DELTA Plus Model & Five Stages Of Analytics Maturity: A Primer

1y ago
128 Views
3 Downloads
2.24 MB
12 Pages
Last View : 2d ago
Last Download : 4d ago
Upload by : Nadine Tse
Transcription

COMPLIMENTARYEDITIONDELTA Plus Model & Five Stages ofAnalytics Maturity: A PrimerADAPATED FROM ANALYTICS AT WORK: SMARTER DECISIONS, BETTER RESULTS(DAVENPORT, HARRIS, MORISON, 2010) AND COMPETING ON ANALYTICS: THE NEWSCIENCE OF WINNING (DAVENPORT AND HARRIS, 2017 & 2007)TOM DAVENPORTUPDATED: AUGUST 2018

DELTA Plus Model & Five Stages of Analytics Maturity: A PrimerBIG IDEAS– The DELTA Plus Model Framework encompasses the fivefoundational elements of a successful analytics program(Data, Enterprise, Leadership, Targets, and Analysts) andintroduces two new elements (Technology and AnalyticalTechniques) required for high performance.Stages of Analytical Maturity, from ‘Analytically Impaired’to ‘Analytical Competitor.’– As organizations build capabilities across the DELTA Pluselements, they become more analytically mature andcompetent.– Enterprises committed to using analytics and data totransform their businesses progress through the FiveIntroductionAs enterprises of all shapes and sizes commit toharnessing the power of data and analytics totransform all aspects of their businesses, leadershipwill inevitably ask these questions: How good are we at using data and analyticsthroughout our enterprise? Are we actually as goodas we think? Are we ahead of or behind our nearest competitors?Are other industries ahead of ours? Are we moving toward becoming an analyticalcompetitor? How can we set a path forward without knowingwhere we stand today?The Five Stages of Analytics Maturity and the DELTAModel have become the industry standard frameworksfor assessing analytics maturity. The five stages ofanalytics maturity were introduced in 2007 by TomDavenport and Jeanne Harris in their book, Competingon Analytics: The New Science of Winning. The DELTAModel was introduced in 2010 by Tom Davenport,Jeanne Harris and Bob Morison in their book, Analyticsiianalytics.comCopyright 2018 International Institute for Analyticsat Work: Smarter Decisions, Better Results. Bothframeworks were updated by Tom Davenport andJeanne Harris in their 2017 revision of Competing onAnalytics. Two new components were added to theDELTA model, creating the DELTA Plus model.The purpose of this research brief is to summarize thekey elements of DELTA Plus and five stages of analyticsmaturity, and discuss how these two frameworks canbe used to understand analytical maturity in yourorganization. For additional discussion on thesetopics, IIA recommends reading Analytics at Work:Smarter Decisions, Better Results (2010) and Competingon Analytics: The New Science of Winning (2017).DELTA Plus ModelThe original DELTA model featured five elements thatmust be in alignment for organizations to succeedwith their analytics initiatives. Without alignment,organizations run the risk of poor or limited results. Tomake real progress and become a data-drivenorganization, the capabilities and assets of these fiveelements must evolve and mature.2

DELTA Plus Model & Five Stages of Analytics Maturity: A PrimerThe five elements of a successful analytics program, asstated in Analytics at Work, are:D for accessible, high quality dataE for an enterprise orientation to managing analyticsL for analytical leadershipT for strategic targetsA for analystsThe continued growth of big data, coupled with theintroduction of new analytics techniques like machinelearning, means there are two additional elements(the Plus factors) that should also be considered:T for technologyA for analytics techniquesDATAFor meaningful analytics, data must be organized,unique, integrated, accessible, and of high quality. Ofcourse, not all organizations have an environment thatiianalytics.comCopyright 2018 International Institute for Analyticsencompasses all the important elements of data, butit’s important to know what to pursue to create thelargest opportunity. The way an organization’s data isstructured influences the type of analysis that can bedone. The ability of an organization to structure andleverage unstructured data also influences the typeand value of analytics that is done. The same is true fordata uniqueness – if an organization can also gatherunique data outside what other companies haveaccess to, then they have an analytical edge and moreopportunity in their analyses. Organizations also needto integrate their data across organizational silos andboundaries. Most organizations have multipletransaction systems in different business units andfunctions, and to fully understand organizationalperformance data from all of them needs to becombined and harmonized.It is also no secret that many organizations face dataquality issues. Once data has been cleaned andintegrated, it must be made accessible to theorganization for analytical purposes. Simply put,analysis cannot be done if the data cannot be located3

DELTA Plus Model & Five Stages of Analytics Maturity: A Primerand accessed. Data warehouses or Hadoop-based datalakes are the primary means to allow analysts andnon-analysts to access data. These repositories can bedeployed on premise, in the cloud or in a hybrid mix ofthe two. Finally, if an enterprise is becoming moremature within all aspects of its data environment, itimplements a dynamic governance strategy to ensurehigh-quality and well-managed data across theorganization.“Without a broad businessperspective, a company cannotaddress the strategic issues atthe core of business performanceand organizationalcompetitiveness.”ENTERPRISE-Tom DavenportAnalytical competitors take an enterprise approach tomanaging systems, data and people. They havecoordinated approaches relying on enterprise-levelorganizational structures, resource allocations andplans.To embrace this approach a company must advocate asingle and consistent perspective for analytics acrossthe organization. This is accomplished by setting ananalytics strategy and building a road map for strategyimplementation. Integrating data and managing aunified data and analytics platform are keycomponents of an analytics road map, as is cultivatinga culture of analytics across the organization.Perspectives from individual managers and businessunits/functions that do not support or advance theenterprise view must be discouraged and replacedwith a single, enterprise wide view of analytics.If analytics goals are not centrally established,organizational silos can develop and lead toduplicated efforts and tools, errors in analysis,ineffective use of resources, conflict among differentgroups, and increased complexity with analyticsprojects. An enterprise approach to analytics willgreatly increase the organization’s competitiveness.iianalytics.comCopyright 2018 International Institute for AnalyticsLEADERSHIPAnalytical organizations have leaders who fullyembrace analytics and lead company culture towarddata-driven decision-making. Beyond the CEO or othertop executives, all levels of leadership within theorganization should support analytics. This isimportant for cultural acceptance of analytics acrossthe enterprise, as well as the accomplishment ofanalytics initiatives. In Analytics at Work, authorsDavenport, Harris and Morison note 12 traits thatanalytical leaders exhibit in analytically 2.Possess people skillsPush for more data and analysisHire smart people, and give them creditSet a hands-on exampleSign up for resultsTeachSet strategy and performance expectationsLook for leverageDemonstrate persistence over timeBuild an analytical ecosystemWork along multiple frontsKnow the limits of analytics4

DELTA Plus Model & Five Stages of Analytics Maturity: A PrimerTARGETSVirtually no organization can afford to be equallyanalytical in all parts of its business. Analytics effortsmust be aligned with specific, strategic targets that arealso aligned with corporate objectives. Organizationswill get lost in all the business opportunities thatanalytics can support if they do not focus on a fewinitial and purposeful use cases and applications.Choosing these targets based on the organization’sstrategic plan is helpful, but not always easy. What istypically required is a group of executives thatunderstands both the business and the analyticalpossibilities for improving it. Enterprises can alsosurvey internal employees for ideas, as well as externalgroups to help understand industry and analyticaltrends. Looking beyond one’s industry is also helpfulto find opportunities in common, cross-industryapplications.When determining what targets to choose, leadersshould narrow in on the best options. This oftenrequires several steps. Leaders need to think about thebig picture for where the business is going, create asystematic inventory of possibilities, and thenprioritize potential uses of analytics based on thebenefit for and capabilities of the organization. Oncean enterprise is mature, its targets become embeddedin the strategic planning process, and are consideredbusiness initiatives, not just analytics initiatives. If theorganization is successful with analytics, its targetscan broaden over time.ANALYSTSOrganizations require analytical talent that covers arange of skills from employees capable of basicspreadsheets to accomplished data scientists. InAnalytics at Work, four analytical types of people aredefined, all of whom play an important role in aniianalytics.comCopyright 2018 International Institute for Analyticsorganization: analytical champions, analyticalprofessionals (now often known as data scientists),analytical semiprofessionals, and analytical amateurs.Recruiting analysts and data scientists can be quitedifficult today and retaining these employees evenmore challenging. Such professionals must havequantitative and technical skills, business knowledge,interpersonal skills, and the ability to coach otherswho may not understand analytics. They also must beadept at navigating new analytics techniques such asmachine learning and AI. Once the right people are inplace, keeping them motivated with creative andchallenging projects is crucial.Of course, the perfect analyst or data scientist with allthe necessary skills for a specific project may bepractically impossible to find. Some may hire businesspeople with the potential to be great analysts, andothers may hire analysts and develop their businessacumen along the way. Other companies employteams to marshal the required range of skills. Becauseanalytical skills are often in short supply,organizational structures and processes are critical forusing them effectively. Both organizing and hiringanalysts will have an impact on how the analyticsstrategy is deployed across the organization, and onrecruiting and retention approaches.TECHNOLOGYAs the technology for analytics rapidly evolves, anorganization’s ability to deploy and manage theunderlying infrastructure, tools and technologiesbecomes increasingly important. Technology wasstable for several decades in analytics, but is changingrapidly today. With the advent of big data, AI, cloudand open source options, creating an effectivetechnology strategy for analytics is a criticalprerequisite for success. Companies are alsoincreasingly expecting “citizen data scientists” to doself-service analytics and reporting. Self-service data5

DELTA Plus Model & Five Stages of Analytics Maturity: A Primerscience platforms can accelerate productivity anddeployment for all types of analytical professionalsand semiprofessionals.“Architectures must supportexperimentation and flexibilitywhile making it feasible tointegrate analytics withproduction systems andprocesses.”As analytics and AI become more critical to anorganization’s success, many will need to developsophisticated architectures for them. The architecturemust support experimentation and flexibility whilemaking it feasible to integrate analytics withproduction systems and processes. The relativeproportions of cloud versus on-premises, open sourceversus proprietary, structured versus unstructureddata capabilities, and other key decisions should bespecified in the architecture. Companies withrelatively specialized data and analytics AIenvironments — for deep learning-based imagerecognition issues, for example — may needspecialized hardware capabilities like graphicsprocessing units.platforms may automatically evaluate hundreds ofdifferent algorithms. “Ensemble” methods employmultiple techniques within a particular model. AImethods such as deep learning raise important issuesof the transparency and interpretability of models.And more traditional approaches to analytics such asreporting and visual analytics haven’t gone away.“Organizations wishing tosucceed with analytics need toassess what types of modelsthey are most likely to need andensure that the relevant tools andskills are available to use them.”Organizations wishing to succeed with analytics needto assess what types of models they are most likely toneed and ensure that the relevant tools and skills areavailable to use them. An informal “technique survey”can compare methods currently used within anorganization to what is available externally.Increasingly, the most sophisticated tools willdetermine which (among a large variety) of techniquesto employ in analyzing data, which may lessen theneed for an organization-specific approach totechnique selection.ANALYTICAL TECHNIQUESAt one time, most commercial organizations primarilyundertook simple regression analysis. Today,however, the plummeting costs of compute andstorage, coupled with widespread adoption of opensource development by Google, Amazon, Microsoftand others, have resulted in an explosion of analyticalmethods and techniques. Some machine learningiianalytics.comCopyright 2018 International Institute for Analytics6

DELTA Plus Model & Five Stages of Analytics Maturity: A PrimerFive Stages of Analytics MaturityOrganizations mature their analytical capabilities asthey develop in the seven areas of DELTA Plus. Thematurity model, described in Competing on Analyticsand developed in Analytics at Work, helps companiesmeasure their growth across the seven DELTAelements. This model enables an organization toassess which elements are strengths and which areweaknesses. For example, an organization mayachieve a stage 4 in analytics leadership maturity, butachieve only a stage 3 in its management and use ofdata. This assessment enables targeted investment tomature analytics weaknesses based on the DELTAModel.Stage 1: Analytically Impaired. These companies relyprimarily on gut feel to make decisions, and they haveno formal plans for becoming more analytical. Theyiianalytics.comCopyright 2018 International Institute for Analyticsaren’t asking analytics questions and/or they lack thedata to answer them. Their leaders may be unaware ofanalytics and what can be done with them.Stage 2: Localized Analytics. Analytics or reporting atthese companies exist within silos. There is no meansor structure for collaborating across organizationalunits or functions in the use of analytics. This oftenleads to “multiple versions of the truth” across acompany.Stage 3: Analytical Aspirations. These companies seethe value of analytics, and intend to improve theircapabilities for generating and using them. Thus far,however, they have made little progress in doing so.Stage 4: Analytical Companies. Companies in thiscategory are good at multiple aspects of analytics.They are highly data-oriented, have analytical toolsand make wide use of analytics with somecoordination across the organization. However, there7

DELTA Plus Model & Five Stages of Analytics Maturity: A Primerremains a lack of commitment to fully compete onanalytics or use them strategically.DELTA Plus TransitionsStage 5: Analytical Competitors. These companiesuse analytics strategically and pervasively across theentire enterprise. They view their analyticalcapabilities as a competitive weapon, and theyalready seen some competitive advantage result fromanalytics.Table 2 outlines the conditions that are typically inplace at each stage of progress in building an analyticsprogram.DELTA Plus Elements Across the FiveStages of MaturityThe factors described above drive the relative maturityand sophistication levels of an organization’sapproach to analytics. In the tables below aredescribed typical attributes of each DELTA element fora given maturity level, and the types of changes thattypically accompany a move from one maturity levelto the next.Organizations that desire to increase their maturitylevel can use this figure as a guideline for capabilitiesand improvements to pursue. Of course,circumstances will vary across companies, but thesedescriptions are illustrative of the most commonattributes and change types.Study this table with your current conditions andanalytical ambitions in mind. What do you need to do toleverage your strengths, shore up your weaknesses,become more DELTA Plus ready, and increase thebusiness impact and value of analytics? As you consideryour course of action, be sure to avoid these commonpitfalls: Focusing too much on one dimension of analyticalcapability (most often technology and data) at theexpense of others Devoting too much time, energy and money toanalytical initiatives that have low business impact(less valuable targets - even if that’s what thebusiness is asking for) Attempting to do too much at once.11Davenport, Harris, Morison, Analytics at Work: SmarterDecisions, Better Results, 2010 (Harvard Business SchoolPublishing), 185 – 188.iianalytics.comCopyright 2018 International Institute for Analytics8

DELTA Plus Model & Five Stages of Analytics Maturity: A PrimerTable 1Stage 1:Stage 2:Stage 3:Stage 4:Stage 5:Success ticalAspirationsAnalyticalCompaniesAnalytical CompetitorsDataInconsistent, poorquality andorganization; difficultto do substantialanalysis; no groupswith strong dataorientation; basicreporting tools anddescriptive analytics.Much data useable,but in functional orprocess silos; seniorexecutives don’tdiscuss datamanagement; BI andbasic analytics tools.Identifying key datadomains and creatingdata warehouses ordata lakes; expansioninto unstructuredNoSQL data.Integrated, accurate,common data incentral warehouse;data still mainly an ITmatter; little uniquedata: use ofunstructured NoSQLdata analysis.Relentless search fornew data and metrics;organization separatefrom IT overseesinformation; datamanaged as strategicasset.EnterpriseNo enterpriseperspective on data oranalytics. Poorlyintegrated systems.Islands of data,technology andexpertise deliver localvalue.Process or businessunit focus foranalytics.Infrastructure foranalytics beginning tocoalesce.Key data, technologyand analysts aremanaged from anenterpriseperspective.Key analyticalresources focused onenterprise prioritiesand differentiation.LeadershipLittle awareness of orinterest in analytics.Local leaders emerge,but have littleconnection.Senior leadersrecognizingimportance ofanalytical capabilities.Senior leadersdeveloping analyticalplans and buildinganalytical capabilities.Strong leadersbehaving analyticallyand showing passionfor analyticalcompetition.TargetsNo targeting ofopportunities.Multiple disconnectedtargets, typically notof strategicimportance.Analytical effortscoalescing behind asmall set of importanttargets.Analytics centered ona few key businessdomains with explicitand ambitiousoutcomes.Analytics integral tothe company’sdistinctive capabilityand strategy.AnalystsFew skills, and thoseare attached tospecific functions.Disconnected pocketsof analysts;unmanaged mix ofskills.Analysts recognized askey talent and focusedon important businessareas.Highly capableanalysts explicitlyrecruited, deployedand engaged.World-classprofessional analysts;cultivation ofanalytical amateursacross the enterprise.TechnologyDesktop technology,standard officepackages, poorlyintegratedtransactional systems.Individual analyticalinitiatives, statisticalpackages, descriptiveanalytics, databasequeryin

source development by Google, Amazon, Microsoft and others, have resulted in an explosion of analytical methods and techniques. Some machine learning platforms may automatically evaluate hundreds of different algorithms. “Ensemble” methods employ multiple techniques within a particular model. AI