Being scientific about the business requires more than treating data scientifically.

Extracting value from big-data analytics is about changing the way in which business is done.¹ Intuition and tradition must give way to evidence and rigor. Kotter (Leading Change²) may actually tell us at least as much about how to achieve this as Bartlett (Practitioner's Guide to Business Analytics³).

Some analysts assert that "Big data business benefits are hampered by 'culture clash'."⁴ Allegedly, traditional "requirements-driven" approaches to managing enterprise IT systems impede "opportunistic analytics and exploring answers to ill-formed or nonexistent questions." Traditional IT bureaucracy purportedly suffocates entrepreneurial creativity of a "new breed" of big-data specialists.

I previously wrote that extracting value from big-data analytics requires us to simultaneously apply scientific methods to both our data and the business questions we seek to answer. I observed that data-science methods — by which big-data analytics yield information bearing economic value — can be classified either as data-centric or business-centric. The business-centric methods emphasize scientific methods for business questions. Data-centric methods narrowly focus on the data.

Guiding an organization's ubiquitous adoption of data-driven practices leads us beyond the scientific method for business questions. We must be systematic and methodical about the business itself. We consider the entire span of businesses attributes over which we can reasonably assert control. The following discussion summarizes a pair of frameworks by which we guide the application of opportunistic analytics by entrepreneurial big-data specialists.

I first describe a framework originally formulated to guide the adoption of Service-Oriented Architecture (SOA).⁵ We decompose the business into separable components. I illustrate using Michael Porter's Value-Chain Framework⁶ as a starting point. A second framework — characterizing business functions' operational criticality — focuses our attention on the risk of insufficiently-managed change. 

These two methods establish boundaries and conditions under which experiments in opportunistic analytics may be applied to core-business functions. They give decision makers control points by which to manage the pursuit of big-data opportunities. 

Business-capabilities analysis — identifying targets for big-data experimentation.

The scientific method makes extensive use of models. Developing a model for the organization often represents the first step in identifying and prioritizing opportunities for big-data analytics. Academia and the consulting industry have produced a variety of useful approaches to the modeling of organizations. Examples include:

• Enterprise architecture methods (e.g., The Open Group Architecture Framework (TOGAF)⁸) characterizes organizations in technology-centric terms;
• Business-Process Modeling takes a functional view of organizations;
• Lean/Six-Sigma emphasizes quality and consistency of business outputs; and
• Quantitative business models characterize financial and economic attributes of business functions.¹⁰

Figure 1 illustrates a functional model for a manufacturing business. It decomposes a value-chain⁶ view into increasingly more-granular functions. This approach was developed to identify and prioritize opportunities for technology investments leading to improved business performance. Consulting organizations, exemplified by IBM,⁹ have more formally developed similar methods.

Figure 1  — A Business-Capabilities Analysis (BCA) prioritizes and establishes the scope for opportunistic experimentation with big-data analytics. (From Merrifield.⁷) 

The business' functional breakdown in Figure 1 is annotated with two pieces of information. A label at the top of each box indicates the importance of the function to the business. This rating might be assigned based on contribution to competitive advantage, essentiality to business operations, or other criteria.

The boxes themselves are color-coded. The coloring indicates how well the function is performing. This performance characterization could compare the individual business against industry benchmarks. It might also indicate disproportionate consumption of management attention or other resources.

Priorities for experimentation with big-data analytics are assigned based on these two criteria. Big-data analytics resources — of which scarce, high-priced talent is often the principal component — are apportioned judiciously. High-value business functions requiring attention are usually assigned the highest priorities.

Characterizing the risk of change associated with new applications of big-data analytics.

Change to an organization is often difficult — even, frequently, turbulent. Introducing new big-data analytics business functions requires organizations' members to change how they think about the business. The maxim "If it ain't broke, don't fix it" usually represents good advice. 

Figure 2 contains a framework for thinking about how business functions contribute to an organization's mission. This framework was originally developed to characterize IT's contribution to a business en total. It is equally valid for thinking about business analytics. 

Figure 2 — A strategic-impact grid gives us a rigorous framework for characterizing the risk of experimentation and change associated with introducing new big-data analytics to a business function. (From Nolan.¹¹)

Information quality provides the big-data-analytics equivalent to the IT-reliability factor. Factory- and strategic-mode business functions demand high-confidence, relevant data. Figure 2 emphasizes the availability of IT services in general. Operational capabilities using outputs from big-data analytics must also provide high-confidence, actionable information.

I provided another 2×2 grid for characterizing data quality in Figure 2 of my previous installment on treating business questions scientifically. Factory- and strategic-mode functions demand outputs from big-data analytics that fall in the "Differentiating Information" quadrant. Even of the service based on big-data analytics is continuously available, lapses in information quality may lead to catastrophically flawed decisions.

Characterizations based on Figure 2 in this installment might also inform characterization of business functions the framework of Figure 1. The vertical axis in Figure 2 might inform our business-value assignment. The need for new business analytics — the horizontal axis of Figure 2 — might inform the performance-spectrum assignment.

Experimentation with big-data analytics as the launchpad for business innovation.

We began this discussion with the observation about an alleged conflict⁴ between traditional enterprise-information management — data warehouses and business intelligence — and and the "new breed" of opportunistic big-data specialists. In reality, both are required. Innovation based on big-data analytics requires a disciplined, managed approach to change. 

Changes to factory- and strategic-mode business functions from Figure 2 must be made cautiously. The business depends upon their reliability. Thomke and Manzi¹² describe a disciplined approach to business experimentation. Their examples extend beyond opportunistic big-data analytics. Nonetheless their approach also applies here. 

We perform big-data analytics experiments through controlled, non-disruptive pilots. Experimenters are free to explore new opportunities using business-representative data. Their activities are isolated from core business operations, particularly those that are factory- or strategic-mode in nature.

But what about scenarios in which big-data analytics identify opportunities for disruptive innovation? 

This question addresses scenarios leading to extreme change. The innovation opportunity may require a business to begin cannibalizing existing revenue streams before competitors exploit the opportunity first.¹³ 

Two principles guide the decision to embark on such a radical course. First, we previously observed that the information from big-data analytics assumes value commensurate with the economic returns from its exploitation. This is a foundational tenet of Business Analytics. 

Capturing this value may require re-orchestration of activities by the entire organization. Operationalizing new big-data analytics capabilities may lead to reconfiguring the business model exemplified in Figure 1. "New-breed" big-data specialists cannot do this by themselves. Taking action to capture value resides beyond the scope of most analytics specialists' activities.

Second, big-data specialists do not by themselves make the decision to embark on major organizational change. This is particularly the case for decisions about whether to divest or unwind revenue streams that are highly vulnerable to market disruption. These decisions are made by the organization's leadership.  

"New-breed" big-data specialists aid in identifying these scenarios. They serve as trusted advisors to the organization's leaders about how best to exploit the opportunities. They also assist in working through practical details of operationalizing the opportunities they discover. The results of their work — particularly when applied to factory- and strategic-mode business functions — ultimately however become subject to the disciplined governance of "traditional" data-warehouse and business-intelligence functions.

References

¹ S. LaValle, et al, "Big Data, analytics, and the path from insights to value," MITSloan Management Review, Winter 2011, pp. 21 - 31, http://goo.gl/MY5VK6. 

² J. P. Kotter, Leading Change, Boston: Harvard Business Review Press, 1996, http://goo.gl/EqglOJ. 

³ R. Bartlett, A practitioner's guide to business analytics, New York: McGraw-Hill, 2013, http://goo.gl/o6dTOS.

⁴ F. Buytendijk and D. Laney, "Big data business benefits are hampered by 'culture clash'," The Gartner Group, Report G00252895, September 12, 2013, http://goo.gl/8GLRgk.

⁵ "New to SOA and web services," IBM DeveloperWorks, https://www.ibm.com/developerworks/webservices/newto/.

⁶ M. Porter, Competitive Advantage of Nations, New York: The Free Press, 1990, Figure 2-3, p. 41, http://goo.gl/Qn1xGn.

⁷ R. Merrifield, et al, "The next productivity revolution," Harvard Business Review, June 2008, https://hbr.org/2008/06/the-next-revolution-in-productivity.

⁸ "The Open Group Architecture Framework," version 9.1, The Open Group, http://www.opengroup.org/togaf/. 

⁹ "Component business models: Making specialization real," IBM Institute of Business Value, http://goo.gl/QhucjH.

¹⁰ J. Tennant and G. Friend, Guide to Business Modelling, third edition, London: The Economist Newspaper, 2011, http://goo.gl/b36blO.

¹¹ R. Nolan and F. W. McFarlan, "Information technology and the board of directors," Harvard Business Review, October 2005, http://goo.gl/AqCdLK.

¹² S. Thomke and J. Manzi, "The discipline of business experimentation," Harvard Business Review, December 2014, pp. 70 - 79, https://hbr.org/2014/12/the-discipline-of-business-experimentation.

¹³ C. M. Christensen, The Innovator's Dilemma, Boston: Harvard Business School Press, 1997, http://goo.gl/kg2u9Y.

© The Quant's Prism, 2014