Analytics and Data Warehousing: The Power of Simplification

Business Intelligence
  • Smaller Small Medium Big Bigger
  • Default Helvetica Segoe Georgia Times

In a time when technology is making it increasingly easy to use sophisticated analytics in business, the best approach for most small to mid-size companies is to keep things simple.


The computer industry has come a long way in terms of enabling data storage, and the trade press is filled with stories of big data. But ultimately, the biggest challenge for any company isn't capturing and storing more data; it's finding easier, affordable ways to access, analyze, and share data that business users will embrace.


The analytics software marketplace includes many third-party firms with off-the-shelf models for common business functionsfrom marketing and sales to customer loyalty, inventory, logistics, and others. Often these are hosted, Web-based services offering key performance measures built over a packaged data model. Companies subscribing to the service are typically given instructions for what data elements to extract and transmit to the service provider who then loads the data into the packaged model and generates the standard outputs. The process isn't difficult and the cost may be low, but meaningful subtleties about a business can be lost in this transfer and translation process. Getting business analysts and executives to trust and use the service can also be a challenge.


Some ERP software vendors have taken steps to embed data marts and even template data warehouse models in their solutions. These models usually come with standard reports, dashboards, and other outputs intended to minimize implementation time and cost. Unfortunately, there is often a substantial lack of common ground between those who:


  • design industry-, application-, or function-specific data models
  • understand your business processes
  • input data into your application software database
  • write programs that load digitally gathered data into your online transaction processing (OLTP) database
  • expect to leverage the data to identify opportunities and trends
  • authorize information technology investments


These factors make it difficult to design a standard model that can satisfy the analytics and reporting needs of multiple companies. Performance indicators and business metrics are subtly personal and different from company to company. The data elements one manager finds meaningful may be extraneous detail to another manager, even in the same company. Some managers don't trust summaries and need to see all the detail. Others simply want totals and charts.


Being able to access data and knowing how to use it to improve business tactics or convince people to change how they perform their jobs is not the same thing. While big data and analytics dominate technology industry headlines and advertising campaigns, in many mid-size companies, printed reports, on-screen inquiries built into business applications, and manually maintained spreadsheets remain the primary methods of monitoring and analyzing business performance. Each of these "old-school" methods was cited as being used more than twice as often as dashboards and analytics or data mining and five times more often than online analytical processing (OLAP) in a recently completed survey of MC Press Online, LLC readers conducted by New Generation Software, Inc. (NGS). The survey wasn't scientifically conducted, but the consistency of the responses across the range of companies and industries strongly suggests a significant gap between business practices and today's technological possibilities.


Anecdotal experience points to many reasons for this gap, including cost, complexity, a shortage of trained staff, and frequently an outdated awareness of what's possible with existing tools. As a result, when evaluating any packaged data model, it's wise to ask:


  • What will the business users see?
  • Is the data available in summary, graphical, and detailed, tabular form?
  • How easy is the software to use?
  • How current will the data be?
  • Can the data model be modified, expanded? If yes, what's the process? Who does the work and at what cost?
  • Can data be exported for offline analysis and ad hoc exploration using query tools and commonly available software like Microsoft Office?
  • If hosted rather than stored on premises, can the data model be transferred to another service provider or brought in-house at a future date? If yes, at what cost?


If you're an IT professional who has been assigned to a business analytics project team, understand from the outset that business people have a low tolerance for software that requires them to learn a new way of thinking or working. Give them a way to interact with business data the way they visualize it, which is most likely in the form of a spreadsheet or single table, and then there's a greater chance of long-term success. The greatest value will come to the surface when skilled business people are given versatile, self-service tools that shield them from database complexity, but enable them to investigate the data in their own personal way. Ease of use, not advanced features or the size of the investment, is the primary barometer of how much any analytics software will appeal to these users.


Business executives and analysts usually have a pretty good sense of the pressure points they need to mine in pursuit of critical insights. But business people without relational database backgrounds tend to have a "flat file" view of data. Give them a single, de-normalized table holding the data elements they want and they will probably know what to do next. Expecting them to learn SQL, the structure of a multi-table data model, and how to perform extraction, transformation, and loading (ETL) functions is the fast path to disappointment. Information technology professionals who understand your application software, database, and business are the ideal people to perform those data preparation and aggregation tasks. These database savvy users want robust function above all else. Interestingly, in the MC Press Online, LLC reader survey, respondents said that, when working with IBM i software applications, they prefer a 5250 user interface or an application that offers both a 5250 and graphical user interface 67% of the time.


Until just a few years ago, the limited scalability of spreadsheets like Microsoft Excel made their use impractical for more than summary analysis in all but the smallest of companies. Some of the biggest names in business intelligence software grew rapidly by filling that market niche, but often by implementing solutions that minimized IBM i utilization and required moving data to another server and database. In the MC Press Online, LLC reader survey, 38% of the respondents said their companies maintain a data warehouse on another platform instead of IBM i. This data management infrastructure usually leads to a need for additional staff between the business application development and support people with the best understanding of how the data originated and the analysts and executives who are expected to study it. Any database architecture or organizational structure that puts distance between the people with the deepest knowledge of your data and the people who need to explore it is very likely to have a disappointing return on investment.


Given their significant investments, we might assume companies with data warehouses on another platform are bigger users of dashboards, analytics/data mining, or OLAP software. However, the MC Press Online, LLC reader survey responses did not support that assumption. The respondents in the companies using IBM i directly and those with data warehouses on another platform reported very close to the same rate of dashboard use. With regard to data mining/analytics, 59% of those reporting they have a data warehouse on another platform did not indicate their companies use these techniques. OLAP usage followed the same trend with the companies who maintain a non-IBM i data warehouse indicating they used OLAP at very nearly the same rate as the companies who didn't move their data from IBM i. Further research should be conducted to investigate these responses, but the numbers again suggest that just because the data is being captured and delivered to a database intended to support analytics, that doesn't mean it's more likely to be used. The possible points of disconnection mentioned previously may be partially responsible.


Given the apparent gap between the availability and effective use of business data, what should a small to mid-size company with limited staffing do? A good starting place is to get your team together to define your analytics priorities and look for opportunities to simplify your data management processes.


Microsoft SQL Server frequently serves as the data warehouse database role in IBM i shops. But today, with the Excel Power Pivot columnar database (available as a no-fee add-on to Office 2010 and as a standard feature of Office 2013 and Office 365), IBM i customers hoping to unlock the analytical potential in their data by supporting a SQL Server–based data warehouse or other relational database should revisit their strategy to see if the non–DB2 on i database is still necessary. This need for a strategy review is especially true for companies with excess capacity on their IBM Power System. DB2 on i is a very solid, accessible, relational database, and Power Pivot can efficiently store and aggregate multiple millions of rows of character data. From there, Excel 2013 and Office 365 users can experiment with Microsoft's Power View and Power Map visualization and mapping features alongside their familiar spreadsheet functions. The cost and the end-user learning curve are very modest.


Encouraging communication across departments and skill levels is critical and too often rare. Can you gain agreement on what aspects of your business your management wants to explore and change? Do you have or can you obtain the data you need to support this goal? Who on the team has the best understanding of the applications and programs that capture the data related to this area? Are you underutilizing existing software applications that could help you achieve your goals? Are there inexpensive add-ons to these tools that could help people be more productive? What mobile operating systems and Web browsers does your solution need to support? Can you get some training funds added to your budget? Who among the analysts or executives has the best grasp of where the effort should lead? Would it help to have your analysts take a refresher business statistics class? Most of us haven't thought much about statistical analysis methods in many years and won't apply what we can't understand.


Surprisingly, in a time when technology is making it increasingly easy to use sophisticated analytics in business, the best approach for most small to mid-size companies is to keep things simple. Leveraging IBM DB2 on i with robust back-end ETL and query solutions that can deliver tables to users of commonly available tools is your low-risk, potentially high-reward strategy.