Business Intelligence – Wow Factor

Posted on May 22, 2011 by David

Introduction

I my last post, I suggested that one of the reasons many BI projects fail is that the focus turns to implementing the tool, and not on solving the problem. However focusing on the tool is not all bad, you can gain insight into exploring a tool’s capability.

Today’s post will first explain technically what I did to put a “WOW” factor on my reporting. It will then give arguments on why on why it was a good idea, and then it will give arguments on why it was a bad idea. You can then decide on your own what you would do.

Technical Challange

I decided that my PHP pie chart was not satisfing for my reporting needs. To put it another way, I was still hungry for a better pie. My current PHP pie sometimes made some of the slices the exact same color. In addition, for a few random totals it filled the whole pie chart background with a color of a pie slice, making the pie picture very messy and unreadable.

Rather then fixing the PHP code, I decided to try something completly different. I googled “open source business intelligence” tools and found eclipse BIRT. I have used eclipse before to experiment with java, so I decided that I that I try BIRT out. BIRT is short for Business Intelligence Reporting Tool. There is also a Easy Reporting Nice Intelligence Engine, or ERNIE. ( nahhh,.., just kidding about that)

I was very excited to learn that BIRT could connect to my MYSQL database through a JDBC connection. I tried for awhile without success. I studied the syntax and tried again. After a lot of trial and error, I found out that my web host I use disallows “external” access to the MYSQL databases. Bummer. However, that was not the end of the story. I saw the BIRT could use XML as input, and I knew I could write to my MYSQL database from PHP, so what I needed to do was create a web service in PHP to read my database to create XML. ( Got it?) Could this be done? Believe it or not, this is really not as hard as it sounds. I thought for awhile that I would have to make this XML “point in time”, and create PHP code that writes a file to the server, but BIRT uses the PHP code as a a “bridge” and creates the XML all at once. Bingo, real time reporting.

Here is the pie chart I created from BIRT. ( This is a picture, not a report)

I also created a table so you can drilldown on the detail. This all happens in real time. Below is a drilldown when you click on one of the “webgroups”. ( This too is a picture, not a report)

So I know what you are thinking. Why didn’t you make the reports real time like the last Post on Business Intelligence? I did try, but BIRT runs primarily in a J2EE environment, and my web host does not support Tomcat. My web host run a “flavor” of debian linux. I seems technically possible to run BIRT on debian linux, but I decided that I was happy enough to create the reports from inside eclispe for now. You can also run tomcat locally on a pc, but that would not help out readers of the BuddyBlog any.

To be continued…
Look for next : Why this was good? Why this was bad? Conclusion will include the bubble chart…

Bubbe Chart?

Posted in Uncategorized | 1 Comment

Business Intelligence

Posted on May 9, 2011 by David

Introduction:
What is Business Intelligence(BI) anyway?   Have we been not so smart in business up until now?    If you claim to be an expert in this area, you are in high demand in today’s job  market. The Garter Group also claims that 80% of BI projects end in failure.  That does not sound very attractive.  You can Google this and read many opinions and powerpoints about this phenomenon. Are BI projects really this hard?  This post is going to attempt to simplify the subject ( with pictures).   I will also suggest some steps to get your project in the 20% success category.

A Data Problem:
I host a number of websites, and I have an interest in what web pages are being accessed. The hosting company has many “canned” reports about access. I have looked at them, and I have derived little value from their content.  The reports just are not the information I am seeking.  In addition, there is a daily log file.  It looks like this:

Now there is a lot of information in this log file, but it is quite complicated to interpret.  I parallel this log file with the state of information in today’s business systems.  You know the answer is somewhere, but it is hard to get.  You are probably ahead of the game if you can even define your exact question. I think what business really wants to the ability to accurately predict the future.  Business Intelligence should do this based on patterns in the data.  To get to this data utopia, is just not possible to get there by just looking at the log file. Similarly, you cannot get there in the present state of data in many of today’s business systems.
To solve my information gap dilema,  I decided that I would start my own Business Intelligence project.  I created a MYSQL database, and was able to populate it from with PHP.  If you are not familiar with this technology, that is OK, I just learned it too.  They are an open source database and an open source programming language.  The are readily available and powerful.  I created some PHP code to insert to my new MYSQL database for all the webpages in which I really had an interest.  That enabled me to have a nice baseline of dependable and quality data for my analysis.

The Result:
Once the database was established, I queried this information and created a list, again through PHP.  This information is really what I have been pursuing.  This points out what should be an important rule of Business Intelligence.  For any piece of information, you should have a potential call for action.  This report passes my test.  Here are some of the potential actionable items from my list: 1) It is worth the effort it takes to post meeting agendas for the nonprofit site? 2)   It is worth the effort it takes to post recorded church services? 3) How many people are reading the outdated content on the church web site?  4) Does action need to be taken to better promote the classic tractor newsletter?

LIVE Report From Database of Last 30 Days of Web Page Access:

Adding The Pizzazz:
I really liked my data table, but I really wanted more.  So I turned the information into something with a little pizzazz. Turning to my new found tool PHP, I created a nice pie chart.  The data presented is really not very meaningful, but it does have the wow factor.  This shows one of the dangers of BI and how many fall into the 80%.  The BI product or tool is not the silver bullet.  In other words, we should not get carried away with the tool. In addition, some of these tools have little quirks that take a long time to track down. For example, for certain denominator totals, my pie chart background will turn completely pink. If I go and access a page being tracked, which adds to the total, the background goes back to white. ( I am sure I could detect the reason, but this would take time, and I am happy with it for now.) However, this soft side factor is not something to be ignored.  You want people to use your tool, and the pie chart really catches your attention.  It was also fun to create, and I found myself changing how much data was in scope to see how the slices changed. Again, not meaningful, but interesting none the less.

The ideal state is to combine the wow factor with actionable information.  For example, perhaps I should figure out how to add a drilldown capability for each slice of the pie.  I am only interested in looking at one slice at a time anyway, and this will give me a nice colorful starting point for my analysis. That way, we can have our cake pie and eat it too.

LIVE Pie Report From Database of Last 24 Hours of Web Page Access:

Project Success:
So how do you reach BI project success?   Number one, you have to check your information against the meaning test. That is, the information being presented should call for a potential action.   Number two, since the ultimate goal of a BI project to give you predictive results into the future, you are in danger if your project plan demands to know the exact questions before you start. Since BI projects are about making the unknown known, the project plan should value people over process.  IT and Business need to work closely together for a BI project’s success.  This approach is actually very Biblical. (Yes , we should not be surprised that God has great insight into BI projects)  We need to work together for a common goal to get to the best product. Agile or Prototype methodology, where you are not afraid to take a step backward to take some steps forward, will deliver much more long term value than a waterfall methodology, where you want to know everything before you get to the next phase.  If you use a waterfall methodology, you will likely deliver what the user specified, which is not really what they want. In summary, for a successful BI project: 1) Gather the cross functional team 2) Plan a little 3) gather the data foundation 4) build answers and questions against “actionable” criteria 5) repeat until very profitable.

Next Post Time:
Combining Business Intelligence wow factors with meaningful data,  or relating the quality of bathroom caulk to the quality information systems. For now, time to take the dog out for a stroll.

Posted in Uncategorized | Leave a comment