For more than 50 years, Auerbach Publications has been printing cutting-edge books on all topics IT.

Read archived articles or become a new subscriber to IT Today, a free newsletter.

This free newsetter offers strategies and insight to managers and hackers alike. Become a new subscriber today.


Partners




Contact

Interested in submitting an article? Want to comment about an article?

Contact John Wyzalek editor of IT Performance Improvement.

 

Big Data and Business Analytics

Joe LaCugna, PhD
Enterprise Analytics and Business Intelligence
Starbucks Coffee Company

The promise and potential of big data and smart analysis are realized in better decisions and stronger business results. But good ideas rarely implement themselves, and often the heavy hand of history means that bad practices and outdated processes tend to persist. Even in organizations that pride themselves on having a vibrant marketplace of ideas, converting data and insights into better business outcomes is a pressing and strategic challenge for senior executives.

How does an organization move from being data-rich to insight-rich— and capable of acting on the best of those insights? Big data is not enough, nor are clever analytics, to ensure that organizations make better decisions based on insights generated by analytic professionals. Some analysts' work directly influences business results, while other analysts' contributions matter much less. Rarely is the difference in impact due to superior analytic insights or larger data sets. Developing shrewd and scalable ways to identify and digest the best insights while avoiding the time traps of lazy data mining or "analysis paralysis" are new key executive competencies.

Information Overload and a Translation Task

How can data, decisions, and impact become more tightly integrated? A central irony, first identified in 1971 by Nobel Prize winner Herbert Simon, is that when data are abundant, the time and attention of senior decision makers become the scarcest, most valuable resource in organizations. We can never have enough time, but we can certainly have too much data. There is also a difficult translation task between the pervasive ambiguity of the executive suite and the apparent precision of analysts' predictions and techniques. Too often, analysts' insights and prescriptions fail to recognize the inherently inexact, unstructured, and time-bound nature of strategically important decisions. Executives sometimes fail to appreciate fully the opportunities or risks that may be expressed in abstract algorithms, and too often analysts fail to become trusted advisors to these same senior executives. Most executives recognize that models and analyses are reductive simplifications of highly complex patterns and that these models can sometimes produce overly simple caricatures rather than helpful precision. In short, while advanced analytic techniques are increasingly important inputs to decision making, savvy executives will insist that math and models are most valuable when tempered by firsthand experience, deep knowledge of an industry, and balanced judgments.

Limitations of Data-Driven Analysis

More data can make decision making harder, not easier, since it can sometimes refute long-cherished views and suggest changes to well-established practices. Smart analysis can also take away excuses and create accountability where there had been none. But sometimes, as Andrew Lang noted, statistics can be used as a drunken man uses a lamppost—for support rather than illumination. And sometimes, as the recent meltdowns in real estate, mortgage banking, and international finance confirm, analysts can become too confident in their models and algorithms, ignoring the chance of "black swan" events and so-called "non-normal" distributions of outcomes. It is tempting to forget that the future is certain to be different from the recent past but that we know little about how that future will become different. Mark Twain cautioned us, "History doesn't repeat itself; at best it sometimes rhymes." Statistics and analysts are rarely able to discern when the future will rhyme or be written in prose.

Some of the most important organizational decisions are simply not amenable to traditional analytic techniques and cannot be characterized helpfully by available data. Investments in innovation, for example, or decisions to partner with other organizations are difficult to evaluate ex ante, and limited data and immeasurable risks can be used to argue against such strategic choices. But of course the absence of data to support such unstructured strategic decisions does not mean these are not good choices—merely that judgment and discernment are better guides to decision making.

Many organizations will find it beneficial to distinguish more explicitly the various types of decisions, who is empowered to make them, and how. Many routine and tactical decisions, such as staffing, inventory planning, or back-office operations, can be improved by an increased reliance on data and by automating key parts of the decision-making process— by, for example, using optimization techniques. These rules and decisions often can be implemented by field managers or headquarters staff and need not involve senior executives. More consequential decisions, when ambiguity is high, precedent is lacking, and trade-offs cannot be quantified confidently, do require executive engagement. In these messy and high-consequence cases, when the future is quite different from the recent past, predictive models and optimization techniques are of limited value. Other more qualitative analytic techniques, such as field research or focus groups, and new analytic techniques, such as sentiment analysis and social network graphs, can provide actionable, near-real-time insights that are diagnostically powerful in ways that are simply not possible with simulations or large-scale data mining.

Even in high-uncertainty, high-risk situations, when judgment and experience are the best available guides, executives will often benefit from soliciting perspectives from outside the rarefied atmosphere of their corner offices. Substantial academic and applied research confirms that decisions made with input from different groups, pay grades, and disciplines are typically better than decisions that are not vetted beyond a few trusted advisors. Senior executives who find themselves inside "bubbles" of incomplete and biased information may be misled, as when business cases for new investments are grounded in unrealistically optimistic assumptions, or when a manager focuses on positive impacts for her business unit rather than the overall organization. To reduce this gaming and the risks of suboptimization, there is substantial value and insight gained by seeking out dissenting views from nontraditional sources. In strategically important and ambiguous situations, the qualitative "wisdom of crowds" is often a better guide to smart decision making than a slavish reliance on extensive data analysis—or a myopically limited range of perspectives favored by executives. Good analysts can play important roles too since they bring the rigor and discipline of the scientific method above and beyond any data they may have. The opportunity is to avoid the all too-common refrain: we're doing it because the CEO said so.

Many executives may need to confront the problem of information distortion. Often this takes the form of hoarding or a reluctance to share information freely and broadly across the organization. Its unhelpful twin, "managing up," may also manifest itself: sharing selectively filtered, makers. These practices can impair decisions, create silos, truncate learning, accentuate discord, and delay the emergence of learning communities. In the past, hoarding and managing up have been rational and were sometimes sanctioned; now, leadership means insisting that sharing information up and down the hierarchy, transparently and with candor, is the new normal. This is true both when insights confirm existing views and practices and also when the data and analysis clash with these. Conflicting ideas and competing interests are best handled by exposing them, addressing them, and recognizing that they can improve decisions.

Evolving a Data-Driven Learning Culture

For organizations that have relied on hard-won experience, memorable events, and other comfortable heuristics, the discipline of data-driven decision making may be a wholly new approach to thinking about how to improve business performance. As several chapters in this volume indicate, it is simply not possible to impose an analytic approach atop a company's culture. Learning to improve business performance through analytics is typically piecemeal and fragile, achieved topic by topic, process by process, group by group, and often in fits and starts. But it rarely happens without strong executive engagement, advocacy, and mindshare—and a willingness to establish data-driven decision making as the preferred, even default approach to answering important business questions.

Executives intent on increasing the impact and mindshare of analytics should recognize the scale and scope of organizational changes that may be needed to capture the value of data-driven decision making. This may require sweeping cultural changes, such as elevating the visibility, seniority, and mindshare that analytic teams enjoy across the company. It may mean investing additional scarce resources in analytics at the expense of other projects and teams, much as Procter & Gamble has done in recent years, and for which it is being well rewarded. It may also require repeated attempts to determine the best way to organize analytic talent: whether they are part of information technology (IT), embedded in business units, centralized into a Center of Excellence at headquarters, or globally dispersed. Building these capabilities takes time and a flexible approach since there are no uniformly valid best practices to accelerate this maturation.

Likewise, analytic priorities and investments will vary across companies, so there are clear opportunities for executives to determine top-priority analytic targets, how data and analysts are resourced and organized, and how decision making evolves within their organizations.

No Simple Recipes to Master Organizational Complexity

The chapters in this volume offer useful case studies, technical roadmaps, lessons learned, and a few prescriptions to "do this, avoid that." But there are many ways to make good decisions, and decision making is highly idiosyncratic and context dependent: what works well in one organization may not work in others, even for near-peers in the same businesses or markets. This is deeply ironic: we know that strong analytic capabilities can improve business results, but we do not yet have a rigorous understanding of the best ways for organizations to build these capabilities. There is little science in how to build those capabilities most efficiently and with maximum impact.

Smart decisions usually require much more than clever analysis, and organizational learning skills may matter more than vast troves of data. High-performing teams identify their biases, disagree constructively, synthesize opposing views, and learn better and faster than others. Relative rates of learning are important, since the ability to learn faster than competitors is sometimes considered to be the only source of sustainable competitive advantage. There is a corresponding, underappreciated organizational skill: a company's ability to forget. Forgetting does matter, because an overcommitment to the status quo limits the range of options considered, impairs innovation, and entrenches taken-for-granted routines. These "core rigidities" are the unwelcome downside to an organization's "core competencies" and are difficult to eradicate, particularly in successful firms. Time after time, in market after market, highly successful firms lose out to new products or technologies pioneered by emerging challengers. Blinded by past successes and prior investments, these incumbent companies may be overly confident that what worked in the past will continue to work well in the future. In short, while big data and sophisticated analyses are increasingly important inputs to better decisions, effective team-learning skills, an ability to learn faster than others, and a fierce willingness to challenge the status quo will increase the chance that databased insights yield better business outcomes.

Executives confront at least one objective constraint as they consider their approach to data-driven decision making: there is a pervasive shortage of deep analytic talent, and we simply cannot import enough talent to fill this gap. Estimates of this talent gap vary, but there is little reason to think it can be filled in the near term given the time involved in formal education and the importance of firsthand business experience for analysts to become trusted advisors. With some irony, Google's Hal Varian believes that statisticians will enjoy "the sexiest job for the next decade." Analysts who combine strong technical skills with a solid grasp of business problems will have the best choices and will seek out the best organizations with the most interesting problems to solve.

There is also an emerging consensus that many managers and executives who think they are already "data driven" will need to become much more so and may need deeper analytic skills to develop a more nuanced understanding of their customers, competitors, and emerging risks and opportunities. Much as an MBA has become a necessary credential to enter the C-suite, executives will increasingly be expected to have deeper knowledge of research methods and analytic techniques. This newly necessary capability is not about developing elegant predictive models or talking confidently about confidence intervals, but about being able to critically assess insights generated by others. What are the central assumptions and what events could challenge their validity? What are the boundary conditions? Is A causing B or vice versa? Is a set of conclusions statistically valid? Are the findings actionable and repeatable at scale? Is a Cronbach's alpha of 5% good or bad?

There is nothing automatic or easy about capturing the potential value of big data and smarter analyses. Across several industries, markets, and technologies, some few firms have been able to create competitive advantages for themselves by building organizational capabilities to unearth valuable insights and to act on the best of them. Many of these companies are household names—Starbucks, Walmart, FedEx, Harrah's, Expedia— and there is strong evidence that these investments have been financially prudent, richly strategic, and competitively valuable. Rarely did this happen without strong and persistent executive sponsorship. These leading companies invested in building scalable analytic capabilities—and in the communities of analysts and managers who comb through data, make decisions, and influence executives. These companies are not satisfied with their early successes and are pioneering new analytic techniques and applying a more disciplined approach to ever more of their operations. Embracing and extending this data-driven approach have been called "the future of everything." The opportunity now is for executives in other firms to do likewise: to capture the value of their information assets through rigorous analysis and better decisions. In addition to more efficient operations, this is also a promising path to identify new market opportunities, address competitive vulnerabilities, earn more loyal customers, and improve bottom-line business results.

Big data is a big deal; executives' judgments and smart organizational learning habits make big data matter more.

Read more IT Performance Improvement

This article is an excerpt from:

Industry reports indicate that by 2018, there will be a shortage of up to 180,000 analysts in the U.S. for examining "big data" problems. At the same time, emerging academic research shows that organizations who use business analytics to guide their decision making are more productive and experience higher returns on equity than their competitors who don't. Thus, big data problems and business analytics will be areas of growing interest in the years ahead, especially for addressing organizational, national, and societal issues.

Big Data and Business Analytics consists of case studies in "big data" domains including cybersecurity, finance, emergency management, healthcare, and international development. It is an easy-read for CEOs and senior managers who need to quickly grasp the key factors and trends to make their organizations more competitive. Business, management, and technology educators will also find this book to be helpful in their courses. The book contains case studies from Fortune 100 companies and U.S. Government agencies.