브랜드저널리즘을위한소셜데이터기반의기업콘텐츠전략에관한연구

Abstract

Theeffectivenessofmassmedia,theexistingmarketingmeasureforcorporatebrandstendstobegraduallydecreased,astherelevantinformationismoreactivelyandselectivelycollectedinthecustomer-friendlyonlineenvironment.Manycompanies,therefore,adoptthe‘brandjournalism’forcreatingbrandstoriesbasedontheirowncontents(i.e.,Firm-generatedcontent)andcommunicatingwithcustomersasanewmarketingmethodology,regardlessoftheirtypesofindustry,andfocusonSNSasameansforembodyingit,whichcanreducethecostfortheestablishmentofonlineplatforms.Ontheotherhand,althoughthecompanies’onlinebrandsjournalismactivitiesusingSNShavebeengeneralized,itisoftencriticizedforalackofdiscussionsoncontentstrategiesreflectingcustomers’responses.PreviousstudiespaidattentiontothefunctionalconvenienceofSNS,asafactorenhancingthecorporateperformance,butSNSnolongerhasanadditionaleffect,asitisfixed,andithasdifficultyindirectlyprovidinginsightsforcontentstrategies. Hence,thepurposeofthisstudyistoexaminecorporatecontents,especially,toenhancethecontentpopularity,notthefunctionalconvenienceofSNS,forthesuccessfulcontentstrategiesonSNS.Becausethecontinuouslyhighcontentpopularity,asacorporateassetcontributestocreatingactualrevenue,whilelowcontentpopularityandcustomers’continuouslynegativeperceptionsonthemmayresultinthecustomerdefectionandadropinsales. Toprovidewideinsightsforstrategiesforenhancingthecontentpopularity,thisstudyattemptstoresolvequestionsregardingtheexplorationonantecedentsofthecontentpopularitythroughstatisticalanalysesandthedesignofamodeltopredictthecontentpopularityusingdeeplearningalgorithms. First,previousstudiesfocusedontheeffectsofcustomers’experientialvalueortheSNSskill,toexploretheantecedentsofthecontentpopularity.Inthisregard,theneedtoexaminefactorscontributingtothecommunicationofbrandmessagesrelatedtothecompositionofcontentswasemphasized,buttherearestilllittlediscussionsonit.Thisstudy,therefore,testeddirecteffectsofsub-factorsofeachfeaturesonthecontentpopularity,byquantifyingvisual,textualandtimeframefeaturesfromfirm-generatedcontent.Inaddition,itconductedanempiricalanalysisonamediationeffectofcontentconsistency,anstrategicapproachforimprintingbrandmessagesandformingthebrandidentity,asitisnecessarytoexploremediationfactorsforenhancingthecontents’abilitytotransmitthebrandmessages.Then,itproposesapracticalcontentcompositionstrategyandusesthePROCESSMacroModel4asananalysismethodology. Second,althoughthepredictionofcontentpopularityisusefulforunderstandingcustomers’experienceanditispossibletousefullyusethepredictionmodelasanewbusinessmodel,thedesignofahybridmodelbasedondeeplearning,foreffectivelyunderstandingtheinformationofcontentsthatareveryvariouslycomposediscontinuouslyrequired.Thisstudy,therefore,triedthemulti-classificationofthecontentpopularity,afterusingvisual,textual,timeframeandengagementfeaturesoffirm-generatedcontentasinputvariables,basedonCNN-TabNet,ahybridmodelreplacingtheCNN’sclassifier,adeeplearningalgorithm,withTabNet.Itvectorizedimagesandtexts,atypicaldatathroughalearningalgorithm,tominimizethelossofthem.Itproposedamodeltopredicttheimprovedcontentpopularityanddrewamethodforcomposingpopularcontentsthatcanbeinterpretedfrompredictions,toincreasetheusabilityofthepredictionmodel. Itcollectedatotalof5,044contentswhichacompanyselectedbyForbes’2020theworlds’mostvaluablebrandsissuedontheglobalaccountofInstagram,andusedSPSS26.0andPython3.8.10foranalyzing. Thefindingsareasfollows:First,theimprovementofimagequalitythroughthebrightnesscontrolofcolors,andthediversificationofcolorsorobjectsareimportantinexpressingimages,forenhancingthecontentpopularity.Thereductioninthelengthofsentencesorwordsandthediversificationofkeywordsarerequiredforhighcontentpopularity.Also,consideringthetimeslotinwhichSNSusers’issuingrateislowerwasfoundtobecontributetoheightenthecontentpopularity.Second,itwasfoundthattheconsistentexpressionaboutvisualobjectsortextualtopicsisastrategyforcontentcompositionwhichcanenhancethecontentpopularity,byincreasingtheimprintingeffectofbrandmessages. Next,theCNN-TabNetthisstudyproposedtopredictthecontentpopularityshowedthepredictionperformancethatisabout10~40%higherthanthatofmulti-inputCNNaswellasthatofSVMandXGBoostbasedonmachinelearningalgorithms,notbasedonneuralnetworks.Inaddition,thisstudyexaminedhowpopularcommunicativeandinformativefirm-generatedcontentandinformalSNScontentsarecomposed,basedonthetheoreticalbackgroundfromthepredictionresultsoftheCNN-TabNet.ThesefindingsshowthattheCNN-TabNetwhichcancomplementofthestructuraldisadvantageofdeeplearningalgorithmsiseffectiveinpredictingthecontentpopularityandthatitispossibletoincreasetheusabilityofthepredictionmodelasthecontentstrategy,byinterpretingthepredictions. Thisstudyhasanacademicimplication,inthatitcontributestoextendingthescopeofresearch,byviewingcontentsfromthecorporateperspective,conductinganempiricalanalysisonthecontentcompositionandthedesignofthepredictionmodelandprovidingversatileinsightsoncontentstrategies.Inaddition,thefindingshaveapracticalimplication,inthattheycancontributetoimprovingthemanagementforthebrandjournalismonSNS.|고객이친숙한온라인환경에서더욱능동적이고선택적으로정보를수집하게되면서기업브랜드의기존마케팅수단인매스미디어의효과성은점차감소하는추세이다.이에따라업종을불문한많은기업이자체적인콘텐츠로브랜드스토리를만들며고객과소통하는‘브랜드저널리즘’을새로운마케팅방법론으로채택하게되었으며,구현수단으로는온라인플랫폼구축비용을절감할수있는SNS가주목받고있다.한편,SNS를활용한기업의브랜드저널리즘활동이보편화되었음에도불구하고고객의반응을반영하는콘텐츠전략에대한논의가부족하다는비판이많다.기존연구에서는SNS에서기업의성과를이끄는요인으로서SNS의기능적편리함에주목하였으나,이는SNS의고착화로인해더이상영향력이발현되지않으며직접적인콘텐츠전략에관한통찰력을제시하기어렵다는한계가있기때문이다. 이에본연구는SNS에서의성공적인콘텐츠전략을위해SNS의기능적편리함이아닌기업콘텐츠자체를연구대상으로하며구체적으로콘텐츠인기도의향상을연구목적으로하였다.왜냐하면지속적으로높은콘텐츠인기도는기업의자산으로서실제수익창출에기여하는반면,낮은인기도와함께지속적인고객의부정적인인식은고객이탈과매출하락을야기하기때문이다. 본연구는콘텐츠인기도를높일수있는전략에관하여폭넓은통찰력을제시하고자,통계분석을통한콘텐츠인기도의선행요인탐색과딥러닝알고리즘을활용한콘텐츠인기도예측모델설계에대하여연구문제를나누어수행하였다. 먼저콘텐츠인기도의선행요인탐색에관하여,기존연구는고객의경험적가치나SNS숙련도의영향력에초점을맞추었다.이에대해콘텐츠구성과관련하여브랜드메시지전달에기여하는요인규명의필요성이제기되었으나여전히논의가부족한실정이다.따라서본연구는콘텐츠의구성으로부터이미지특성,텍스트특성,시간특성을정량화한것을콘텐츠구성특성으로하고,각특성의하위요인이콘텐츠인기도에미치는직접효과에대해검증하였다.또한콘텐츠의브랜드메시지전달력을더욱높일수있는매개요인탐색의필요성에따라,브랜드메시지각인과브랜드정체성형성을위한전략적접근인콘텐츠일관성의매개효과를실증분석하였다.이를통해실질적인콘텐츠구성전략을제안하며분석방법론으로는PROCESSMacroModel4를활용하였다. 다음으로콘텐츠인기도예측은고객경험파악및새로운비즈니스모델로서유용하지만,구성방식이매우다양한콘텐츠의정보를효과적으로이해할수있는딥러닝기반의하이브리드모델설계의필요성이꾸준히제기되고있는실정이다.이에본연구는딥러닝알고리즘인CNN의분류기를TabNet으로대체한하이브리드모델인CNN-TabNet을활용하여콘텐츠의이미지,텍스트,시간특성및인게이지먼트특성을입력변수로한뒤콘텐츠인기도를다중분류한다.이때,비정형데이터인이미지와텍스트는정보유실을최소화하고자학습알고리즘을통해벡터화하였다.이를통해개선된콘텐츠인기도예측모델을제안하고,콘텐츠전략으로서예측모델의활용성을높일수있도록예측결과로부터해석가능한인기있는콘텐츠구성방식을도출하였다. 연구수행을위해Forbes’2020Theworlds’mostvaluablebrands로부터선정한기업이인스타그램의글로벌계정에발행한총5,044개의콘텐츠를수집하였으며변수추출및분석을위해SPSS26.0과Python3.8.10을사용하였다. 연구결과는다음과같다.첫째,콘텐츠인기도향상을위해이미지표현에는색상의밝기조정을통한화질의개선,색상의종류및물체의다양화가중요한반면,텍스트표현에는문장및단어길이의절감과키워드의다양화가요구되며SNS이용자의콘텐츠발행률이낮은시간대를고려하는것이도움이된다는것을확인하였다.둘째,이미지의물체나텍스트의주제에대한일관적표현은브랜드메시지각인효과를높여콘텐츠인기도를형성할수있는콘텐츠구성전략임을확인하였다. 다음으로콘텐츠인기도예측을위해본연구에서제안한CNN-TabNet은다중입력CNN뿐만아니라신경망기반이아닌머신러닝알고리즘인SVM및XGBoost보다10~40%가량높은예측성능을보였다.또한CNN-TabNet의예측결과로부터이론적근거를토대로인기있는소통성및정보성콘텐츠와이를포괄하는SNS콘텐츠의구성방식을규명한결과,딥러닝알고리즘의구조적단점을보완하는방식의CNN-TabNet이콘텐츠인기도예측에효과적이며예측결과에대한해석을통해콘텐츠전략으로서예측모델의활용성을높일수있다는것을확인하였다. 본연구는기업의관점에서콘텐츠를바라보고,콘텐츠구성방식과예측모델설계에대해실증연구하여콘텐츠전략에관한다각도의통찰력을제시함으로써연구범위의확장에기여한다는학문적의의가있다.또한본연구의결과는SNS에서의브랜드저널리즘을위한경영개선에기여할수있다는실무적의의가있다.