The side effects of the information age are not pretty. The glut of data coming in has Forecasting and Market Research departments feeling like they are under siege. At the same time, the need to share and make sense out of data has become vital.

That is how it was at Gateway. Sure, the PC maker's Forecasting and Market Research departments were in the same building, but there was not an organized way to share information in a timely manner. In fact, many decisions that determined PC features and pricing were based on whatever data the particular decision maker could get plus and gut-instinct.

Tim Stoughton, Senior Manager Forecasting & Predictive Modeling, had the benefit of understanding both departments. A 10-year Gateway veteran, he had worked in both departments, and saw the need for a model that would bring all this data together.

"The model had to be able to determine which product features were most important to customers and how those features affected sales," Stoughton says.

In 2002, Gateway created a complex and interactive simulation of the U.S. consumer PC market that pulled data in from an unprecedented number of internal and external sources.

Here are five tips from Stoughton on getting a model such as this off the ground:

Tip #1. Get buy-in at all levels

Any project that takes up lots of company resources needs management buy-in. Get it early. Management is open to any idea that provides ROI, but sometimes defining ROI is difficult.

Show management how the model will help the company:
- Save money,
- Save time on analysis,
- Get a handle on complex interactions in the marketplace, and
- Make better business decisions.

Do not forget:

Management is not the only group whose support you will need. "The people responsible for the data that will feed the model have to be brought in from the start," Stoughton says.

Some data suppliers might become users of the model; for others, it is simply helpful to know how the data will be used, creating a sense of ownership. When resources are being piped into your model and other departments are waiting longer than usual for data, you need as many people as possible backing your project.

Gateway wanted the model to be useful to multiple departments, so that whoever had queries on PC market forecasting could get quick answers. It would also allow analysts to understand the market at a complex yet realistic level, rather than in abstract ways.

In other words, it would make people's jobs easier. Tell people this whenever you have the opportunity.

In selling the product to colleagues, Stoughton had to address the following questions frequently:
- Can I rely on it?
- Will it be kept current?
- Will there be continued support for it?

Stoughton's answer: "If you use it, it'll be kept up."

People were also worried that the model would threaten their jobs or replace their functionality. The model is "a decision support tool and not an expert thing by itself," Stoughton says. "It's not intended to take jobs away."

Tip #2. Make it simple, but not too simple

The team decided that it was acceptable to make the model sophisticated enough that users would need a half day or full day of training. "If it's so easy that a fool could use it, only a fool would be interested in it," Stoughton says.

"We didn't want to dumb down the interface too much, but we needed other people in the company to be able to make sense out of it, too."

Since the program went live in September, 2002, two dozen people have been trained on the model, and those people have helped others learn the system. "This work by definition is complicated. We found a way to simplify it so it could be understood without stripping away the complexity," Stoughton explains.

Tip #3. Bring in outside help

The kind of PC market model Gateway wanted was far beyond the ability of a simple spreadsheet, and it was not something that could be done internally in a timely manner.

For assistance with internal data collection and external market research, Gateway turned to Cincinnati-based Catalina Marketing Research Solutions. The model itself was put together with assistance from Salt Lake City-based specialists, The Modellers.

Tip #4. Select useful data

Although other companies have taken a year or more to build complex models, Gateway's feasibility studies for the model and the collection of data only took three to six months; and, all the input data was chosen from the point of view of Gateway's lifetime value.

"Some of our products have lifetimes on that order," Stoughton says, so the team needed to look carefully at the period of time that the data would be useful. Data that "went bad" too quickly would not be useful.

The team also realized that some factors can not currently be modeled easily or successfully enough to be worth the effort. "It can't anticipate what the economy will do," Stoughton says. "And it's not very granular in terms of feature sets, like difference if you change the speed of CPUs up one megahertz."

If you missed Stoughton speaking at IIR's "Return on Marketing Investment: The Promise of Modeling & Other Marketing Metrics" conference, February 24-25, 2003, at the Wyndham Bonaventure Resort & Spa in Ft. Lauderdale, Florida, and if you are interested in hearing more about the next "Return on Marketing Investment" event, email anicholl@iirusa.com to receive the update on when and where the next event will be held.