The Esoteric World of Predictive Analytics

Let me start with the defintion of Predictive Analytics as used in literature – “The nontrivial extraction of implicit, previously unknown and potentially useful information from data”. If that doesn’t sound esoteric enough, you are probably more advanced than what this post gives you credit for!

For a BI practitioner, it is important to get an understanding of Predictive Analytics (also known as Data Mining) as this subject definitely deserves a place in the wide spectrum of Business Intelligence disciplines. BI at a broad level is about optimizing business through “Hindsight, Insight and Foresight”. Predictive analytics adds the powerful “Foresight” part to business decision making.

Most BI practitioners tend to equate statistics with predictive analytics and this post explains why such a view is inaccurate. To understand this let’s start at the very beginning (a la Alice in Wonderland). Broadly, this world is divided into 2 types of systems:

• Physical Systems – Has causality and hence can be modeled mathematically with relative ease
• Human Behavioral Systems – Lacks causality and can be modeled only with specialized techniques

Predictive analytics for business decision making is all about modeling human behavioral systems.

Why Traditional Statistics is insufficient?

Though the entry into predictive analytics requires that we understand the implications of traditional statistical analysis, statistics by itself is insufficient in the business context. Traditional statistical analysis allows us to understand the general group behavior and is primarily concerned with common behavior within the group – the central tendencies.

In business we generally develop models to anticipate human behavior of some type. Human behavior is inconsistent, lacks causality and distributions based on human behavior almost always violate the assumptions of traditional statistical analysis (like normal distribution of data, stability of mean and standard deviation etc). The strength of data mining comes from the ability of the associated techniques to deal with the tails of the distributions, rather than the central tendencies, and from the techniques’ ability to deal with the realities of the data in a more precise manner.

In the realm of predictive analytics, we are concerned with modeling human behavior and hence are interested with the tail of our distribution – small percentage of the population that responds to a campaign, commits a fraud, leave our business or purchase the next service.

Though there are specialized techniques used for Predictive Analytics (viz. Non-linear statistics, Induction Algorithms, Cluster Analysis, Neural Networks to name a few), a BI practitioner is only expected to appreciate its usage in different business situations, prepare and model data as required by the tools and interpret the results correctly (a much less daunting task indeed!)

Typically the model development process involves the following steps – a) Define Project, b) Select Data, c) Prepare Data, d) Transform Variables, e) Process Model, f) Validate Model, g) Implement Model. I will explain these steps in more detail in subsequent posts.

Fundamentally, an end-to-end BI view requires the practitioner to learn the concepts around statistics and predictive analytical techniques as available in tools (like say SQL Server Analysis Services) in addition to their technology bag of tricks around data integration, data modeling and OLAP.

Read More About  Predictive Analytics

Business Objects Security

In the current business scenario, securing the data and restricting the users from what rows and columns of data they can see and what rows and columns of data they cannot see is very important.  We can secure the rows of data by row level security. Some people call this as ‘Fine grained access control’.  We can secure the columns of data by column level security. This is popularly called in Business Objects as ‘Object level security’

ROW LEVEL SECURITY

There are various ways through which the row level security can be implemented in a Business Objects environment.

One way is by securing the datamart. In case of this approach, the datamart is secured – meaning the security policies and rules are written in the datamart. Technically, a security table can be created and maintained having the users / groups with corresponding access rights.  Security policies can have a logic to compare the active logged in user and security table. All the users accessing the datamart are provided access to their data only after executing the security policies. We can also embed the security policies and rules in a view. A good example for row level security is — Non-Managers cannot see the data of   co-workers however managers can see the data of his / her sub-ordinates. In Oracle (for example), we can create a non-manager and manager views with the security rule (<security_table.user> = “USER”). The security views are imported in the Business Objects ( BO) universe and the reports use these security views through the universe. The main ADVANTAGE of securing your datamart is that your security rules can also be used by many other BI tools ( Cognos, Microstrategy )  as the rules are built at the datamart and NOT at the Business Objects)

Second way is by building the security rules at the Business Objects. Here the security rules comparing the logged in user and security data can be written in a virtual table of your Business Objects. These virtual tables are nothing but the universe derived table. BO Reports use the derived table to access the datamart tables. Alternatively, we can also define security filters in a BO universe. The filters are called as condition / filter  objects in the BO universe world. With this approach, you can take the maximum ADVANTAGE of the BO features however the disadvantage is that when you are going to a different BI tool like Cognos you need to rewrite the business security rules in your new tool.

In case of the projects dealing with the migration of Peoplesoft transactional reporting to Business Objects analytical reporting. We can potentially reuse / import some security tables  and security policies from Peoplesoft into our analytical datamart. These reusable components can save time in building the secured datamart and reporting environment.

COLUMN LEVEL SECURITY

Like ‘Row level security’, we can implement the column level security either at the datamart or Business Objects. In the financial industry, the business users do not want their revenue amounts, social security number , tax id number and other sensitive columns to be shown to unauthorized users.  Given this instance, we can mask the sensitive columns by a restricted tag in the place of sensitive columns. Non-sensitive columns like first name , last name , gender , age can be left and shown as it is to the end business user. These logic can be technically implemented in the business objects universe derived table or datamart views using a decode / ‘if then else’ / case statements.

Alternatively , we can use the universe object restriction feature in the BO designer to define restriction on the universe objects. So whenever a business user tries to drag the restricted object from the universe , the restriction rules get invoked , authorization occurs and the object access is given to the end user if he / she is successfully authenticated to access that object.

I’m signing off this BO security blog for now. The contents are based on my knowledge and BO experience in various projects.  Thanks for reading.  Please share your thoughts on this blog. Also, please let me know your project experiences pertaining to row and column level security in Business Objects.

Read More About  Business Objects Security

Business Intelligence Challenge – Product Upgrades & Migrations, Moving the Code – 4

Last time we discussed about Impact Assessment , the next logical step after this is to perform the actual upgrade or migration of the code.

Moving the Code: Performing Upgrade or Migration of the Objects

When we talk about product upgrades, always the product vendor provides tools by which the objects in the earlier version can be upgraded to the latest version. Yes we would see some objects failing through while using such tools; these are the ones that would need rework after the upgrade process.

When we talk about product migration like moving from Cognos to Business Objects or Business Objects to Cognos, there is good scope for us to look for some ways to automate the code migration. Earlier discussions have been on how to leverage the metadata for understanding the environment, now we are looking at an option on how to manipulate or transform the metadata so that an object in platform ‘A’ becomes compliant to platform ‘B’.

Steps involved in building an automated product migration process

Perform metadata level object mapping between the two platforms, determine the gaps. This would actually be a ‘by product’ of ‘Step 2’ in Impact Assessment

Build individual components that would

• Read the metadata from the source platform and prepare a repository
• Have the knowledge of the match & gap between the platforms, could be reference tables
• Transform the ‘source’ metadata and write out as understood by the ‘target’ platform by using the reference tables

Benefits of Automated Migration

• Helps avoid creation of objects from scratch
• Ensures availability of time for testing (core task) than code development
• Enables team to have a flexible skillset
• A faster way of delivering things when a ‘one to one’ migration from the source platform is seen as a must

Automated Migration Challenges

Transforming the source metadata to the target platform would be a challenge, especially with data manipulation functions. Having a good understanding of the gaps will help; a reference table mapping the functions between the platforms would be useful. In scenarios where a function cannot be converted to the target platform, a comment can be written into a log file enabling quicker attention.

Have seen good success in writing such automated migration components though not 100%. With almost every products providing good SDK kits for reading and as well writing metadata and as well with the support for XML structures, writing such bridges for object migration are getting easier.

Whether the objects in a product are migrated/upgraded in an automated way or not, the following activity of ‘Validation’ plays a key role in ensuring the final quality, next time let us discuss on some of the means for effective validation ….

Read More About  Business Intelligence Challenge

Zachman Framework for BI Assessments

The Zachman Framework for Enterprise Architecture has become the model around which major organizations view and communicate their enterprise information infrastructure. Enterprise Architecture provides the blueprint, or architecture, for the organization’s information infrastructure. More information on the Zachman Framework can be obtained at www.zifa.com.

For BI practitioners, the Zachman Framework provides a way of articulating the current state of the BI infrastructure in the organization. Ralph Kimball in his eminently readable book “The Data Warehouse Lifecycle Toolkit” illustrates how the Zachman Framework can be adapted to the Business Intelligence context.

Given below is a version of the Zachman Framework that I have used in some of my consulting engagements. This is just one way of using this framework but does illustrate the power of this model in some measure.

Some Salient Points with respect to the above diagram are:

• The framework answers the basic questions of “What”, “How”, “Who” and “Where” across 4 important dimensions – Business Requirements, Conceptual Model, Logical/Physical Model and Actual Implementation.
• Zachman Framework reinforces the fact that a successful enterprise system combines the ingredients of business, process, people and technology in proper measure.
• It is typically used to assess the current state of the BI infrastructure in any organization
• Each of the cells that lies at the intersection of the rows and columns (Ex: Information Requirements of Business) has to be documented in detail as part of the assessment document
• Information on each cell is gathered through subjective and objective questionnaires.
• Scoring Models can be developed to provide an assessment score for each of the cells. Based on the scores, a set of recommendations can be provided to achieve the intended goals.
• Another interesting thought is to create a As-Is Zachman framework and overlay that with To-Be one in situations where re-engineering of a BI environment is undertaken. This will help us provide a transition path from the current state to the future.

Thanks for reading. If you have used the Zachman framework differently in your environment, please do share your thoughts.

Valuing your Business Intelligence System – Part 1

Sample these statements:

• Dow Jones Industrial Average jumped 200 points today, a 2% increase from the previous close
• The carbon footprint of an average individual in the world is about 4 tonnes per year which is a 3% increase over last year
• The number of unique URL’s as on July 2008 in the World Wide web is 1 trillion. The previous landmark of 1 billion was reached in 2000
• One day 5% VaR (Value at Risk) for the portfolio is $ 1 Million as compared to the VaR of $ 1.3 Million a couple of weeks back

Most of us buy into the idea of having a single number that encapsulates complex phenomena. Though the details of the underlying processes are important, the single number (and the trend) does act like a bellwether of sorts helping us quickly get a feel of the current situation.

As a BI practitioner, I feel that it is about time that we formulated a way for valuing the BI infrastructure in organizations. Imagine a scenario where the Director of BI in company X can announce thus: “The value of the BI system in this organization has grown 15% over the past 1 year to touch $50 Million” (substitute your appropriate currencies here!).

The core idea of this post is to find a way to “scientifically put a number to your data warehouse”. Here are a few level setting points:

1. Valuation of BI systems is different from computing the Return on Investment (ROI) for BI initiatives. ROI calculations are typically done using Discounted Cash Flow techniques and are used in organizations to some extent
2. More than the absolute number, the trends are important which means that the BI system has to be valued using the same norms at different points in time. Scientific / Mathematical rigor helps in bringing the consistency aspect.
3.  

My perspective to valuation is based on the “Outside-in” logic where the fundamental premise is that the value of the BI infrastructure is completely determined by its consumption. Or in other words, if there are no consumers for your data warehouse, the value of such a system is zero. One simple, yet powerful technique in the “Outside-in” category is RFM Analysis. RFM stands for Recency, Frequency and Monetary and is very popular in the direct marketing world. My 2-step hypothesis for BI system valuation using the RFM technique is:

• Step 1: Value of BI system = Sum of the values of individual BI consumers
• Step 2: Value of each individual consumer = Function (Recency, Frequency, Monetary parameters)

Qualitatively speaking, from the business user standpoint, one who has accessed information from the BI system more recently, has been using data more frequently and uses that information to make decisions that are critical to the organization will be given a higher value. A calibration chart will provide the specific value associated with RFM parameters based on the categories within them. For example: For the Recency parameter, usage of information within the last 1 day can be fixed at 10 points while access 10 days back will fetch 1 point. I will explain my version of the calibration chart in detail in subsequent posts. (Please note that the conversion of points to dollar values is also an interesting, non-trivial exercise)

Am sure that people acknowledge the fact that valuing data assets are difficult, tricky at best. But then, lot more difficult questions on nature and behavior have been reduced to mathematical equations – probably, the day on which BI practitioners can apply standardized techniques to value their BI infrastructure is not too far off.

Read More About  Business Intelligence System