Open Enrollment | Subscribe to Printing Impressions HERE
Connect
Follow us on
Advertisement
 

Dickeson--The Analytical Power of OLAP

June 1998
Diapers and beer sell best after 6 p.m. After six, husbands are sent out to get diapers, and often grab a six pack for themselves. So the rack with diapers should be located next to the beer storage to increase impulse buying by men.

Web breaks increase on a web press as relative humidity in the pressroom decreases.

People on the third shift decrease the density of cyan ink on a running press form because of a loss of natural light perception incurred while they sleep in the daytime.

Are these statements true? Perhaps. Perhaps not. To find the answer, do some "mining" on your "data warehouse." Use OLAP.

OLAP (online analytical processing) is a hot topic these days, and with good reason. OLAP enables you to "slice and dice" the sales or operating data in your computer databases in an almost infinite variety of ways to find relationships that can provide support for decisions that need to be made.

A study was conducted by the GATF that identified, as I recall, more than 1,600 variables that affect press operations. Chaos theory teaches us that slight changes in any variable can have substantial impact on output—productivity. We all know this. One press manufacturer provides a database with 2,600 event codes (op codes) to analyze press operations.

When we try to quantify the impact of all those thousands of variables and take affirmative action, we throw up our hands in frustration. No way to analyze this mass of data. We revert to anecdotes: "Let me tell you what happened on press 22 last Thursday..."

Maybe we're getting a tad closer to finding significant and relevant relationships by using the analytic power of our personal computers to establish such relationships and trends. With OLAP, we take a multidimensional view of the data via MDBMS (multidimensional data- base management).

Until recently we've mostly used two-dimensional databasing (RDBMS—relational database management systems), where the elements appear in rows and columns like a spreadsheet. With the RDBMS applications, we're confined to OLTP (online transactional processing). OLTP is optimized for creating, updating and retrieving individual records, such as our accounting systems, for receivables, payables, payrolls, etc.

In contrast, OLAP is used by analysts and managers who frequently want a higher-level, aggregated view of the data. OLAP databases are optimized for analysis.

The data world of RDBMS is two-dimensional—like a flat spreadsheet. The MDBMS multidimensional world is a cube, three-dimensional—like a Rubik's Cube. With it you can drill up, down and across. The RDBMS data can become a "virtual" cube for OLAP inquiries with certain software, but with restricted speed and efficiency. Or the data can be stored as a true cube in a multidimensional server set of "arrays" that are interlinked by dictionaries and pointers.

Using the MDBMS true cubic aspect, analysis is far faster and more efficient. If this technical stuff intrigues you, have a look at these Web sites: www.pilotsw.com/ olap/olap.htm; www.datamation. com/plugin/workbench/olap/stories/virt.htm; www.picksys. com/product/whtpapr.html.

Enterprise Effectiveness
Most of the literature about OLAP deals with marketing phenomena such as the correlation between diapers and beer. That's fine.

If some of you reading this column are printing sales or marketing people, give this some thought. Think about products, materials and customer base. Could online analysis lead to conclusions about core competency of your shop? Account preferences? Market segmentation?

If so, you'd be using OLAP for insights into enterprise effectiveness. I'm suggesting that we think "out of the marketing box" about OLAP—that we use this tool to improve printing process efficiency. There may well be more usefulness in printing process management for OLAP than in product marketing.

As E.F. Codd, the dean of the database world, puts it: The relational databases "were never intended to provide powerful functions for data synthesis, analysis and consolidation (the functions collectively known as multidimensional data analysis)." Online analysis—the MDBMS—is a concept whose time has now come for print process management!

We need to aggregate, synthesize and analyze our printing process information beyond the world of our job cost systems that presently use transactional analysis of an RDBMS system. There are 1,600 variables and 2,600 press events, and more than that must be aggregated, compared and related. Trends must be found and analyzed. Covariances and correlations that lead to actions and decisions need to be identified. Unscheduled stoppages of equipment must be reduced to improve time and materials efficiency.

It's beyond my competence to do more than point a direction. I don't know how my VCR works, but I use it to keep Blockbuster solvent.

OLAP isn't really new, just newly rediscovered and touted. In 1977, I witnessed a demonstration of an MDBMS on a microcomputer system called the Microdata Reality, and I could hardly believe the ease of query and analysis I was observing. It was named the Pick System, after its developer Richard Pick, who pioneered it for the military at about the same time in the late '60s that UNIX was being perfected. Pick Systems are now running in thousands of applications worldwide.

A year or two ago a new version of Pick, called D3, was introduced for UNIX, Windows 95, Windows NT and Visual Basic 5. It can run on terminals or PC workstations and is available with open database connectivity and SQL. I don't know how it works, but I'd sure like to use it, or one of the other dozen or so systems that are now available with OLAP capability, to analyze our printing processes.

—Roger V. Dickeson

About the Author
Roger Dickeson is a printing productivity consultant based in The Woodlands, TX. He can be reached via e-mail at rogervd@bigfoot.com or rogervd@pdq.net; or via fax at (281) 362-7572.
 

COMMENTS

Click here to leave a comment...
Comment *
Most Recent Comments: