What Are You Telling People?

A person might reasonably wonder about a headline like “Small Business Use of Internet to Grow Slightly.” My first reaction was something like, “Perhaps, but does a business that does use the Internet effectively stay small?” If I survey “small business” use of the Internet in Year 0, and again survey “small business” use in Year 1, might I overlook firms that started small but made lucrative use of Internet technology—and consequently failed to qualify for my “small business” sample the second time around? Absent detailed information on sample strategy, an artifact like this could easily slip by.

Ive lately had similar suspicions about continued popular focus on the Dow Jones Industrial Average as a measure of financial markets vigor. The DJIAs distortions of overall market action are legion, and Ive been trying to train myself to look instead at broader and less oddly weighted measures like the Wilshire 5000—but my brain has decades of DJIA back story in it that make the DJIA feel like a quicker and stronger signal of “what happened today?” If I cant get a daily update on DJIA numbers and trends, Im less well-informed—the DJIAs defects notwithstanding.

Issues like these ought to be part of the discussion whenever someone starts to talk about writing a business intelligence application or implementing a BI “dashboard” for management, sales, operations, or any other discipline or constituency in an enterprise. Its incredibly easy to measure the right thing in the wrong way, or measure something thats objectively useful but not usefully familiar to the people who are supposed to benefit from the information.

Even simple decisions like the manner of presenting performance charts require thought about why the graph is desired. If youre concerned about upward or downward fluctuations around some target level of some measure, thats one thing, and a linear scale with a vertically centered horizontal axis will give you prompt warnings. If youre concerned about maintaining growth rates, a logarithmic vertical scale should be used; if youre trying to get a sense of what may be a non-linear relationship between two variables, a log-log plot is almost always the best way to go. If you merely take the default choices of a common business graphics tool or a charting component in a Web site widget collection, youre likely to get ugly and uninformative results.

If it doesnt “inform,” youre on thin ice if you call it “information technology.” The application developer should not be merely a conduit between poorly chosen measures and poorly designed presentations of whats happening in the real world. Developers should be bold enough to ask about alternative data types and sources, and should take their professional duties seriously enough to educate themselves on the options their tools offer and the interpretations their users are likely to make when an application tries to tell it like it is.

Tell me what stories youve heard lately at peter_coffee@ziffdavis.com.

Check out eWEEK.coms for the latest news, reviews and analysis in programming environments and developer tools.