Billions and Billions of Reports (a la Carl Sagan)
By Ken Hilburn
November 7, 2008
Find more about:
bi
I recently came across a white paper on the "five styles of BI" and thought that would be an interesting read. As it turns out, more interesting than I expected. In this paper, the vendor (in order to protect the innocent, we'll just call them MacroTactics) made a statement regarding the performance capacity of this particular vendor's solution: 72,000 reports per hour. Let's see, 72,000 reports per hour... that would be 576,000 reports in an 8 hour day... and 149,760,000 reports per year. Wow. Who's reading that stuff?
Now, I fully buy in to the fact that applications that deal with lots and lots of data need to be hugely scalable, but what I don't buy is how this is in any fashion a measure that anyone can use to figure out if a particular BI solution is right for them. I can just imagine the requirements spec for that solution: "15.1.182.f - Solution must be capable of creating 70,000 reports per hour. Alternately, solution will be able to generate 140,000,000 reports per year." 140 million reports! Incredible. (Now, what did I do with my mini-me?)
Seriously, here's the thing. More reports is rarely the answer. We already have plenty of data and plenty of reports. What buyers and users really want is fewer reports and more information that helps them get their jobs done better and faster.
We'd encourage business intelligence vendors to think of themselves more as data storytellers than data factories churning out generic report widgets…even if they can do it at incredibly high speeds. From this perspective, you wouldn't want to hear Steven Spielberg bragging about his ability to pump out a dozen movies a year or J.K. Rowling trumpeting her ability to write 1000 pages a year (hmm, wait a sec).





2 comments
michael said:
I work on the supplier side and this has become an initiative. The issue here though is what do we really consider a report. For some of my clients I've almost fully automated the "reports" but, really these are just data dumps. The just want the all the data points summarized in an excel workbook. Useful eh?
But for other clients we provide written reports. The problem we're facing now is that many of our research teams are 'trained' to write their reports like a data dump, Q1 here is the distribution here is the mean. Next, Q2 here is ... This is what I'm currently dealing with. Researchers with over 10 years experience and no desire to make their product better.
Alan said:
I think you are extrapolating the figures the wrong way. A system that can turn out 72,000 reports in an hour can do 1,200 in a minute. This may still seem high, but depends on your definition of a report. For instance a monthly sales report can be distributed to 1000 people, each with different access to the data, and therefore different data content. If you consider each of these a spearate report, then you only have 12 core reports for each user to consume. When a warehouse completes a load, sending out updated report pack of 12 reports to all users is not beyond the realms of possibility.
So while I think the claims of the MacroTactics are silly, I can see a case where high performance is an issue
said:
Add a comment