Gartner-首席数据和分析官如何领导数据科学和机器学习领域的技能提升计划_第1页
Gartner-首席数据和分析官如何领导数据科学和机器学习领域的技能提升计划_第2页
Gartner-首席数据和分析官如何领导数据科学和机器学习领域的技能提升计划_第3页
Gartner-首席数据和分析官如何领导数据科学和机器学习领域的技能提升计划_第4页
Gartner-首席数据和分析官如何领导数据科学和机器学习领域的技能提升计划_第5页
已阅读5页,还剩25页未读 继续免费阅读

下载本文档

版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领

文档简介

GartnerResearch

HowCDAOsCanLead

UpskillingInitiatives

inDataScienceand

MachineLearning

PeterKrensky

15December2022

HowCDAOsCanLeadUpskillingInitiativesinDataScienceandMachineLearning

Published15December2022-IDG00780993-12minread

ByAnalyst(s):PeterKrensky

Initiatives:ChiefDataandAnalyticsOfficerLeadership

Asthehiringboomfordatasciencetalentcontinues,initiativestoupskillquantitativeprofessionalsremainjustasprominent.CDAOsshouldusethishigh-levelguidancetohelpdevelopin-housetalentandimprovedatascienceandmachinelearningliteracy.

Overview

KeyFindings

■Machinelearningliteracyremainslowinmanyorganizations;concertededucationandculturechangearenecessarybutdifficultduetothegapbetweendatascientists’technicalexpertiseandbusinessusers’domainknowledge.

■Abundanteducationalopportunitiescombinedwithtalentacquisitionandretentionchallengesmotivateorganizationstoupskilltheirdataprofessionalsatalllevelsofsophistication,especiallywithanaimtogrowtheircitizendatascientistpopulations.

■Anoverwhelmingnumberoftoolsandapproachesareavailabletoexpertandcitizendatascientists.CDAOsmustnavigateavastlandscapetomatchdiverseuserstoappropriatesolutionsandcorrespondingeducationalpaths.

Recommendations

CDAOsresponsibleforanalytics,BIanddatasciencesolutionsshould:

■Raiseoveralldatascienceandmachinelearningawareness,adoptionandliteracybyprovidingcentralizededucationalresourcesandshowcasingexistingusecasesandsuccessstories,bothinternalandexternal.

Gartner,Inc.|G00780993

Page1of12

■Identifycitizendatascience(CDS)candidatesintheirorganizationbycreatinganinventoryofin-houseskillsandambitions.MatchupskillingpathstothevariousbackgroundsandaspirationsofCDScandidates.Looktobuildinterconnectedcommunitiesofdatascientists,citizendatascientistsandotherMLpipelinestakeholders.

■Buildrepeatableandsustainableeducationprogramsbydesigningdifferentupskillingroadmapsforaverageconsumersofanalytics,CDScandidatesandexpertdatascientists.

StrategicPlanningAssumptions

By2024,75%oforganizationswillhaveestablishedacentralizeddataandanalytics(D&A)centerofexcellencetosupportfederatedD&Ainitiativesandprevententerprisefailure.

By2025,50%ofdatascientistactivitieswillbeautomatedbyartificialintelligence,easingtheacutetalentshortage.

Introduction

Thetalentgapindatasciencemayneverbefullyclosed—butitcanbenarrowed.ManyGartnerclientsstillreportdifficultyfindingandattractingtalent.Retainingproductivedatascientistsforlongtenuresisalsoamajorchallenge.CDAOsneedtolearnhowtobuilddevelopmentpathsforexpertsandsupportbuddingcitizendatascientistswiththerighttools,trainingandstructure(seeNote1forCDAOroledefinitionandNote2foradefinitionofcitizendatascientist).Evenorganizationsthatbuildhighvolumesofcomplexandaccuratemodelshavetodiligentlyfosterdataliteracyandproperadoptionofsolutions.Forexample,CDAOswhoinvestmuchmoreinresourcesandtalentare1.8xmoreeffectiveandsuccessfulwiththeirdataliteracyprograms.1Upskillingshouldbepromotedthroughouttheorganization,withtargetedtrainingforaselectgroupofindividuals—bothexpertsandnewCDScandidates—aswellasgeneraleducationforallconsumersofanalytics.

Thegreatestopportunityformostorganizationstogrowtheirtalentpoolfordatascienceandmachinelearningisthroughtheupskillingofcurrentstaff.

Gartner,Inc.|G00780993

Page2of12

Mostsuccessfulupskillingincorporatessomeformaleducationandtraining,butoftenthemostimpactfullearninghappens“onthejob”duringthecompletionofanewprojectortheassumptionofnewanalyticalduties.Manyself-identifieddatascientistshavefewerthanfiveyearsofexperienceworkingwithML.Leadersshouldexpectsomegrowingpainsandsteeplearningcurves.Timelinesfordevelopmentshouldbeflexiblewherepossible,andearlyprojectsshouldbelimitedinscopeandrisk.

Analysis

RaiseMachineLearningLiteracyAmongConsumersandPromote

CollaborationWithDataScientists

Raisingthelevelofdiscoursearounddatascienceandmachinelearningisthefirststeptowardupskillingyourworkforce.Beginbyensuringthatallline-of-business(LOB)leadersanddecisionmakershaveaclearunderstandingofhowdatascientistscreatevalue.Thiscanbedonethroughsimpleworkshopsorexerciseswheredatascientistsand/orotherparticipantsbegintoquestionsomeexistingKPIs,dataormetrics.Aspiringmodelbuildersandheavyconsumersshouldhaveafoundationalunderstandingofthemachinelearninglifecycle—particularlydatapreparation,featureengineering,testingandtraining,anddeployment(seeFigure1).Emphasizetoconsumersthatanalyticsconsumersarethekeytogeneratingvaluefromtheworkofdatascienceteamsandgivingfeedbackforthefutureprojectsandmodeliterationsthatwillgothroughthiscycle.

Gartner,Inc.|G00780993

Page3of12

Figure1.TheMachineLearningLifeCycle

Helpenthusiasticindividualsbecomefamiliarwiththebasicsofseveralmachinelearningtechniques,suchasregression,clusteringandclassification.Encourageorrequiredatascientiststoregularlyholdopensessionstodiscussacurrentproject(inlayperson’sterms)orintroduceanaspectofdatasciencetheyarepassionateabout.Considergamificationofdevelopmentthatencouragesupskillingindividualstoattendregulartraining,engagenewsubjectsorenterintohealthycompetitionwithpeers.

CreateaSkillsInventorytoIdentifyIn-HouseCDSCandidatesandFosterInterconnectedDataScienceCommunities

Talentalignment,careerdevelopmentandretainingtalentarethemajorleadershipdemandsneededforsustainingsuccessfulupskillinginitiatives.TheroleofthecitizendatascientistpresentsuniquedemandsforCDAOsintermsoftalentrecognitionanddevelopment.CDAOsneedtounderstandwhattheCDSpersonais(seeFigure2)andrecognizetheskillsandprofilesthatmakeforgoodCDScandidates.

Gartner,Inc.|G00780993

Page4of12

Figure2.BreakingDowntheCDSPersona

Talktopotentialcandidatestogaugetheirinterestandaspirationsaroundcareersindatascience.Getcandidatestocompleteself-evaluationsoftheirbackgroundssothatadetailedinventoryofskillscanbeestablished.Thisinventorywillbeinvaluablefordesigningtrainingprogramsandmakingtechnologyinvestments.Themostpromisingcandidatesforupskillingoftenhaveeducationalandprofessionalbackgroundsinphysics,chemistry,biology,actuarialscience,computerscience,engineering,finance,economicsandmathematics.

UsethatdatafromtheinventorytoidentifycommunitieswithinITandbusinessunitsandacrossbusinessfunctions.Supportandgrowthesecommunitiesviaformalmentorship,structuredcollaborationandstrategicsoftwareinvestments.Lookforleadersandpractitionersthatwillkeepthesecommunitiesactiveandtalkingtoeachother.Formoreonspreadingupskillinginitiativesthroughcommunities,seeUseSocialInfluencerPrinciplestoBuildCommunitiesandIncreaseDataandAnalyticsAdoption.

Gartner,Inc.|G00780993

Page5of12

DesignUpskillingRoadmapsforExpertCDSCandidatesandAnalyticsConsumersandSupporttheContinuousDevelopmentofExperts

Conductfoundationalupskillinginitiativesarounddatascienceandmachinelearningwithintwogroupssimultaneously:CDScandidatesandanalyticsconsumers(seeFigure3).

Figure3.SimultaneousUpskillingRoadmap

ThePathforCDSCandidates

TheCDScandidategroupwillembarkonastructuredupskillingprogramacrossthreestages.

Stage1:ApproachSelectionandFormalTraining

Regardlessoftheselectedapproach,allCDScandidatesshouldsetoutwithaclearvisionoftheskillstheywillneedtoacquireandhone.CDAOsshouldtasktheirteamstoworkwithHRtodesigncareerpathsforcitizendatascientists,includingdifferentspecializations.

Gartner,Inc.|G00780993

Page6of12

Formaltrainingcanbeacourse,bootcamportooltrainingprogram.CDAOsandCDScandidatesinterestedinpursuingupskillingthroughfreestudyshouldexplorethemassiveopenonlinecourses(MOOCs)offeredbyorganizationssuchasCoursera,DataCamp,edX,UdacityandUdemy.CDAOsshouldensurethatthenumeroustutorialsavailableonYouTube,KaggleLearnandelsewhereareinvestigated.LeadersandCDScandidatesshouldinvestigatelocalin-personoptionsaswell.Whenoverseeingafreestudyprogram,developclearincentivesandmilestonesforCDScandidatesinordertoaccuratelymeasuretheirprogress.Classroomlearningwillserveasafoundationforextensiveproject-basedlearning.

Stage2:ExperimentationandPrototyping

OnceCDScandidatescompletetheirformaltraining,theyshouldbeginexperimentingwiththeirnewskillsanddesigningprototypes.LeadersshouldprovidesandboxenvironmentsfornewMLpractitionerswhereuserscanworkwithdesensitizeddatafortrialanderror.Wheneverpossible,placecitizendatascientistsunderthementorshipofdatascientistswhocanreviewtheirworkandprovidefeedback.LeadersshouldensureCDScandidatesalso:

■Beginhoningthecommunicationskillsnecessaryforsuccessfuldatascienceandevangelizethemethodologyandpotentialoftheirworkwithmachinelearning.

■Establishregularcontactandcollaborationwithanalyticsconsumersandexpertdatascientists.

■Examineanalyticsconsumers’suggestionsfornewmodels.

■Sharetheirdomainexpertisewiththeirdatascienceteammates.

■Initiateareverseknowledgetransfertofillgapsamongeventhemosttenureddatascientists.

Stage3:DeliveryandOperationalization

ThethirdstageofCDSupskillingisthedeliveryandoperationalizationofnewmodels.TheemergingrolesofchiefdatascientistandMLengineershouldtakeanactiveroleinshepherdingcitizendatascienceprojectsfromexperimentationtoproduction.Citizendatascientistsshouldalsopromoteutilizationamonganalyticsconsumersandworktoestablishcontinuousfeedbacktoimproveexistingmodelsandexchangeideasfornewones.

Gartner,Inc.|G00780993

Page7of12

Gartnerhaspublishedresearchthatfurtherexploresbestpracticesformanagingcitizendatascientistsandcommonpitfallstoavoid(seeLessonsFromDataScientistsonTheirEducationandCareerDevelopment).

ThePathforAnalyticsConsumers

Thesecondgrouprequiringsimultaneousstructuredupskillingisconsumersofanalyticsthroughouttheorganization—againacrossthreestages.

Stage1:IntroductiontoDataScienceandMachineLearning

Begintheupskillingroadmapfortheaverageconsumerwithintroductory-leveleducationondatascienceandmachinelearning.Thisshouldcomeintheformofanexternalguestspeakerordedicatedtimewithaninternaldatascienceprofessional.Showconsumersinternalcasestudieswithmeasurablebusinessvalueandrecruitthemasenthusiasticstakeholdersinyourentiredatascienceinitiative.CDAOsshouldviewthisphaseasakeyopportunitytoestablishnewworkingrelationshipsbetweenthedatascienceteamanddifferentLOBs.

Stage2:BrainstormIdeasonMetricsandAnalytics

Followupthisintroductorytrainingwithbrainstormingsessionsonmetricsthatanalyticsconsumerswouldliketobetterpredictandoptimize.Thisinturnshouldleadtoproposalsforinvestmentsinpackagedapplicationswiththehighesteaseofimplementationanduse.

Atthisstage,putanalyticsconsumersincontactwithCDScandidatestosharetheirideasandhearwhatmaybepossibleuponthecompletionoftheupskillinginitiative.Analyticsconsumerscanhelpbrainstormideas,andthetwogroupscanthencollaborateonfeasibilityandprioritization.

Stage3:TesttheFunctionalityoftheModels

Finally,analyticsconsumersshouldreceiveandutilizethefirstroundofmodelscreatedbycitizendatascientists.CDAOsshouldtasktheirteammemberstomonitortheconsumptionofnewmodelstoensureactiveparticipationattheLOBlevel.Analyticsconsumersshouldofferfeedbacktodatascienceteamsonthestrengthsandshortcomingsofnewmodelsandparticipateinbrainstormingforthenextroundofmodeldevelopment.

Gartner,Inc.|G00780993

Page8of12

DriveUpskillingofExpertDataScientists

Successfuldatasciencerequiresconstantlearningatalllevels—especiallyamongexperts.Encourageandsupporttheircontinuedknowledgedevelopmentinallareasofthefast-movingAIspace.Createspaceindatascienceprofessionals’schedulesforindependentlearning.

Commonareasofstudyanddevelopmentforexpertdatascientistsinclude:

■Domainknowledge

■Communicationskills

■Strategicandexecutiveskills

■Deeplearning

■Reinforcementlearning

■Augmenteddatascienceandmachinelearning

■InternetofThingsandedgecomputing

■Computervision

■Naturallanguageprocessing

■Machinelearningoperations(MLOps)

■AItrust,fairnessandexplainability

■Digitalethics

InadditiontohardskillsandhottopicsinAI,expertdatascientistsespeciallybenefitfromleadershipandhumanskillsdevelopment.Datascientists,particularlyjuniordatascientists,tendtohavelessstrengthandexperienceinareassuchaspresentationsskills,mentorship,collaborationandprojectprioritization.Formoreonthistopic,seeAnExecutiveLeader’sGuidetoStaffingEffectiveDataScienceTeams.

Gartner,Inc.|G00780993

Page9of12

CDAOswhohavetenureddatascientistsshouldcollaboratewiththeseindividualstounderstandtheirambitions,thendesignappropriatecareerpaths.Doyourdatascientistswanttobepurepractitionersordevelopintomorehybridroles?Seniororchiefdatascientistscanstepintoaroleprimarilyfocusedonmanagementandarenolongerday-to-daypractitioners,whichmaynotappealtomanyexpertdatascientists.ReservesuchaleadershiproleformaturedatascienceoperationsthathavemultiplejuniordatascientistsandanestablishedCDSinitiative.Formoreonthistopic,seeRolesandSkillstoSupportAdvancedAnalyticsandAIInitiatives.

Evidence

GartnerChiefDataOfficerAgendaSurveyfor2022:ThisstudywasconductedtoexploreandtrackthebusinessimpactoftheCDOroleand/ortheofficeoftheCDOandthebestpracticestocreateadata-drivenorganization.TheresearchwasconductedonlinefromSeptemberthroughNovember2021among496respondentsfromacrosstheworld.Respondentswererequiredtobethehighestleveldataandanalyticsleaderintheorganization:chiefdataofficer,chiefanalyticsofficer,themostseniorleaderinITwithdataandanalyticsresponsibilities,orabusinessexecutivesuchaschiefdigitalofficerorotherbusinessexecutivewithdataandanalyticsresponsibilities.Thesurveysamplewasgleanedfromavarietyofsources(includingLinkedIn),withthegreatestnumbercomingfromaGartner-curatedlistofmorethan4,519CDOsandotherhigh-leveldataandanalyticsleaders.ThestudywasdevelopedcollaborativelybyGartnerD&AanalystsandthePrimaryResearchTeam(SeeCDAOAgenda2022:FocusonValue,TalentandCulturetoPullAhead).

Disclaimer:Resultsofthisstudydonotrepresentglobalfindingsorthemarketasawholebutreflectsentimentoftherespondentsandcompaniessurveyed.

Note1:CDAORoleDefinition

Chiefdataandanalyticsofficer(CDAO)referstothebusinessleadershiprolethathastheprimaryenterpriseaccountabilityforvaluecreationbymeansoftheorganization’sdataandanalyticsassets,andthedataandanalyticsecosystem.Equivalenttitlesforthisrolearechiefdataofficer,chiefanalyticsofficer(iftheCDAOroleorequivalentisnotintheenterprise),chief/headofdataandanalyticsandothervariations.

Gartner,Inc.|G00780993

Page10of12

Note2:CitizenDataScientist

Acitizendatascientistisapersonwhocreatesorgeneratesmodelsthatusepredictiveorprescriptiveanalytics,butwhoseprimaryjobfunctionisoutsidethefieldofstatisticsandanalytics.Thepersonisnottypicallyamemberofananalyticsteam(forexample,ananalyticscenterofexcellence)anddoesnotnecessarilyhaveajobdescriptionthatlistsanalyticsashisorherprimaryrole.ThispersonistypicallyinalineofbusinessoutsideITandoutsideabusinessintelligence(BI)team.However,anITorBIprofessionalmaybeacitizendatascientistiftheirworkonanalyticsisonlyasecondaryrole.Citizendatascientistsare“powerusers”whoareabletousesimpleandmoderatelysophisticatedanalyticsapplicationsthatwouldpreviouslyhaverequiredmoreexpertise.

RecommendedbytheAuthor

SomedocumentsmaynotbeavailableaspartofyourcurrentGartnersubscription.

LessonsFromDataScientistsonTheirEducationandCareerDevelopment

HypeCycleforDataScienceandMachineLearning,2022

3StepstoBuildandOptimizeaPortfolioofAnalytics,DataScienceandMachine

LearningTools

TopTrendsinDataandAnalytics,2022

Tool:DataLiteracyPersonastoDriveaData-DrivenCulture

RolesandSkillstoSupportAdvancedAnalyticsandAIInitiatives

Gartner,Inc.|G00780993

Page11of12

©2023Gartner,Inc.and/oritsaffiliates.Allrightsreserved.GartnerisaregisteredtrademarkofGartner,Inc.anditsaffiliates.ThispublicationmaynotbereproducedordistributedinanyformwithoutGartner'spriorwrittenpermission.ItconsistsoftheopinionsofGartner'sresearchorganization,whichshouldnotbeconstruedasstatementsoffact.Whiletheinformationcontainedinthispublicationhasbeenobtainedfromsourcesbelievedtobereliable,Gartnerdisclaimsallwarrantiesastotheaccuracy,completenessoradequacyofsuchinformation.AlthoughGartnerresearchmayaddresslegalandfinancialissues,Gartnerdoesnotprovidelegalorinvestmentadvice

anditsresearchshouldnotbeconstruedorusedassuch.Youraccessanduseofthispublicationaregovernedby

Gartner’sUsagePolicy

.Gartnerpridesitselfonitsreputationforindependenceandobjectivity.Itsresearchisproducedindependentlybyitsresearchorganizationwithoutinputor

influencefromanythirdparty.Forfurtherinformation,see"GuidingPrinciplesonIndependenceand

Objectivity."

Gartner,Inc.|G00780993

Page12of12

Webinar

GartnerPanel:TheRelationship

BetweenData&Analytics

TechnologyandTalent

Explorecomprehensiveapproachesdata&

analyticsleadersneedtoattractandretain

talentfortheirteams.

WatchNow

eBook

WhatAretheEssentialRolesfor

DataandAnalytics?

Learnhowtoequipyourbusinesswiththe

rightcompetencies.

DownloadNow

Actionable,objectiveinsight

ExploretheseadditionalcomplimentaryresourcesandtoolsforData&Analyticsleaders:

HYPERLINK"/en/webinar/453

温馨提示

  • 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
  • 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
  • 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
  • 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
  • 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
  • 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
  • 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。

评论

0/150

提交评论