Strategy & LeadershipHow strategists use big data to support internal business decisions, discovery and productionThomas H. Davenport

Article information:To cite this document:Thomas H. Davenport , (2014),"How strategists use big data to support internal business decisions, discovery andproduction", Strategy & Leadership, Vol. 42 Iss 4 pp. 45 - 50Permanent link to this document:http://dx.doi.org/10.1108/SL-05-2014-0034

Downloaded on: 26 May 2015, At: 20:13 (PT)References: this document contains references to 0 other documents.To copy this document: permissions@emeraldinsight.comThe fulltext of this document has been downloaded 1296 times since 2014*

Users who downloaded this article also downloaded:Michael E. Prescott, (2014),"Big data and competitive advantage at Nielsen", Management Decision, Vol. 52 Iss 3 pp.573-601 http://dx.doi.org/10.1108/MD-09-2013-0437Victoria Louise Lemieux, Brianna Gormly, Lyse Rowledge, (2014),"Meeting Big Data challenges with visual analytics:The role of records management", Records Management Journal, Vol. 24 Iss 2 pp. 122-141 http://dx.doi.org/10.1108/RMJ-01-2014-0009Stephen Fox, Tuan Do, (2013),"Getting real about Big Data: applying critical realism to analyse Big Data hype", InternationalJournal of Managing Projects in Business, Vol. 6 Iss 4 pp. 739-760 http://dx.doi.org/10.1108/IJMPB-08-2012-0049

Access to this document was granted through an Emerald subscription provided by 460696 []

For AuthorsIf you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors serviceinformation about how to choose which publication to write for and submission guidelines are available for all. Please visitwww.emeraldinsight.com/authors for more information.

About Emerald www.emeraldinsight.comEmerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio ofmore than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of onlineproducts and additional customer resources and services.

Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on PublicationEthics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.

*Related content and download information correct at time of download.

Dow

nloa

ded

by G

rand

Can

yon

Uni

vers

ity A

t 20:

13 2

6 M

ay 2

015

(PT)

How strategists use big data to supportinternal business decisions, discoveryand productionThomas H. Davenport

Thomas H. Davenportsmost recent book is BigData@Work: Dispellingthe Myths, Uncovering theOpportunities (HarvardBusiness Review Press,2013). He is a Professorof Information Technologyand Management atBabson College, a fellowof the MIT Center forDigital Business andcofounder and Director ofResearch at theInternational Institute forAnalytics. This article isbased on research forthe book.

Over the last 45 years, the general activity of making sense of data has evolved fromdecision support, to executive support, to online analytical processing, to businessintelligence, to analytics and now to big data (see Table 1-3). But what is big

data? A general definition is, the collection and interpretation of massive data sets, madepossible by vast computing power that monitors a variety of digital streams such assensors, marketplace interactions and social information exchanges and analyses themusing smart algorithms. In short, big data searches comb massive amounts of digitalinformation looking for useful correlations. For example, colloquially speaking, big datacould be used as a smart way to discover which are the most valuable needles in a needlefactory or who recently has been sewing with needles near haystacks.

Technical issues aside, from a strategic management perspective, how is big data differentfrom previous data analysis systems? The primary purpose behind traditional small dataanalytics that all managers are more or less familiar with is to support internal businessdecisions. Examples include: What offers should you present to a customer? Whichcustomers are most likely to stop being customers soon? How much inventory should wehold in the warehouse? How should we price our products? With some creative tweaking,big data analysis can be used to produce insights on these issues. But big data also offersa promising new dimension: to discover new opportunities to offer customers high-valueproducts and services.

Big data is notably different from traditional information management and analytics in thisregard. Instead of just creating reports or presentations that advise senior executives oninternal decisions, big data scientists commonly work on customer-facing products andservices.

This is particularly true in big data start-ups, but its also the case in larger, moreestablished companies. For example:

Reid Hoffman, the cofounder and chairman of LinkedIn, made his data scientists a lineproduct team for the company, and they have developed such products as People YouMay Know, Groups You May Like, Jobs You May Be Interested In, Whos Viewed MyProfile, and several others.

General Electric is primarily focused on using big data for improving services and isalready using data science to optimize the service contracts and maintenance intervalsfor industrial products.

Google, of course the ultimate big data firm uses data scientists to refine its coresearch and ad-serving algorithms.

Zynga uses data scientists to target games and game-related products to customers.

DOI 10.1108/SL-05-2014-0034 VOL. 42 NO. 4 2014, pp. 45-50, Emerald Group Publishing Limited, ISSN 1087-8572 STRATEGY & LEADERSHIP PAGE 45

Dow

nloa

ded

by G

rand

Can

yon

Uni

vers

ity A

t 20:

13 2

6 M

ay 2

015

(PT)

Netflix created the well-known Netflix Prize for the data science team that could optimizethe companys movie recommendations for customers.

The testing firm Kaplan uses its data scientists to begin advising customers on effectivelearning and test-preparation strategies.

These companies big data efforts are directly focused on products, services andcustomers. This has important implications, of course, for the organizational locus of bigdata and the processes and pace of new product development.

For those uses of big data that do involve internal decisions, new management approaches arestill necessary, but not yet fully resolved in practice. This is because big data just keeps onflowing. In traditional decision support situations, an analyst takes a pool of data, sets it asidefor analysis, comes up with a model, and advises the decision maker on the results. However,with big data, the data resembles not so much a pool as an ongoing, fast-flowing stream.Therefore, a more continuous approach to sampling, analyzing and acting on data isnecessary.

This is particularly at issue for applications involving ongoing monitoring of data on stakeholderperceptions, as in social media sentiment analysis. Sentiment analysis allows an organizationto assess whether the comments about its brands and products in blogs, tweets, andFacebook pages are positive or negative on balance. One potential problem with suchmonitoring applications is the tendency for managers to view a continuing stream of analysisand reports without making any decisions or taking any action. Sentiment is up . . . no, itsdown . . . hooray, its back up again! For ongoing monitoring work, there should be processesfor determining when specific decisions and actions are necessary when, for example, datavalues fall outside certain limits. Such information helps to determine decision stakeholders,decision processes and the criteria and timeframes for which decisions need to be made.

Even the United Nations an organization typically not known for its agility is getting in on thisnew approach to deciding. The UNs Global Pulse innovation lab has developed a bigdata-related tool called HunchWorks, which is clearly a monitoring-oriented application of bigdata. The lab describes HunchWorks as the worlds first social network for hypothesisformation, evidence collection, and collective decision-making.[1] The idea is that as databegin to reveal a trend or finding say, for example, weather data suggesting a drought thatcould lead to famine in a part of Africa an analyst would post the hunch and the data on whichit was based, and others could weigh in with new analyses and data. Such suggestivehypotheses have been described as digital smoke signals.[2] One goal is to determine howlikely the hunch is to be worthy of detailed analysis and action. But the idea that the UN wouldhave a system for circulating data-driven hunches marks a major change in that organizationsculture.

Table 1-3 Terminology for using and analyzing dataTerm Time frame Specific meaning

Decision support 19701985 Use of data analysis to support decision makingExecutive support 19801990 Focus on data analysis for decisions by senior

executivesOnline analyticalprocessing (OLAP)

19902000 Software for analyzing multidimensional datatables

Businessintelligence

19892005 Tools to support data-driven decisions, withemphasis on reporting

Analytics 20052010 Focus on statistical and mathematical analysisfor decisions

Big data 2010present Focus on very large, unstructured, fast-movingdata

Source: From Big Data@Work: Dispelling the Myths, Uncovering the Opportunities by Thomas H.Davenport (Harvard business Review Press, 2013), by permission

PAGE 46 STRATEGY & LEADERSHIP VOL. 42 NO. 4 2014

Dow

nloa

ded

by G

rand

Can

yon

Uni

vers

ity A

t 20:

13 2

6 M

ay 2

015

(PT)

Whether the analysis and decision processes are social or individual, the continuing stream ofbig data suggests that organizations need to think about new ways of making decisions withthis resource. If its worth investing in the collection and analysis of big data, its also worththinking about how the outcome of the analysis will have an impact on decisions and actions.

Customer satisfactionNowadays its possible to use big data analysis methods on new, less-structured datasources and utilize the resulting information to make better internal decisions. For example,United Healthcare, a large health insurance company, is focused on the customersatisfaction and attrition issue. A lot of data about how the companys customers feel issitting in recorded voice files from customer calls to call centers. The level of customersatisfaction is increasingly important to health insurers because it is being monitored bystate and federal government groups and published by organizations such as ConsumersUnion. In the past, that valuable data from calls couldnt be analyzed.

Now United is turning it into text and then analyzing it with natural language processingsoftware, a way to extract meaning from text. The analysis process can identify customerswho use terms suggesting strong dissatisfaction. The insurer can then make some sort ofintervention perhaps a call exploring the source of the dissatisfaction. The decision is thesame as in the past how to identify a dissatisfied customer but the tools are different.

Customer journeysA number of major financial services firms Wells Fargo, Bank of America, and Discover arealso using big data to understand aspects of the customer relationship that they couldntpreviously get at. In that industry as well as several others, including retail the big challengeis to understand multichannel-customer relationships. They are using customer journeysthrough the tangle of websites, call centers, tellers and other branch personnel to betterunderstand the paths that customers follow through the organization, and how those pathsaffect attrition or the purchase of particular financial services.

The data sources on multichannel customer journeys are unstructured or semi-structured. Theyinclude website clicks, transaction records, bankers notes, and voice recordings from callcenters. The volumes are quite large 12 billion rows of data for one of the banks. The firms arebeginning to understand common journeys, describing them with segment names, ensuringthat the customer interactions are high quality, and correlating journeys with customeropportunities and problems. Its a complex set of problems and decisions to analyze, but thepotential payoff is high half a billion dollars is the estimate at one of the banks.

Supply chain riskBusiness decisions using big data can also involve other traditional areas for analytics, such assupply chains, risk management, or pricing. The factor that makes these big, rather than small,data problems is the use of large volumes of external data to improve the analysis. In supplychain decisions, for example, companies are increasingly using external data to measure andmonitor supply chain risks. External sources of supplier data can furnish information onsuppliers technical capabilities, financial health, quality management, delivery reliability,weather and political risk, market reputation, and commercial practices. The most advancedfirms are monitoring not only their own suppliers but their suppliers suppliers.

Big data [. . .] offers a promising new dimension: to discovernew opportunities to offer customers high-value productsand services.

VOL. 42 NO. 4 2014 STRATEGY & LEADERSHIP PAGE 47

Dow

nloa

ded

by G

rand

Can

yon

Uni

vers

ity A

t 20:

13 2

6 M

ay 2

015

(PT)

Competitive intelligenceCompetitive and market intelligence used to be a rather intuitive exercise, but big data isbeginning to change that approach. If you can get more detailed data and do more systematicanalysis on it, the activity will probably improve your strategic decisions. As Joey Fitts, CEO ofMatters Corp., explains, Historically, market and industry intelligence consisted primarily ofcompany directories indicating who companies are basic information such as their physicallocation, phone numbers, SIC code, credit score, etc., but now we can explain whatorganizations do in the market. Market factors which were hidden are now visible data,enabling trend analysis, benchmarking, segmentation, modeling, and recommendations. Its amuch broader set of data, at significantly greater scale, and more real-time. Companies canlead rather than react.[3]

For example, a leading software provider sought to better understand the landscape of partnersupport for competitive software platforms versus their own offerings. The company usedMatters Corp., which monitors market activities in that industry, to untangle the web ofunstructured data on partnerships and platform support. The data revealed that thecompetition was garnering as much as three times the partner attention and platform support.In customer terms, this meant that a much richer ecosystem of partner services technologyadvisory services, assessments, implementation, applications, solutions, technologyextensions and support was available to customers of the competition than they could offerto their own customers. This recognition resulted in the company leadership proposing $100million in additional budget to close the competitive partner gap. The company was able to usethe same tools to monitor their relative impact on partner capacity and watch the gap close overtime.

PricingPricing has a long history of applying analytics successfully. Almost every airline and hotelchain, for example, now uses pricing optimization tools to determine the best price for aseat or room. Pricing optimization was originally done with internal structured data on whatgoods historically sold at what price, and thats still a key element. But pricing softwarecompanies such as PROS now often incorporate external, and somewhat less structured,big data into the algorithm. For example, a PROS user in the oil industry can incorporateweather data (which would influence consumer demand) and competitor prices, which canoften be scraped from the internet, into pricing algorithms.

Discovery and experimentationPerhaps the highest and best use of big data mining it for discovery and experimentation is still in the learning phase in most companies. To date, the primary focus of business andtechnology organizations has been to automate data analysis processes such as marketing,sales and service. Analytics has been used to understand and tune such business processes,keeping management informed and alerting them to anomalies exception reporting hasbeen a key aspect of business intelligence.

Big data flips this approach on its head. The basic tenet is that the world and the data thatdescribes it are in a constant state of change and flux, and those organizations that canrecognize and react quickly and intelligently have the upper hand. Increasingly the most prized

Big data [. . .] resembles not so much a pool as an ongoing,fast-owing stream. Therefore, a more continuous approachto sampling, analyzing and acting on data is necessary.

PAGE 48 STRATEGY & LEADERSHIP VOL. 42 NO. 4 2014

Dow

nloa

ded

by G

rand

Can

yon

Uni

vers

ity A

t 20:

13 2

6 M

ay 2

015

(PT)

business and IT capabilities are discovery and agility rather than stability. Data scientistsworking with big data tools and technologies will be able to continuously mine new and existingdata sources for patterns, events and opportunities at an unprecedented scale and pace.

Increasingly, corporate strategists are recognizing that big data architecture andmanagement should be designed so that discovery and analysis is the first order ofbusiness.[4] Data scientists, as well as general business analysts, need continuous accessto an analytics platform that supports ready insight to enterprise and external data. Theplatform needs to facilitate integrating new data, ad hoc queries and visualization toaccelerate human understanding. As valuable insights emerge from this platform, theybecome the requirements for changes to production systems and processes.

Companies also need to adopt new methodologies for insight and data-based productdevelopment. Traditional waterfall highly structured approaches that only yield a result atthe end of a long process methods have been increasingly forced out of systemdevelopment processes in favor of faster and more flexible Agile/Scrum processes. SuchAgile approaches, in which relatively little time is spent specifying a system up front andmore emphasis is placed on creating small deliverables quickly, can also apply to analyticsand big data. Imprecise, slow requirements gathering for a new analytical system orprocess is replaced by iterative experimentation, insight and validation.

Facilitating big data discoveryDiscovery is most often done in business units rather than IT organizations, typically bypeople who are focused on innovation, product development and research. Somecompanies organize them into data labs or analytics sandboxes or a group with asimilar name. They are typically found within the most data-intensive business units of theirorganizations for example, the online or distribution channels functions within banks ormarketing functions in retail and consumer products. They know the latest tools, know howto construct and monitor experiments with data, and arent averse to failing. The classicview of a data scientist is someone who fits this profile.

The desired outcome of data discovery is an idea a notion of a new product, service, orfeature, or a hypothesis with supporting evidence that an existing model can beimproved. There will be more of the incremental improvements than the grandbreakthroughs; most discoveries are relatively minor. One might find a new factor to betteridentify customers who are about to leave, or how to better target an offer. If you keep atit and have good people and a supportive culture, youll eventually find something big.

The payoff at production timeThe ultimate value of an insight follows from the decision to use it or not use in productionprocesses and systems. Once an insight has been extracted from data, it needs to beclassified as irrelevant to the business, interesting but not useful or the basis for an action.Similarly, if the big data analysis has led to a new product or feature, it needs to be adoptedor dropped or considered for adoption at a later date.

A number of major nancial services rms are usingcustomer journeys through the tangle of websites, callcenters, tellers and other branch personnel to betterunderstand the paths that customers follow through theorganization, and how those paths affect attrition or thepurchase of particular nancial services.

VOL. 42 NO. 4 2014 STRATEGY & LEADERSHIP PAGE 49

Dow

nloa

ded

by G

rand

Can

yon

Uni

vers

ity A

t 20:

13 2

6 M

ay 2

015

(PT)

The production stage for big data applications is just that putting the application intoproduction processes at scale. It might mean merging new data and scoring approachesinto a pricing algorithm, or moving a new product feature from beta release to a full-featuredoffering. It requires scale, reliability, security and all those pesky attributes that customers,partners and regulators care about.

Of course, not all discovery ideas should go into production. Not all ideas fit anorganizations culture or processes or have a clear payoff. If youre sending even half ofyour discovery projects into production, youre probably being a bit too liberal.

But things can also go astray if not enough discovery projects go into production. For example,I recently conducted interviews at a health-care organization that has done a large number ofdiscovery projects involving big data, including capturing and analyzing physicians notes,radiology images and patient behaviors. Many of these projects held considerable potential.However, the organizations electronic medical record (EMR) system, while being well suited tocapturing transaction data, had limited capabilities to export data for analysis. Each attempt todo so involved large amounts of time, money and frustration. As a result, the discovery projectsseldom made it into production at all. This organization knew it had a problem in the discovery/production handoff. It had invested a lot in its EMR system, but it hadnt yet marshaled theresources and time to create an enterprise warehouse for clinical data.

Whatever task its applied to internal decisions, discovery or production the return oninvestment from big data comes from the processing and analysis of it and the insights,products and services that emerge and become recognized as adding value. The comingsweeping changes in big data technologies and management approaches need to beaccompanied by similarly dramatic shifts in how data supports decisions and product/service innovation. There is little doubt that big data analytics can transform organizations,and the firms that recognize the full extent of their opportunities will seize the most value.

Notes1. http://www.unglobalpulse.org/technology/hunchworks.

2. Steve Lohr, Searching big data for digital smoke signals, New York Times, August 7, 2013,http://www.nytimes.com/2013/08/08/technology/development-groups-tap-big-data-to-direct-humanitarian-aid.html.

3. Correspondence and conversations in 2013 between the author and Joey Fitts, CEO of Matters Corp.

4. I am grateful to Paul Barth for many of his ideas on experimentation. Some of them appeared inThomas H. Davenport, Paul Barth, and Randy Bean, How big data is different, MIT SloanManagement Review (Fall 2012), http:// sloanreview.mit.edu/the-magazine/2012-fall/54104/how-big-data-is-different/

Corresponding authorThomas H. Davenport can be contacted at: jdevoll@hbr.org

To purchase reprints of this article please e-mail: reprints@emeraldinsight.comOr visit our web site for further details: www.emeraldinsight.com/reprints

Increasingly, corporate strategists are recognizing that bigdata architecture and management should be designed sothat discovery and analysis is the rst order of business.

PAGE 50 STRATEGY & LEADERSHIP VOL. 42 NO. 4 2014

Dow

nloa

ded

by G

rand

Can

yon

Uni

vers

ity A

t 20:

13 2

6 M

ay 2

015

(PT)

How strategists use big data to support internal business decisions, discovery a ...Customer satisfactionCustomer journeysSupply chain riskCompetitive intelligencePricingDiscovery and experimentationFacilitating big data discoveryThe payoff at production time