Read - Part I - Self-Service Analytics: What Could Possibly Go Wrong?

One reason self-service analytics is so hard is that it means different things to different people.

For example, an executive might think self-service analytics is the ability to view an online dashboard during an operational review meeting. A manager, on the other hand, views it as the ability to drill, sort, and filter dashboards and reports. And a marketing analyst thinks it’s about creating custom data sets from corporate and demographic data.

With self-service analytics, one size does not fit all. To successfully deploy self-service analytics, organizations must tailor the analytics experience to each and every individual in an organization.

Know Your Business Users

Identify. The first step in customizing analytics is to know your users. I’m always surprised when I consult with BI teams how few have identified all the business users in their organization who use data and reports. This is the bread and butter of BI. Unless you document your business users (and your technical architecture), you don’t have a strong foundation upon which to build a BI strategy. In essence, you’re leading blind.

Classify. The next step is to classify business users into relevant categories based on how they analyze and produce data. On the analysis side, do they simply want to view static reports or do they want to click and drill into details? On the production side, do they want to save live snapshots of reports for later viewing or do they want to assemble reports or dashboards from predefined charts, tables, metrics, and dimensions? (See figure 1.)

Figure 1. Self-Service Hierarchies

Understanding where users fall in functional hierarchies for analyzing and producing data is a key step in delivering a tailored self-service analytics experience.

Mapping Users to Tools

 I’ve defined two major categories of business users and four sub-categories. (See figure 2.) This scheme works well for most organizations, although more granularity is better. I use the grid below to help organizations map types of business users to categories of analytic and data tools that they require.

Figure 2. Tools Mapping

Casual users use data to do their jobs, while power users are paid to analyze data full time. Casual users demand “silver service”—that is data and reports tailored to their specific roles—or delivered on a “silver platter.” In contrast, power users demand true self service, the ability to create data sets and reports without IT involvement.

Casual User Tools. On the casual user side, data consumers are content to use reports and dashboards to make decisions, while data explorers want to go a step further and customize the reports or create new ones from a predefined business model called a semantic layer. Data explorers may also use data wrangling tools—lightweight data preparation tools—to merge report data with spreadsheets and governed data sets to create a tailored, ad hoc report for themselves or their departmental colleagues.

Power User Tools. On the power user side, data analysts—the proverbial spreadsheet jockeys in every department—mash up data from various systems, both governed and ungoverned, using data preparation tools and then analyze and visualize the results using visual discovery tools. Data scientists are glorified data analysts who know how to program using SQL, Java, Python and other languages. And the best are also statisticians who create analytical models using machine learning algorithms.

Tailored Fit. It’s a big mistake to expose tools geared to data explorers or data analysts to data consumers. They generally freak out when they see all the buttons and menu items and never give the tools a chance. They pick up the phone and ask their local analyst or the IT department to create a custom report for them. Similarly, giving data scientists tools geared to data explorers or even data analysts will undermine their productivity, if not prompt them to quit.

 Mapping Users to Data

It’s not enough to map business users to tools; you must also map them to an information supply chain. This is especially true for power users whom IT traditionally locks out of the data architecture for fear that they will bog down performance with runaway queries and create spreadmarts and data shadow systems. Figure 3 maps categories of business to a modern information supply chain.

Figure 3. Mapping Users to an Information Supply Chain

Data scientists generally want to access data in its raw, most granular form so they can format the data in an optimal way based on the analytic algorithms they need to apply. So giving them access to the Staging Area makes sense. In contrast, data analysts are partial to the Data Hub since it consolidates data by subject area (e.g., customer) into wide, flat tables, much like a spreadsheet.

On the casual users side, data explorers would rather access a sanitized, business view of data rather than database tables. Created by IT or a tech-savvy data analyst, the views make it easy for data explorers to drag and drop predefined report objects (e.g., metrics, dimensions, objects) onto a canvass to edit or create simple, ad hoc reports. Finally, data consumers just want to view or interact with reports and dashboards. At most, they only data they want to create is a live snapshot that they can view at a later time with fresh data.

Self-Service Supply Chain

Every type of user wants to augment corporate data in some way, shape, or form. Providing users the leeway to do that within a governed environment is the key to succeeding with self-service analytics. Without proper mappings and governance, all hell breaks loose with over- and user-use, as described in the previous article in this series. 

Thus, the tricky part of self-service analytics is stitching together the formal data flows in the IT- managed information supply chain with the informal, ad hoc data flows initiated by business users. Figure 4 below shows these dual supply chains and how they map to users and each other.

Figure 4. Mapping Dual Information Supply Chains

Where Do We Go From Here? 

Mapping users to tools and information supply chains is the job of a BI and data architect. This provides a straight and narrow pathway to success. However, the promise of self-service analytics cannot be achieved without adequate organizational structures, governance processes, and educational programs. That is the focus of the next article in this series.

 Also, stay tuned for an upcoming Eckerson Group report that addresses these issues in more detail. “A Reference Architecture for Self-Service Analytics: Balancing Agility and Governance in the Modern Era” and a companion infographic will be published in early September. 

Read - Part III - Self-Service Workflows: Curate, Create, Consume