Better BI on Bigger Data
September 1, 2010
“Can we do BI without a significant infrastructure effort?”
What’s the most common BI tool? Microsoft EXCEL! If you are considering a DW initiative, you are likely encountering resistance from your hard-core, Excel based, analysts. Taking away Excel from them means that they will have less flexibility and spend more time waiting on changes. I’m guessing that they may not be on board with your vision.
1010Data is a zero-footprint, browser based data warehousing product. It has an Excel like feel, but it can handle billions of rows of data (not thousands like Excel). It’s fast, flexible and requires no infrastructure support from IT (if you are in IT, my guess is that you may have already stopped reading this article). The system is not an OLAP system, it requires no design, and not even indexing strategies – but it’s so fast!!
This is a new paradigm shift: from ETL to ELTAR – extract, load and transform as required. Hmmm….cleansing and summarization at query run time; views are used for data governance.
Don’t think it’s true? Dollar General is using the platform and was up and running in 5 weeks to 115 users with a 100% ROI.
What could you do for your customers if you could have their DW up in just a few weeks instead of a few months??
- Jodie
Data Mining – We want it Now!
September 1, 2010
“We are drowning in information but starving for knowledge.”
What do you really understand about data mining? Dr. Candace Gunnarsson was formerly a professor of statistics at Xavier University. Her view of data mining is all about getting information out of the systems you have today and predicting future results. Her view is also about needing a DW modeled in a way that makes the mining experience easier and more meaningful.
Her view is that there is prescriptive (automated) data mining and descriptive data mining. I think most people think of data mining to be an automated process. Truly, the manual exploration must happen first. What do I think the drivers should be? If it’s predicting a purchase, I need to understand you as a customer. What magazines do you read? How old are you? Male/female? How often do you come to my store? All of these become data points in an algorithm that will lead to prescriptive data mining. If I can predict that you will make a purchase, I then need to be able to test my theory and provide results back to my model. Data Mining is Avery iterative process.
To have truly effective data mining, you need to have a multi-disciplinary team. Be sure to bring in your IT, Marketing, Finance, and operations-focused team members. They will all have different views of your customers and will understand their transactional data better than anyone else. Use all of these views to create a better view.
BIG Question – can you start data mining before you have your data organized in a DW?
I’d be interested in hearing your data mining stories.
- Jodie
The Science of Visual Analytics
September 1, 2010
Today LUCRUM hosted our second BI Symposium. Once again, it was well attended and we had some great speakers! I’m hoping that this becomes a regular event. If you have yet to attend, I encourage you to come to our next event (to be scheduled).
Our first speaker was Mr. Stuart Woodward, President OD OcuCue. (http://ocucue.com/) OcuCue is an interesting start-up that’s all focused on data visualization. I always love listening to visual experts. There is such a science to visual design. It’s about understanding the psychology of how users think and perceive what they see. If you are creating a dashboard, you have to design it in the way people think – we read from left to right, heavy color should be at the lower left hand side, etc.
“Good design has two key elements. Graphical elegance is often found in the simplicity of design and complexity of data.” – Edward Tufte
Mr. Woodward’s company creates meaningful dashboards that are icon based. They go beyond speedometers and graphs and actually create a customized dashboards with icons that are meaningful to the company using them. One example that he showed was for a hospital. There are some rooms that can only take female patients or only male patients. To show bed availability, their dashboard has a pink pillow or a blue pillow to represent which rooms are available. Hmmm…never thought of that!
How are you presenting data to your users? Are you simplifying the message? Setting up the information from left to right? Are your colors meaningful? (ie Red should mean bad, green is good)
OK…gotta run and listen to the next speaker!
- Jodie
