhones will Apple sell this year? How many people will fly from Pensacola to Naples, Florida on any given day?

No one has a corner on what's going to happen, and the future is anyone's call. But analytic tools and techniques are making it possible to gain some measure of predictability.

The basic premise goes like this: analyze past events and current trends, and roll that into future projections to plot what is likely to happen.

Winning on probability, even to a small degree, can bring big payoffs – in terms of business strategy, knowing where your customers will be, and getting ahead of the competition.

Beating rivals at their own game

Michael Lewis' Moneyball offers up a basic lesson in predictive analytics. It's based on a simple question: "How did one of the poorest teams in baseball, the Oakland Athletics, win so many games?"¹

The answer, explains Lewis, lies in two areas. First, by understanding how teams actually score runs (through players with high on-base percentages), instead of how they're believed to be scored (players with high batting averages). Second, that it's not how much money you spend, but how you spend it.

Armed with the second-lowest payroll in the majors, Athletics' general manager Billy Beane focused on acquiring players with high on-base percentages, thus building a team that could generate runs and win games better than any other.

Google: Prediction markets and customer demand

Prediction markets are a variation on the theme. The idea is to aggregate the collective knowledge of an organization to forecast future events.

It's a trading system where people bid on outcomes they think will happen – who will be the next U.S. president, what will be our total sales next quarter.

Harvard Business Review says prediction markets can be "surprisingly accurate decision-support tools." It highlights the case of Google, which uses some 300 of them to determine customer demand for new products.²

Among the questions employees speculate on: How many Gmail users will there be on January 1, 2009? When will the first Android phone hit the market?

This kind of analysis puts many minds to work to understand and anticipate market trends in a fast-changing environment.

DayJet: Where people will want to go

DayJet is a fast-growing air taxi service centered in Boca Raton, Florida. Its business model is very different from a commercial airline. Instead of offering scheduled flights, the company allows passengers to choose when and where they want to go.

To ensure there are planes and pilots available to serve its "on-demand" market, the airline uses a complex analytic process to predict how many people will want to use an air taxi on any given day in a given place.

It consists largely of simulation modeling and "ant farming" (a form of deductive reasoning: how many paths a colony of ants, or agents, will likely take to retrieve something).

"To predict how many Floridians would pay to fly from Pensacola to Naples, they start not by gathering gross-travel or population figures but by trying to simulate the decisions that hundreds of thousands of individual travelers will make," says The Atlantic.

These results are "based on average income in each city, business relations, and other factors, and are constantly tuned to reflect real data."³ In this case, predicting the future isn't pie in the sky: the ability to use data and analytics to anticipate demand is at the core of DayJet's business.

Analyzing what will be

When mining future scenarios and probabilities, organizations use different analytics strategies. One common approach is to use predictive modeling.

Predictive modeling is at the far end of the analytics spectrum – following on analytic reporting, trending, and forecasting. This type of future-focused analysis is not about what has been or what if, but what will be.

In this case, you're looking to predict outcomes by uncovering patterns, applying algorithms, and mining data.

Another way of looking at the process: it is the ability to study a result by studying its variables and examining the relationships among them.

SPSS: Predictive analytics

Companies such as SPSS support this kind of capability through visual modeling and statistical software. SPSS predictive analytics encompass advanced analysis and decision optimization tools. On the analytics side are statistical, mathematical, and other algorithmic techniques.

From there, decision-support tools such as scoring and rules engines help users determine and test the actions that will lead to the best outcomes.

BI and the predictive organization

As a business intelligence capability, predictive modeling can inform and improve performance in areas such as churn analysis, fraud prevention, demand forecasting, and new product ideas.

To build such a predictive organization, what companies need is the ability to do statistical and predictive analytics, combine the data with current and historical analysis, and communicate the results across the organization using a common platform.

Through a joint SPSS and Cognos solution, organizations can do just that. They can take advanced statistical models, analyze results, and push them to decision-makers using reports, analysis, dashboards, and scorecards.

Summary

Businesses benefit most when they understand what has happened and what will happen. BI gives them the ability to analyze past and present trends. Predictive analytics provides future insight.

The future can never be known for certain. But with the right data mining and models, it is possible to predict outcomes with some degree of success.

That can add huge value to an organization's ability to respond to changes in markets, business risks, and customer trends.

Find Out More Cognos Performance 2008: Sign up now!

---

Sources

¹ Michael Lewis, Moneball: The Art of Winning an Unfair Game, W.W. Norton & Company, 2004.

² Bala Iyer and Thomas Davenport, Reverse Engineering Google's Innovation Machine, Harvard Business Review, April 2008.

³ James Fallows, Taxis in the Sky, The Atlantic, May 2008.