[PDF] What Makes a Great Manager of Software Engineers? - Microsoft PDF kalliamvakou-tse-2018.pdf

As software development today is done in teams, managers are essential to organize the effort of creating good software and manage the people that carry it out

[PDF] [PDF] Become An Effective Software Engineering Manager

Ebook Title : Become An Effective Software Engineering Manager - Read Become An Effective Software Engineering Manager PDF on your Android, iPhone,

Managing Software Engineers and their Knowledge - CORE

successful use in Information Technology-related companies, relatively few are specifically for software engineering Most reported systems in these companies

[PDF] What Every Software Manager Must Know To Succeed With Object

management Lead and manage successful software development projects 97 Things Every Engineering Manager Should Know-Camille Fournier

[PDF] Software Engineering For Technical Managers - Caribbean

10 mar 2021 · software-engineering-for-technical-managers 3/14 Downloaded from cep unep on March 10, 2021 by guest The Successful Software

[PDF] effectiveness of fitness apps

[PDF] effects of cosmetics on human body

[PDF] effects of cosmetics on human skin

[PDF] effects of exercise on the body

[PDF] effects of exercise on the muscular system

[PDF] effects of exercise on various systems

[PDF] effects of exercise on various systems of human body

[PDF] effects of free trade

[PDF] effects of poison

[PDF] effects of skipping breakfast to students

[PDF] effects of the french revolution

[PDF] effects of the thirty years

[PDF] effects of the thirty years war on germany

[PDF] effects of the thirty years war quizlet

[PDF] effects of urbanization

1

What MakesaGreat Manager

of SoftwareEngineers? EiriniKalliamv akou,Student member,IEEE, ChristianBird,Member,IEEE , Thomas Zimmermann,Member,IEEE , AndrewBegel,Member,IEEE , RobertDeLine, Member,IEEE ,

Daniel M.Ger manF

Abstract

-Havingg reatmanagersisas critical tosuccess asha ving a goodteam ororganization. Ingener al,a great managerisseen as fuellingth eteamthey manage, enablingitto useitsfullpotential. Though softwareengineering researchstudiesfactors thatma yaff ect the performanceandproductivityof software engineersand teams(lik e tools andskill), ithas ov erlook edthesoftwareengineeringmanager.The softwareindustr y"sgrowthand changeinthelastdecadesis creating a needf oradomain-specificvie wof management.On theonehand, expertsarequestioning how theab undantwork inmanagement applies to softwareengineering. Ontheotherhand, practitioners arelooking to researchers forevidence-based guidanceonho wto managesoftware teams.W econductedamix edmethods empirical studyofsoftware engineeringmanagement atM icrosoftto investigate whatmanager at- tributesdev elopersandengineeringmanagersperceiv eimpor tantand why.We presentaconceptual frame wor kofmanagerattr ibutes,and find thattech nicalskillsarenot thesign ofg reatnessf oran engineering manager.Through statisticalanalysis we identify howengineersand managers relatein theirvie ws, andhowsoftwareengineer ingdiff ers from otherkno wledgework groupsinitsperceptionsabout whatmakes greatmanagers .Wepresent strategiesfor puttingthe attributesto use, discuss implicationsf orresearchandpr actice, andoff eravenuesf or furtherw ork.

IndexT erms

-softwareengineer ingmanagement;empirical studies; softwarecompanies

1 INTRODUCTION

Case studiesfr omdiverseindustriesshow thatgr eatman- agers makea signicantdif ference intheperformanceof teams andor ganizations[1],[2],[3]. Conversely, thewr ong person ina managerr olehas detrimentaleffectson em- ployee engagement,pr oductivity,andthequalityof pro- duced results[4].As software developmenttoday isdone in teams,managers are essentialtoorganize theef fortof creatinggood software andmanagethepeople thatcarry it out. The manager'srole ismultifaceted.Oneof theirr espon- sibilities isto delivera product thatmakes theorganization successful. Thisis generallycaptur edby variousmetricsof productivity,performance, protabilityetc.The manageris also responsibleforcr eatingconditions whereemployees feel motivatedand productive. Successhereis captured in the employees'per ceptions,whichstudiesshow determine behaviour andimpact organizational outcomes[5],[6],[7].

Thus understandingwhat impactsengineers' perceptions oftheir managersis ofhigh importance.Unfortunately ,we stilldon't knowwhat tolook forin agr eatsoftwar eengineering manager,and howto furtherdevelop theirskills tosupport the teamsthey manage.

As thesof twareindustryundergoestr emendouschange

every year,resear chersmustcontinuallyrethinkthe fac- tors thataf fectthetraditional conceptof productivity 1. In this vein,our resear chgoalistounderstand howsoftwar e engineering managersfunction andwhat isper ceivedto make them great . Greatmanagerspositively impactmotivation and engagement[8];we aimto raiseawar enessof these aspects, asthey canaf fectsoftwar eengineeringoutcomes, even ifin asecond-or dermanner .Welook forattributes that areperceived tocharacterizegreatsof twareengineering managers,howandwhythese attributesar eimportant,and how theyar eusedspecicallyin thisdomain.

The studywe report inthispaperused amixed meth-

ods approach.We conducted37semi-structur edinterviews with engineersand managersof varieddemographics at Microsoft.W ethenusedtheir inputto create anddeploy a surveyto 3,646engineers andmanagers, usinga question- nairegr oundedoncontextualizedinformation. We found that theengineering managerguides engineersto make decisions, motivatesthem, andmediates theirp resence in the organization.To thatend,asuf cientlevel oftechnical knowledge isnecessary butpeople managementskills are critical forgr eatsoftwareengineering managers.Comparing the perceptionsofmanage rsto engineersinouranalysis, we foundgeneral alignmentbut alsoidentied specic differencesthatcan helptailor managementappr oaches. Our resultshavenovelty forsoftwar eengineer ing,but also linktoo rganizational psychologyandbehaviour,and apply toother knowledge workdomains.Through asepa- rate survey,wer eviewedhow theperceptionsin software engineering relatetothose inother knowledgeworker groupswithin Microsoft. Identifyingthesimilaritiesand dif- ferencesbetween domains'per ceptionscan helpusunder- stand whatconditions are likelytomakemanager practices effective.

In thispaper:

1. RethinkingPr oductivityinSoftware Engineering:https://www .

dagstuhl.de/en/program/calendar/semhp/?semnr=17102This is the author's version of an article that has been published in th

is journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TSE.2017.2768368 Copyright (c) 2017 IEEE. Personal use is permitted. For any other purp oses, permission must be obtained from the IEEE by emailing pubs-permiss ions@ieee.org. 2 we contributea conceptual frameworkof fifteenat- tributes thatcharacterize great engineeringmanagers we offercontextual examplesof howthese attributes areput intoaction, anddiscuss therole oftechnical knowledgefor managersto begr eat we providequantitative evidenceabout howthe attributes rankin perceived importance,whatdemo- graphic differencesexist,andhow thefindings from softwareengineering compare tootherknowledge work domains.

Our studyhas implicationsfor bothpractic eand re-

search.Our conceptualframework canbe usedby new and existingengineering managers,or thosein training, to highlight whichattributes theyshould focuson toimpr ove. The identifiedattributes alsofuel furtherwork, tomeasur e their impacton organizational orengineeringoutcomes.

2 RELATEDWORK

In thispaper ,wesetout toexplor ehow software engi- neering managementworks inpractice ata large software company,and identifythe particularaspects whichar e relevanttoday .Ourworkdraws onmultiple perspectives, and wer elateourfindingsto otherknowledge workgr oups.

Many theorieshavebeen developedar oundhow to

manage organizationsingeneral. Originally, thesetheories focused onhow workersperform tasks.In the1920s, works by Taylor[9],the Gilbreths (describedin [10])andGantt[11] formed theclassic eraof management,when studiesof how tospeed uppr oductiono fferedadvice tomanagers on howto organize thetasksandenvir onmentfor factory workers tobe efficient. Betweenthe1920s-1950s,the fieldof management turnedits attentionto howworkers thought and feltabout theirwork (see[12] fora detailedoverview) aiming tomotivate employeesto identifywith organiza- tional goals[13], [14],and improve performance[15].

Twomilestones havebeen keyto shiftingthe focusof

management onmanaging people.The firstone wasthe introductionof psychology andsociologytheoriesin man- agement, researchingfactorsthatimpact humanneeds [16] to betterunderstandworkers" motivation[17] andengage- ment [18].Especially afterthe 1960s-considered themod- ern eraof management-thefocus ison employeework attitudes andmotivation [19]and ther ecognitionof people management asseparate from workorganizationo rpr oject management [20].The secondmilestone wasthe riseand increasedmobility ofknowledge workers,which turned attention towardsthebehavioural aspectsof employees[21].

Peter Drucker[22]led managementthinkers tosee the

corporation asa socialinstitution andworkers asassets, challenging existingmanagement principlesas fitonly for manual labour.

After decadesof resear chonmanagementingeneral,

therear ealarge numberof theoriesthatdescribemanagers and theirbehaviors inthe organization. However, ifone wants toapply thesetheories ina particulardomain such as softwareengineering,the factorsi ntheir modelsmust be reducedtopractice. Ar ecentr eviewof thisliterature was doneby Lenberg et al., describingthe intersectionof organizationalpsychology with"behavioral software engi-

neering" [23].They found23 relevant paperson leadershipof softwareteamssince 1997,and concludedthat stillmor ehuman-oriented studiesare neededinsoftware engineering.

Although softwareengineeringmanagement isoften

equated topr ojectmanagement[24],[25], [26],some books about softwareengineeringpr ojectmanagement alsomen- tionpeoplemanagement; forexample, Listerand DeMarco"s Peopleware[27] discussedsuch issues insoftwar eprojects early on.Beecham etal. conducteda systematicliteratur e reviewof motivationin software engineeringand reported studies thatshow astr ongimpact ofmanagersandtheir practices onengineers" motivation[28]. Otherbooks fo- cus onadvising developerson how toact inamore col- laborative orsocially-awar emanner[29];or takean en- trepreneurialviewon management[30]; orar eanecdotal and basedentir elyonthe(admittedly extensive)experience of theauthor[31]. Storiesof badmanagers are widespread, often showinghow theirbehavior contributesto product failures[32].

In themajority though,literatur eon softwareengi-

neering managementfocuses onpr escribingformalized ap- proaches(e.g theSpiral andW aterfallmodels) oralternative approaches(e.g Agileand Lean)to scheduling,planning, and deliveringsoftwar eproductson timeandbudget. The SoftwareEngineering Bodyof Knowledge[33] brieflyad- dressesgr oupdynamicsandteamwork, butoverlooks the management ofteams andtheir members. Looking tothe popularpr essfor inspiration,someau- thors havea cynicalview ofmanagement theoriesand prescriptions[34] whilesome offer anecdotalevidence and advice formanagement behaviourthat theybelieve ledto their success[35]. Otherexperts focuson ther elevance of managementprincipl esindomainsthat undergo rapid growthand change[36] (suchas inthe technologyand softwareindustry [37]).

Thereseems tobe aneed forstudies tounderstand

people managementin software engineeringandhowman- agement principlesapply orr elateto thesoftwareengineer - ing domain.The studywe present inthis paperaimsto addressthis. Whilewe acknowledgeand drawon resear ch on generalmanagement, weset outto explore howengi- neering managementworks inpractice andwhich aspects arer elevanttoday,withoutpr esupposing.Wehave also drawn onmultiple perspectives,and related ourfindings to otherknowledge workgr oups.

Twostudies havediscussed aspectsof management,

specifically insoftwar eengineeringorganizations.

Our studyshar essomesimilarityof purposeand find-

ings witha studyfr omLi etal.[38],which investigated softwareengineering expertise. Thestudyidentified53 at- tributes ofgr eatsoftwareengineers. Someoftheattributes that werefoundimportant forengineers were recognized as potentiallyinspir edorfacilitatedby themanager; for example,creatingshar edsuccessfor everyoneon theteam, or creatinga safehaven whereengineers couldmake mistakes without repercussions.Ourstudyindependently identified these asimportant attributesfor engineeringmanagers too, and alsouncover edcomplementaryonesand strategiesto enact them. Recently,r esearchersatGoogle(asoftwar eengineering company ofcomparable sizeto Microsoft) investigatedthe

question, "Domanagers matter?"[39], [40]and found8 be-This is the author's version of an article that has been published in th

is journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TSE.2017.2768368 Copyright (c) 2017 IEEE. Personal use is permitted. For any other purp oses, permission must be obtained from the IEEE by emailing pubs-permiss ions@ieee.org. 3 haviors forgr eatmanagersintheir organization. Ina follow- up studyof whatmakes effective teams,Google founditis important thatteam membersfeel psychologicallysafe [41], and haveclarity abouttheir purposeand goals[42]; these areaspects thatthe managercan influencein apositive way. Wer eferbacktothese studiesin Section9, anddiscuss how they relatetothe findingswe present inthis paper.

Virtualteams oftenpr ovidea contextinwhichman-

agement proficienciesanddeficiencies canhave astr ong impact onthe team.Saxena andBurmann [43]looked at the specialneeds ofvirtual andglobally distributedteams, especially focusingon task-related andculturally-relatedat- tributes thataf fectteamperformance.Managers mustfacil- itate communicationand effective interactionsbetweenfar- flung teammembers andempower themto makedecisions independently (dueto timezone differ ences).Kayworth and Leidneralso focuson virtualteams, identifying factors such asmentoring andempathy ,which helpmakemanagers moreef fectiveleaders[44]. Zhanget al.identify howman- agers evolvefr omcontrollingvirtual teams,tocoordinating work amongteam members.Becoming anef fectivedelega- tor,even ofmanagement functionsand decision-making,is a keyfactor tomaking virtualteams successful.[45].

3 METHODOLOGY

Our researchmethodologycomprisedtwo highlevel

phases. Inthe first,exploratory ,phase, weinterviewed37 softwareengineers andengineering managersto identify perceivedimportant attributesof great software engineering managers. Inthe second,confirmatory ,phase, wedeveloped and deployeda surveyto alar gerpopulation.

3.1 Interviews

Weused interviewsto identifythe importantattributes that make agr eatsoftwareengineering manager,aswell as understand whysuch attributesar eseen ascriticalandhow they manifestin software engineeringcontexts. Participant selection.Wepurposely soughtto interview a diversegroup tocaptureas manyvarying opinionsand experiences aspossible. To thatend,weused astratified purposeful samplingappr oach[46] tor ecruitinterviewees. This selectionstrategy isa formof Maximum Variation Sampling[46] andis appropriate when"thegoalis notto build arandom andgeneralizable sample,but ratherto try to representarangeof experiencesr elatedto whatone is studying." Tocapture multipleperspectives,weinterviewed softwareengineers (those beingmanaged), andmanagers at multiple levels. Softwareengineering manager(or simply,engineering man- ager ) isthe nameof aparticular role atMicr osoft. According to thejob description,these managersar er esponsiblefor delivering resultsthrough oneormore teamsof engineers; they assistthe teamwith goalsetting, handlehiring deci- sions, manager esourcesfortheteam(s),and are responsible for guidingthe engineers"pr ofessionaldevelopment and reviewingtheir performance.As partof communicating business directiontotheir team(s),engineer ingmanagers liaise withother teamsand meetwith uppermanagement.

Beforemajor releases, engineeringmanagersrepr esenttheir team incr oss-teamdiscussionsaboutpr ojectstatus, anddecisions onthe features thatshiptocustomers.

Although ourstudy focuseson theenginee ringmanager role,we elicitedthe perspectivesof managersat multiple levels. Theseincluded team leads(often owninga feature with asmall numberof engineersr eportingto them), engineering managers , andupper levelmanagers (those that hire,advise, andr eviewengineering managers).Sincewe found thatr esponseswerein alignmentacross thedif ferent management roles,wemake nodistinction inthe remain- der ofthe paper;we simplydivide thoseinterviewed and surveyed intoengineers andengineering managers. For bothengineers andmanagers, weselected partici- pants alongthe dimensionsof experience(new tothe role- hiredin thelast 6months-or longtime intheir current role-longerthan 5years), number ofemployers (has their entirecar eerbeenatMic rosoft orhave theyworkedelse- where),gender,organizationallevel (engineer,team lead,en- gineering manager,andupper levelmanager), andproduct group(e.g., Windows,Office, Azure).

Wesent recr uitmentemailstoarandomsample ranging

between 10and 50people, dependingon thesize ofthe stratum. Forthose thataccepted (37persons), wesent a follow upemail askingthem toselect andrank thetop five mostimportant attributesfr oma listof16manager attributes (wer efertothoseas seedattributes inthe rest of the paper).T able1showsthe role andexperience ofthe 37 interviewees. Inthe parentheses weprovidethe numberof participants thatwe sentinvitations tofr omthat stratum.

Weasked onlyfor thetop fiveattributes knowingthat

due tothe cognitiveload ofrankings individualsusually pay moreattentiont othe topfewchoicesrather thancar e- fully rankingall alternatives,r esultingin additionalnoise in thelower rankings[47]. Theonline surveytool weused allowed theparticipants todrag attributesas separateitems and placethem inthe order thatr epresentedtheirranking.

The listof seedattributes wascompiled basedon two

sources.First, weused the11 attributesused inthe an- nual companypoll where Microsoftemployeesevaluate their manager;most ofthe Microsoft Pollattributes canbe traced backto managementliteratur e.The secondsourceof seed attributeswas ourr eviewof additionalmanagement literature[48], [49],[22], [50],[39] (5attributes );we added attributes foundin theliteratur ethat werenotalr eadyin- cluded inthe companypoll orvery similarto those.W ealso providedspace forthe participantsto addother attributes that theyfelt were important.

Interview protocol.Weasked allparticipants-

regardlessoftheir level-tor eferto andtalk aboutthe engineering managerrole.W easkedintervieweesbasic de- mographic questions;the informationcollected wasthe participants" numberof yearsof professional experience, the numberof differ entcompaniestheyworkedin,and their currentrole atMicrosoft.Next, wehad anin-depth discussion ofthr eeoftheattributes weselected from their top fiveseed andwrite-in attributes.W edetermined that a discussionof three attributescouldpragmaticallyfit in the timewe hadwith theparticip antsto bothcollect allthe information weneeded, andnot rush theiranswers. This was confirmedas afitting strategyafter thefirst fewinter -

views. Theattributes were selectedfordiscussion basedon This is the author"s version of an article that has been published in th

Engineer

P1, P2,P3, P4,P5 P6, P7,P9, P18,P20,

P23, P26

(out of50) (outof 40)TeamLead

P27, P30P28, P31,P32

(out of10) (outof 10)Manager

P11, P12,P14, P17,P19,

P21, P22,P25 P8, P10,P13, P15,P16,

P24 (out of40) (outof 40)Upper Manager

P29, P33,P34, P35,P36,

P37 (out of20) the rankinggiven bythe participant;we chosethe highest ranking attributes.When intervieweeshad provided write- in attributesthatthey feltwer emor eimportant, wegave these priorityin ourdiscussion. Gradually, aswe achieved saturation regardingsomeofthe attributes,we intentionally picked fordiscussion attributesthat wehad lessinformation about aslong asthe participantshad highlighted themas important (i.e.wer einthetop five). Weasked whythey felteach attributewas importantfor greatmanagers tohave. We alsoasked aboutstrategiesto gain orutilize theattribute (formanagers) toensur eour understanding ofthe nature oftheattributes,and toof fer actionable insightsfr omourstudy. We intentionallyused the abstractterm "great" withoutprovidinga definition so as notto bias intervieweesandinsteadgain anunderstand- ing ofwhat itmeant tothem. We accomplishedthis by employing a"W arStory"elicitationpr ocedure toexplor e concreteexperiences from theintervieweerelated tothe attribute [51].W eexplainedthatwe were interested inexpe- riences theyhad anytime duringtheir developmentcar eer, not limitedto Microsoft, andalsoaskedthem notto use names orindicat ewhethervariousexperiences, thoughts, etc. referredtotheircurr entor priorteams ormanagers.

Interviews lastedfr om30minutesto anhour andwer e

recordedwiththe interviewee"spermission.

Analysis.The interviewswer etranscribed;wethen

identified allattributes brought upduringtheinterviews and performedan opencar dsort toidentifycategoriesand organizedthem intothemes [52],[53]. Eachcar dr epresented an attributethat wasdescribed inan interview, eitherseed ones orthose thatemer gedfr omtheparticipants;we sorted

83 cardsinto15 categories.The card sortingwas performed

by twoof theauthors (oneof themwas aMicr osoftem- ployee), intwo rounds. Foreachcategory, weexamined the context forevery card inthatcategoryand cameup with a nameand ashort listof examplesfor theattribute. The interview transcriptionswer ethencodedaccor dingto these categories.

3.2 Survey

Wedesi gnedasurvey basedon interviewresultsto validate, and seehow theattributes generalizeto abr oaderpopula- tion.

Survey instrument.The survey"sprimary purposewas

toassess theimportance ofeach ofthe identiedattributes from the interviewsand determineif there wereadditionalones

to add.W easkedrespondents torate eachattributeofengi- neering managerson aten pointscale from "notimportant" to "critical".The displayedor derof theattributeswasran- dom foreach respondent .Weprovidedthe nameofeachat- tribute witha shortdescription ofexamples demonstrating it inpractice (seeT able2 forthetextof each).W epr ovideda write-in questionfor respondents toprovideattributes that they feltwer eimportant;wer eceived123 responses tothat question. Wereviewed theresponsesmanually andfound that theywer eparaphrasingorgivin gconcr eteexamples of one ofthe 15attributes oridentifying asubcase ofone ofthe attributes anddid notgenerate newinput.quotesdbs_dbs19.pdfusesText_25

[PDF] [PDF] What Makes a Great Manager of Software Engineers? - Microsoft