This presentation was done by Gogula Aryalingam (Software Architect at Navantis) at the SLASSCOM Tech Talk - 'Smart Data Engineering' on 26th November 2014.Read more related content at Gogula's blog posts below:
http://dbantics.wordpress.com/2014/12/05/slaughter-analytics-part-1/
https://dbantics.wordpress.com/2014/12/21/slaughter-analytics-part-2/
5. BUSINESS INTELLIGENCE (BI)
Techniques
&
Tools
Raw data Meaningful
&
useful information
Transform
Analyze Make effective decisions
6. BUSINESS INTELLIGENCE (BI)
Heterogeneous
data sources
Extract,
Transform
&
Load
Data Warehouse/
Data Marts
OLAP/Cubes
Visualizations/
Reports
Dashboards/
Scorecards
7. TRADITIONAL BI
(also: Corporate BI, Enterprise BI)
Historical data (Various sources) → Data Warehouse
Periodically updated
— Weekly (Weekend-ly)
— Nightly
— Hourly (In recent times)
Provides
— Hindsight (via Reports, Dashboards, Scorecards)
— Sometimes Insight (via Data Mining)
8. TRADITIONAL BI
(also: Corporate BI, Enterprise BI)
Expensive (Hardware, Software)
Specialist/IT built
Used by:
— Top level management
— Some business users
— In some cases not used at all
9. PROBLEMS
Building a BI solution takes ages (sometimes 2-3
years)
— Things change
IT (or someone else) builds it for you
— You want more? You wait
— Fixed/Pre-decided reports
More than 70% of BI projects fail*
*Source: Gartner
10. WINDS OF CHANGE
A new breed of business user
— Technical/Non-Technical
Wants:
— To gain insight through data discovery
— To mashup data from various sources
(including public domain)
— To access data without going through IT
— The tools to do all this
12. CHARACTERISTICS
Users are self-reliant
— Allows access to data with minimum/no IT intervention
— Allows users to bring in their own sources
Allows for data discovery
Allows for sharing/collaboration
Agile
Is not a replacement for traditional BI
— Makes use of traditional BI
13. SCENARIO
Sales history and product info in cubes (traditional BI)
Month long ad campaign – data on new CRM system
You need to
— Analyze sales against ads – for effectiveness
— Analyze sentiments expressed on Twitter and Facebook about ads
This is a lightning presentation on Self-Service Business Intelligence, presented at the first SLASSCOM TechTalks event titled Smart Data Engineering held on the 26th of November, 2014 at the ICTAD auditorium in Colombo.
#SLASSCOMTechTalks
A quick show of hands of to see how many of the audience have just heard of BI
A show of hands of how many of the audience know what BI is
A show of hands of how many of the audience who have worked with BI – Implementing or as users
Business Intelligence, as I see, has various descriptions. Different people see it from different perspectives, depending on how it influences them and how it is used.
For me, business intelligence is about making decisions that are effective, and having the information to make them.
Hence, my definitions of BI, as depicted in slide is:
A set of techniques and tools to transform raw data into meaningful and useful information, which is analyzed in order to arrive at decisions that are effective.
Another great example is from the book Delivering Business Intelligence with SQL Server 2008 by Brian Larson:
“Business intelligence is the delivery of accurate, useful information to the appropriate decision makers within
the necessary timeframe to support effective decision making”
Typically, a business intelligence solution looked like what is depicted in this slide.
Raw data (i.e. data that is in its native structure, eg: transactional systems, spreadsheets etc.) are extracted from their source systems, brought together, transformed and loaded (a.k.a. ETL or Extract, Transform and Load) into a special type of database called data warehouses or data marts. These are essentially relational databases that are structured differently from a traditional OLTP structure. This structure which is usually designed using the dimensional modelling technique is optimal for reading.
A data mart is essentially similar to a data warehouse, except that it contains data from a single department or silo of an organization, whereas a data warehouse contains information from across the organization.
Data from the data warehouse/data mart is then pulled into a special type of database called OLAP databases, which have structures called cubes instead of tables. A cube has multiple dimensions, much unlike tables which have only two dimensions (rows and columns). A cube can have 2, 3, 4 or more dimensions; making it a very fast database for reading large amounts of data for reporting and analysis.
Finally, the visualizations. From enterprise reports that are 10 pages long that no one would read to nifty dashboards that show the state of the business at-a-glance can be created from the data that is collected.
Traditional BI (or the traditional way of doing BI, a.k.a. Corporate BI because business intelligence was usually used by corporates in a large scale) mostly stored data in a data warehouse. This data was periodically updated as new data came into the source systems. The updates were usually performed on a weekly, or nightly basis; and as time went by, and when technologies became better and cheaper, the updates were performed on an hourly and sometimes every minute.
These systems usually provided hindsight into the data and in certain cases some insight as well.
Still, traditional BI involved high costs for enterprise scale software and hardware. It was specialist-built or built by IT, and in a lot of cases did not exactly cater to what the business users wanted.
These BI systems were usually used by top-level management who mostly looked at the very high level picture of the business (dashboards) and some business users who did some analysis on the data (reports, interactive reports, scorecards).
And in most cases these systems were not used at all.
Most BI projects take a quite a long time to complete. Some take as much as 2-3 years, whereas a few others take almost 5 years… A lot are abandoned part way through. Most of the time it is because of the ambitious nature of trying to build the system for the entire organization, and due to no proper understanding between the technical folk building the system and the business folk who are the stakeholders, and during this (long) time things happen: people move out, new ones come in, requirements change, technologies get better (and others go obsolete) – visions are lost.
Changes take long to be incorporated etc. etc.
A new breed of business user has arrived. They are sometimes technical, sometimes not… But they have this thing for exploring. They have the knack for doing things their way.
These individuals want to gain insight into the business through data discovery. The information that they have from corporate BI systems is not enough, they want to bring other reliable data from within the organization and from the public space, mash them up and have more fine-grained insights. They do not want to wait for IT to serve the data to them – waiting a few days even, could be too late… and they need special tools for this.
Enter SELF-SERVICE BUSINESS INTELLIGENCE
Self-Service BI (SSBI) allows users to be self-reliant. I.e., the do not need the intervention of IT to get the data that they want, while they could bring in data from their own data sources.
It allows for data discovery and then sharing of that information for collaboration.
SSBI allows BI to be performed in an agile way, and is also ideal as a prototyping tool for larger business intelligence solutions.
One thing that has to be in mind is that SSBI is not a replacement for traditional BI, but is a complement and enhancer. Of course in certain cases all you need could be Self-Service BI itself!
Imagine a scenario where you have a traditional BI system. It contains sales history and product information in OLAP cubes. You then run a month long advertising campaign on TV, and record all related information in a new CRM system that you purchased.
You now need to analyze the effectiveness of the ads against the sales that was performed prior and subsequent to the campaign. You are also required to analyze the popularity of the ads against the sales by using social media feeds.
Going to IT to get you this information is going to be a joke, in a lot of cases… Rather, you could get the sales/product info from the cubes, pull in the ad data yourself from the CRM system, get IT to quickly set up a Hadoop cluster on the cloud and some code to pull in social media feeds into it, and then use self-service BI tools to pull in, mash up these data and get insights yourself!
SELF-SERVICE BUSINESS INTELLIGENCE!
Some of the popular tools for SSBI are Tableau and QlikView. However, since I have Excel 2013, I could do everything using that with the help of free Power BI add-ons, and then publish the findings on a subscription based Power BI portal on the cloud.
The example shown here is not much of a business scenario, but it does show how data can be pulled off a website and analyzed for insight.
Data is taken off official livestock slaughter statistics from Sri Lanka.
Apart from what you saw in the video, this is analysis done for the district of Hambantota, which shows the slaughter of cattle declining over the years.
An analysis on the Polannaruwa district shows a similar story, but a little bit of inconsistency here and there...
However, on the whole, these two districts are almost the same in number of slaughters and how the numbers have dwindled in the last few years…
Just goes to show that they are both more or less the same
The slaughter data that you find in the video and these slides come off an official government site, hence this is not the work of any type of computer jilmaart.
(http://www.statistics.gov.lk/agriculture/Livestock/slaughterstatistcs.html)
The aftermath of the demo… What can be done further…
If you would like to learn more about self-service BI, reach out on gogulaa@gmail.com or @gogula on twitter.