Gartner Lee Models The Future

Forced to build forecasts far more complex than Excel can handle, an environmental-consulting firm foregoes more advanced BI tools and puts its faith in a startup modeling specialist.
For Gartner Lee, perhaps the most attractive feature was cost. With a list price of $999 per license (though Gartner Lee paid about $700 each for its ten seats), it certainly compared favorably to its competitors on the enterprise level. Additionally, because Quantrix is desktop-based, it exhibited a flexibility, and accessibility, that server-based enterprise modelers couldn't match.

"The learning curve was much steeper in Cognos," says Gartner's Keir. "The people in our organization doing modeling are largely economists, and Cognos seemed to require a much deeper understanding of programming. So we went the other route. Quantrix is kind of an extrapolation. If you know how to use Excel, you can probably climb into Quantrix rather easily, which is probably not the case with Cognos."

And, of course, maintaining the Excel status-quo wasn't an option for Gartner Lee, either. The Quantrix calculation engine uses a technology similar to an OLAP cube: its formulas are built in three dimensions, unlike the two-dimensional Excel. Keir uses a Lego analogy: the Quantrix models are built out of many smaller modules, which are bolted together like Lego blocks. This makes plugging new variables into a modeling structure much less time-consuming, and thus much less costly.

"We like our models to run what-if scenarios," Keir says. Hundreds of thousands of variables are sometimes at play in a specific project. If a client wants to rejigger the cost of a landfill as part of a massive waste-management system -- a project Gartner is currently working on in Brunei -- along with multiple other proposed capital expenditures, "we couldn't do that in Excel without some serious visual basic programming," Keir says. "The models then become quite big in terms of memory. I think we've got one model, for a large regional government in Canada, running at well over 2 gigs in Excel. That's starting to push the limits of Excel." The size of the same model in Quantrix, on the other hand, is running in the hundreds of thousands of kilobytes, Keir estimates.

This, in turn, increases the model's calculation speed. "One model that runs in Excel takes about seven minutes to do a calculation, because it's hitting 150 million cells. In Quantrix, the model runs in under 30 seconds." And this, in turn, reduces costs. Asked if he could attach a dollar value to the savings, Keir said, "If we were had a project that cost $100,000 to develop in Excel, you'd be looking at $40,000 to develop it in Quantrix."

Next up for Gartner Lee is using a new Quantrix connectivity program to link the modeler directly to databases. Currently, Gartner imports the data by hand. "We've got a project coming up shortly," Keir says, "and we're going to see if we can wire Quantrix to Oracle." The firm also wants to deliver all of its models to its clients using the Quantrix interface, whereas now, it sometimes does the calculations in Quantrix, then imports them Excel for delivery.