August 4, 2014

Performance Analytics Architecture

I tell clients and students that technology can do whatever you want…if you know what you want.  While knowing what you want – the questions you want answered – is the first and most important step, preparing the information to answer those questions is 90% of the work.  The good news is that you don’t have to “boil the ocean” to get started.  Like building anything of quality, the architect must know what it’s going to be when it’s finished, so everything developed and delivered builds on what has already been developed to further leverage the Analytics investment.
It’s really logical when the process is considered:
  1. Decisive action is enabled by questions answered – the right information at the right time.
  2. Questions are answered by business people using Analytics tools
  3. Analytic tools require organized information that is current, consistent and accurate.
  4. Organized information is develop by integrating data from inside and outside the organization.

Let’s expand on this simple process with a little reality.  In most organizations, business users don’t drive Analytics initiatives, so reporting projects become long and expensive because the technology team may not have a good understanding of the business questions that need to be answered, so they work to answer every possible question, i.e., “boil the ocean.”  In the process:
  1. The business user communities don’t have control or much participation in the process, so they don’t feel ownership of the Analytics deliverables.  “It’s another IT initiative.”
  2. Report requests must be submitted to the IT group, where a report development backlog grows, making business users frustrated with IT, so they continue using spreadsheet reports with inconsistent data definitions and calculations.
  3. Information (i.e., reports and analytic deliverables) are “pushed” to business users who don’t understand the information or how to access the data to develop answers to their real questions.
  4. Management receives untimely (i.e., monthly) historical reports with marginal information, allowing opportunities to pass and problems to grow.
  5. Calculations and terms are not defined consistently across all business units causing confusion and arguments in management meetings about which report is right.
  6. Incomplete information from a single application data source provides possibly misleading customer and product segmentation and behavior profiles, which leads to inaccurate analyses and bad decisions.
  7. …and the list goes on.

Providing business users with immediate access to relevant, accurate, consistent business information is the requirement.  While Analytics Initiatives are guided by a continuous stream of increasingly insightful questions from across the organization that build on answers to previous questions, the foundation performance analytics architecture must be planned and in place to maximize an organization’s return on intellectual assets.  The architecture includes:
1.      Enterprise Data Warehouse Architecture
In a perfect world, this is an enterprise data warehouse (EDW) architecture with all of the key components.  In reality, there are a lot of “variations on a theme.”  It doesn’t happen at once, but having the vision of the model will lead to making fewer mistakes.
Automating and iteratively tuning the model and processes as analytic requirements evolve provides a high return on investment.
2.      Extract, Transfer and Load (ETL) Processes
Extracting the application system data required to answer questions, and then transferring it to a staging database, and then converting/standardizing the various data types and definitions as it is consolidated and loaded into the EDW and then on to the functional data marts is a complex, but necessary process to establish a single source for enterprise analytics.  An understanding of the desired data model and data types makes the process a lot easier. Automating the process makes it fast, invisible and cost effective.  This is a good thing.
3.      Data Governance
The most challenging part of the process is getting lines of business and strategic business units to 1) share their data and 2) agree on a common dictionary of terms and calculations. This is why all successful Analytics Programs are sponsored by senior executives.
4.      Data Marts
Functional data marts provide the control, security and perspective to maximize business user information access and productivity.  Access to everything, when security isn’t an issued, will be overwhelming to business users, who need to focus on their business objectives, rather than everything that is going on in the world.  Defining the data requirements and automating the ETL update processes makes data marts a very productive platform.
5.      Security
Security profiles make business user self-service possible.  When individual and team performance scorecards are employed, secure access to performance dashboards that are updated daily is an incredibly productive consideration.  Do it!
6.      Analytics Tools
“How do I love thee?  Let me count the ways!”  Sorry for waxing poetic a little, but I’ve been lovin’ reporting and analytic tools for many years…and I still do.  They are the answer to my questions…and yours.  I used Pilot Software, IBI Focus and IRI Express when I started in 1990.  I’ve been a Cognos Partner since 1995, a Microsoft Partner since 2001, recently designed and developed a Business Objects-based enterprise reporting platform for a $13B global technology client, and taught Visual Analytics at a large New England University, featuring Tableau and QlikView.
The good news is that they have become a commodity – they can all, basically, do the same thing.  There are new ‘data discovery’ tools, like QlikView, Tableau and others, who have very good visual capabilities, but it all goes back to: “What do we want to know?”
It’s interesting that the top 3 vendors in Gartner’s 2014 Business Intelligence and Analytics Magic Quadrant are Microsoft, Tableau and QlikView, the latter two being relative newcomers to the game.  But they’re all good, depending on the questions that need to be answered.  The really interesting thing is that the Microsoft BI platform is not being marketed as a business user platform…when it’s an excellent and very cost-effective reporting and analytics capability.  More on this another time…
7.      Analytics Library
Redundant report development is more of a problem than using spreadsheets.  While trying to answer the same question, but using variations in data and presentation graphics, most organizations have many reports that answer the same question.  An Analytics Library helps organizations improve productivity and save time and expense by not recreating the wheel (i.e., developing a new report to answer the same question).
8.      Business User Self Service
This is where the rubber meets the road.  Reporting and analytics is not an IT initiative – it’s a business initiative that empowers everyone in the organization to make the best possible decisions regarding customer, product and employee processes…which result in optimized quality, productivity and expense management outcomes.  Providing business users to access information that provides them with a context to understand how each of their jobs contributes to the success of the organization is HUGE!
This is a journey with a changing destination.  As dynamic business requirements change, so do Analytic requirements.  Using an architecture that anticipates and accommodates these changes will save a lot of time and expense, while increasing Organization IQ and shareholder value.
Think about it. 

What questions do you have?