BI Desktop—Widgets for people who would rather play than work

I discovered today, thanks to the blogger at Serious about Consulting, that Business Objects is working on a new product called BI Desktop. In a recent conversation with the folks at Business Objects Labs, I was encouraged when they said that they are taking data visualization seriously and are working on some new products that are bound to please. The blogger said that, although he suspected that I would be unimpressed with the desktop widgets that make up BI Desktop, he liked them. He referred to them as “neat stuff to play with.” He was right—I don’t like them, although I would agree that if you want to play, they might keep you entertained for a few minutes. I would love to see a well designed data visualization product from Business Objects, but in its current form, BI Desktop won’t be the one.

I don’t have anything against the concept of simple display widgets that remain on your screen to keep you informed about some important performance measure. I do object, however, to any form of display that doesn’t communicate effectively and doesn’t fit an actual business need.

What kind of business intelligence information deserves to sit on your desktop at all times in the form of an individual display widget? The only kind that qualifies displays a measure that meets the following criteria:

  • The information changes frequently throughout the day.
  • You must monitor it throughout the day.
  • It is displayed in a manner that enables you to assess performance.
  • It tells you enough about what’s going on to know if you must take action.

Let’s take a look at the sample widgets that Business Objects Labs have put on display:

Business Objects Sample Widgets

Now let’s evaluate each of the three widgets, based on the criteria above.

Total Inventory Cost

This widget is a typical gauge, such as those provided in many dashboard products. It suffers from all of the problems that are typical of these gauges: it takes up far more space than necessary to say far too little. It informs us of two facts: the cost of total inventory is $1,410,049 and this is satisfactory (based on the fact that the needle is pointing to the yellow range). Here are a few things that it doesn’t tell us, however:

  • What dollar amount of inventory would be considered good or bad? Because the gauge provides no quantitative scale, we simply can’t tell.
  • What is the cost of total inventory being compared to as a measure of performance? Is there a target? If so, what is it?
  • If you happen to be color blind, which 10% of males and 1% of females are, you probably can’t tell whether it is to the left or right that you ideally want the needle to point.

One of the biggest problems with this example, however, is that total inventory cost is not something that changes so rapidly that you need a widget to keep track of it by constantly sitting on your screen. This is a measure that you might look at once a day at most.

Inventory Breakdown

Pie charts don’t communicate a breakdown of inventory into its parts nearly as effectively as bar charts. It is much harder to compare the sizes of pie slices than the lengths or heights of bars. You might object to this statement, however, by pointing out that the percentage of each slice is revealed as text, to which I would respond, “If you need the numbers to be printed on the slices to understand the chart, what good is the image?”—a table of numbers would work better. Despite the generic problems that pie charts suffer, this pie chart in particular is absurd. What’s missing from the picture? The labels that are needed to tell us what is being measured, what each slice represents. Also, this pie chart fails to display performance information. Assuming that the sizes of the slices are changing throughout the day and that we actually know what they represent, how would we ever know if we should take action? Nothing tells us when particular conditions are good or bad, such as what the targets are for the various measures.

Revenues by Quarter

Once again, here’s a widget that would sit on our screen taking up space all day long to tell us something that we would likely only need to see once in the day. Of the five quarters, only the current quarter could possibly change during the day. The little bit that it might change in a day, however, isn’t worth monitoring. And where is the target or some other point of comparison that gives us some clue as to how revenues are performing. At least with this particular widget, we can compare the current quarter to past quarters, which is useful, but not enough. Unless we’re near the end of the quarter, comparing the current quarter’s revenues to prior quarters doesn’t really tell us how well we’re doing.

My Conclusion

The criticism that I’ve aimed at Business Objects in the past, remains valid for this future product as well. They are either developing or purchasing visualization products (for example, Crystal Xcelsius, which was developed by Infommersion) that are poorly designed. These problems in design would never occur if they stepped back from what they’re doing and asked the questions “What do our customers really need?” and “What functionality and design would effectively address this need?” When a data visualization product is needed, it cannot be developed without a firm understanding of visualization: what works, what doesn’t, and why. To date, Business Objects, much like most other BI vendors, has not demonstrated even a fundamental understanding of data visualization.

Take care,


3 Comments on “BI Desktop—Widgets for people who would rather play than work”

By Carlos Pereira da Cruz. February 27th, 2007 at 9:36 am

IMHO, using gauges in a dashboard promotes a strange way of managing, people look only for the last result, they don’t see the evolution, the trends, the pattern, the behavior of a system.

By Robin. February 28th, 2007 at 12:41 pm

Hi, visiting the Business Objects labs site i also saw the “Visual Data Analyzer”. Have you reviewed this tool? From the screenshots, it looks quite sober / lacking the visual fluff of the xcelcius product line.

By Stephen Few. February 28th, 2007 at 1:40 pm


I noticed the Visual Data Analyzer product as well, and even took a few minutes to look it over. I’ve been trying to decide whether I should comment on it now or wait until it’s released. It is clearly an attempt to duplicate the fine work of Tableau Software, but it is a poor imitation. It has the appearance of a product that only understands what Tableau does at a superficial level. The folks who built Tableau’s product are all recognized experts in information visualization. They worked long and hard to produce a well-designed product.

During my brief review of Business Objects’ product, I recognized attempts to add functionality that does not exist in Tableau (and for good reason), but in every case, it was either poorly designed or completely wrongheaded. For instance, in a regular crosstab, where values appear as text (rather than graphics), it is alright to include values for more than one level in a hierarchy (for example, product line, product family, and product), but this is not appropriate in a visual crosstab, where values are displayed graphically. Mixing bars (as in a bar graph) for various levels in a hierarchy in a single display not only presents a significant scaling problem, it produces a picture that can only be understood with slow and careful study to avoid confusion.