Accenture: Outlook What C Suite Should Know About Analytics 2011
1 like647 views
This article discusses how analytics will fundamentally change business decision making in the future. It argues that analytics will establish an empirical, data-driven basis for decisions at all levels of an organization, replacing many decisions currently made based on assumptions without evidence. The article outlines five dimensions along which companies of the future will operate differently due to sophisticated use of analytics: high analytical literacy, volatility, fine-grained decision making, integrated awareness across data silos, and responsibility for outcomes informed by all available data.
Accenture: Outlook What C Suite Should Know About Analytics 2011
- 1. This article originally appeared in the 2011, No. 1, issue of The journal of high-performance business O n the Edge W hat the C-suite should know about analytics By Kishore S. Swaminathan Chief Scientist Accenture A nalytics is a transformational phenomenon that will fundamentally change how business discourse will be conducted and decisions will be made.
- 2. Case study after case study has buy more of our services or express confirmed the value proposition their “better impression” in any other for analytics across a wide range way. Not unexpectedly, I was asked of business functions, including by the powers that be if this really pricing, demand prediction, targeted was a battle that I wanted to fight. marketing, supply chain optimiza- tion, CRM and HR. In my view, I chose this example to illustrate how analytics is something much more average, mundane decisions are made than a technology with an ROI; it’s in organizations daily based on well- a transformational phenomenon intentioned, plausible yet armchair that will fundamentally change how theories—those that, like Aristotle’s, business discourse will be conducted lack any empirical evidence. While and decisions made. An analogy highly specialized functions such may help in understanding why. as pricing or customer segmentation may be based on sophisticated If you drop a feather and a rock models and empirical data, my con- at the same time from the same tention is that the long-term impact height, which will hit the ground of analytics will be in instilling first? At one point in history, this a culture of data-driven decision was a question for philosophers to making at all levels of an enterprise. resolve. Aristotle opined that the Or, put more bluntly, business rock, because it was heavier, would proposals and decisions—big or fall faster and hit the ground first. small—will have to provide satis- Aristotle’s armchair wisdom was not factory answers to this question: questioned until the 16th century, “Do we think this is true or do we when Galileo, through cleverly know?” (This particular formulation designed experiments, proved him is attributed to Gary Loveman, CEO wrong and established an empirical of Harrah’s Entertainment.) basis for answering such questions about the physical world. A sophisticated and analytically oriented enterprise of the future Much the same way that an em- will behave and operate differently pirically based scientific method from today’s enterprise along five became the basis of our understand- major dimensions. ing of the world around us, analytics will eventually bring empiricism High analytical literacy into business discourse and dethrone Data is a double-edged sword. When many of today’s business practices. properly used, it can lead to sound and well-informed decisions. When Mundane decisions improperly used, the same data can Recently, I received a memo saying lead not only to poor decisions but that all employees at my location to poor decisions made with high would be required to keep their confidence that, in turn, could lead offices clean, subject to inspection to actions that could be erroneous every other Friday. I wanted an and expensive. Let’s consider some explanation, so I asked if there was specific examples. any data to show that clean offices lead to higher productivity. When one has access to real-time data, it’s tempting to make real-time My question, of course, was side- decisions. For instance, if you are a stepped, and I was told that clean retailer and you have real-time access offices would make a better impres- to sales data from cash registers from sion on clients. Undeterred, I asked all your stores and real-time access 2 Outlook 2011 if there was any data to show that to your inventory in your warehouse, Number 1 clients walking through our offices you could be tempted to run sales
- 3. promotions on the fly and manage When you have fine-grained vis- your supply chain in tandem to sup- ibility into your processes, customers, port your real-time promotions. suppliers and competitors, you have the ability to make very fine-grained However, this is unlikely to work decisions. In fact, your decision rules because three types of events—your can capture subtleties such as “stock decisions, the ensuing customer more beer on Sunday nights in loca- behavior and supply chain events— tions where the home football team is operate in different timeframes, so on a winning streak.” Such decisions making decisions any faster than the are highly context-sensitive and can slowest-moving event could be use- change as rapidly as the fortunes of less at best and dangerous at worst. the football team. Another problem with data and ana- lytics is that they give you very fine- Volatility—or rapidly changing grained visibility into your business decisions that are context- and processes, and you could be tempted time-sensitive—will be a big chal- to overoptimize the processes. Highly lenge for enterprises. Decisions optimized processes—just-in-time are no longer easily explainable; inventory being an example—are capital investments cannot be very fragile because circumstances based on mass repeatability but beyond your control could arise, and must cater to endemic volatility. there is little room for error. Integrated awareness A third problem is what’s known as Today’s enterprises have more in- “oversteering,” or making decisions formation than they can act upon when none is needed. So, for example, because the information is siloed in your data could tell you that a project so many ways: technologically (data is behind schedule, which, in turn, in different systems that cannot be may lead you to berate the project brought together), organizationally manager or tell your stakeholders (data in different governance units that the project will be delayed. that cannot be brought together) or Yet neither of these actions may be by ownership (inside versus outside necessary if the project has contin- the enterprise). The enterprise of gency built in, if the status update the future will be (or will be forced has a different frequency than your to be) “conscious” in the sense that sampling frequency or if perhaps the it will know that it must integrate employees who are aware of the proj- everything it has access to. ect delay will put in more work time to get the project back on schedule. As an extreme example of “inte- grated awareness,” let’s consider Volatility pharmaceuticals, an industry that Businesses thrive on stability and has traditionally relied on clinical repeatability. Stable and repeatable trials data as a means of estab- processes justify large-scale capital lishing the efficacy and the side expenses; they justify large-scale effects of a drug. employee training; and they reduce cognitive overhead because processes A pharmaceuticals company today and decisions do not change and can legally and morally claim im- hence their rationale does not have munity from any adverse effect of to be explained repeatedly. a drug that was not revealed during clinical trials—in other words, any By contrast, an analytically based information that it did not explicitly enterprise of the future will have to collect as part of a clinical trial pro- 3 Outlook 2011 be designed around volatility rather tocol. But in a world of blogs and Number 1 than repeatability. social networks, where people share
- 4. this information unprompted and in The third cause of analysis-paralysis public, it will become both a respon- is the fact that most companies sibility and an obligation of phar- do not know or articulate their risk maceuticals companies to monitor tolerance clearly and are much public sources and integrate the public more likely to penalize failed information with their own clini- action than inaction. As a result, cal data. (For more on the business many managers do not act unless impact of social media, see Outlook there is enough data to assure 2011, No. 1.) them of successful outcomes. An analytically literate organization “I should have known” (either for will have a firm grasp of its risk regulatory or competitive reasons) tolerance. With guidelines and will be the new normal, replacing models for action under uncertainty, the “I did not know” or “I could not it will restore the symmetry be- have known” approach to awareness tween how it treats failed action and information integration. and inaction. The end of analysis-paralysis Intuition’s new pulpit In the future, businesses will likely Empiricism and analytics sound be run by managers and leaders who a death knell for such vaunted are no-nonsense empiricists; they business traits as intuition, gut feel, won’t move a finger until after all the killer instinct and so forth, right? relevant data has been gathered and analyzed. A recipe for organizational Wrong. “analysis-paralysis”? This is not an unreasonable fear. But though it may Science is purely empirical and seem counterintuitive, an empirical dispassionate, but scientists are not. enterprise with high analytical Science is objective and mechanical, literacy is less likely to fall prey to but it also values scientists who are this malady than today’s enterprises. creative, intuitive and can take a leap of faith. There are three very distinct ways that organizations can fall into Data, by itself, can be interpreted in the analysis-paralysis trap. One is many ways. Imagine a physical or a managerial tendency to “over-fit business phenomenon that produces the curve”—a statistical term that the following sequence of data: 1, 2, refers to the diminishing value of 6, 24, 33. Perhaps it’s a factorial additional data once a pattern (or sequence with 33 as noise, or a curve, in the graphic sense) has been sequence where every fourth term found. Data collection has a price, is twice the multiple of the previous inaction has a price and an analyti- three. Or perhaps every fifth terms cally literate organization will clearly if the sum of the previous four. understand the cost of over-fitting. All are indeed correct. To prove or The second cause of analysis-paraly- disprove any theory, you need the sis is waiting for data that simply does next several terms of the sequence. not exist, which reflects an inability A good scientist knows when there to design experiments to generate is enough data to warrant a theory, the needed data. As mentioned above, when there isn’t, what new data to experimentation has a price and in- gather and how to design an experi- action has a price, so an analytically ment to gather the right data. literate organization will be charac- terized by a clear understanding Apple’s Steve Jobs is known to ex- 4 Outlook 2011 of data gaps and the value of experi- plicitly discount the value of surveys Number 1 mentation to break the logjam. and focus groups for designing new
- 5. products. How do you explain this It should be noted that some prod- apparent anti-empiricism? ucts—in Apple’s case, it was the Newton—do not succeed and are One explanation is that, much terminated. Intuition, creative leaps like a creative scientist, people and clever experimentation are not like Jobs recognize when there is incompatible with empiricism; in not enough data or the right kind fact, the value of these traits will be of data to form a theory. They even better understood in the future recognize that, for completely new enterprise by analogy to theoretical lines of products that will change and experimental scientists. a user’s experience or behavior, the only useful data is experien- The enterprise of the future, based on tial data, not commentary and re- empiricism and analytical decision actions from those who have never making, will indeed be considerably used the product. different from today’s enterprise. You may well ask: “Do you think Jobs and people like him are akin to this is true or do you know?” scientists who recognize what type of data is needed to support a theory Touché. (in this case, whether a product will succeed), recognize that such data Kishore S. Swaminathan is based cannot be gathered through focus in Beijing. groups (one type of experiment) and boldly design new types of ex- k.s.swaminathan@accenture.com periments (release the product and gather experiential data). Outlook is published by Accenture. © 2011 Accenture. All rights reserved. The views and opinions in this article should not be viewed as professional advice with respect to your business. Accenture, its logo, and High Performance Delivered are trademarks of Accenture. The use herein of trademarks that may be owned by others is not an assertion of ownership of such trademarks by Accenture nor intended to imply an association between Accenture and the lawful owners of such trademarks. For more information about Accenture, please visit www.accenture.com 5 Outlook 2011 Number 1