While most business software products have merely added bells and whistles during the past few years, Business Forecast Systems Inc.'s $995 Forecast Pro XE 4.0 has gained capabilities that yield genuine productivity improvement in the complex and crucial business task of time-series analysis—a job that's too often tackled with calculators or spreadsheets using simple straight-line curve-fits. Slated for release this month, the software still fits on two floppy disks.

In tests of the final prerelease version of the professional edition, PC Week Labs found solid features to guide rapid e-business expansion but also found a continued lack of attention to obvious usability issues. A standard edition, with a smaller range of statistical tools and less capacity to automate the processing of multiple forecasts, sells for $595.

Since its release as a DOS-based product, Forecast Pro has offered automatic guidance in applying a rich and sophisticated arsenal of both rigorous algorithms and proven rules of thumb to predict future time-series behaviors in tasks such as inventory management. Last year, Forecast Pro outperformed four other automated forecasting tools and all but one of 19 expert teams in the M3 competition sponsored by the International Journal of Forecasting. (Results from the competition can be found at www-marketing.Wharton.upenn.edu/forecast/m3-competition.html.)

In Version 4.0, BFI has closed some mathematical loopholes found by users. Many variable values have a bell-curve distribution, but bell curves have infinite tails at both ends and therefore imply the possibility of negative values. Some forecast variables, such as hit rates on a Web site, have no meaning at negative values, and Forecast Pro XE therefore lets the user decide what's appropriate.

The product does not provide the level of automation that one might logically expect in handling negative values that may have meaning, as when product returns are treated as negative sales. Opening a data series with negative values will trigger a warning of a "nonpositive series" but will not produce a default behavior of allowing negative forecasts—or even prompt the user to consider enabling that option. Usability problems such as this have plagued this otherwise fine product through several versions.

When the user knows enough to manually adjust a forecast, the 4.0 update streamlines the process by enabling direct visual adjustment of points on a graph as well as providing the spreadsheetlike editor found in previous versions. Group operations, such as increasing all forecast values by 10 percent, can also be easily done in the new visual editing mode, which treats all changes as tentative until the user gives a commitment command. At that point, related curves, such as confidence limits on a forecast, are immediately updated.

As with previous versions, however, we were troubled by the lack of an automatic transcript of our manual adjustments. Also still falling short of our expectations are the product's data import facilities from common spreadsheet files: The user must still build a spreadsheet with a specific layout, rather than simply naming ranges to be used as forecast variables.

Executive Summary: Forecast Pro XE 4.0

USABILITY  C  CAPABILITY  A  PERFORMANCE  B  INTEROPERABILITY  C  MANAGEABILITY  B  Setting up a powerful Web- and e-mail-based customer service operation is neither difficult nor outlandishly expensive with RightNow Technologies' RightNow Web solution for CSM. The software learns as it goes and quickly streamlines customer interactions. Short-term Business Impact: Within minutes of installation, Forecast Pro can be put to work transforming business data archives into sophisticated forecasts that capture both trends and cycles and that aid in identifying risks. Long-term Business Impact: Users who gain familiarity with Forecast Pro's many options will be able to explore many interpretations of past data and will be able to offer managers increasingly specific predictions. Immediate access to automated aids without loss of manual control. Limited interoperability with standard business spreadsheets; inadequate record of manual adjustments.