pro rata and user centric distribution models: a comparative study · 2018. 2. 21. · therefore,...
TRANSCRIPT
DIGITALMEDIAFINLANDOY|digitalmedia.fi
FinnishMusicPublishers’AssociationFinnishMusicians’UnionFinnishSocietyofComposersandLyricistsTheSocietyofFinnishComposersProRataandUserCentricDistributionModels:AComparativeStudy30.11.2017JariMuikku
2(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
TABLEOFCONTENTS
1. Introduction.......................................................................................................................31.1. Background.........................................................................................................................................31.2. Aims.....................................................................................................................................................31.3. Credits.................................................................................................................................................32. DistributionModels...........................................................................................................42.1. General................................................................................................................................................42.2. ProRataModel...................................................................................................................................52.3. UserCentricModel.............................................................................................................................63. ResearchMaterial..............................................................................................................64. AnalysisandResults...........................................................................................................74.1. TheResultsoftheStatisticalAnalysis.................................................................................................74.2. TheResultsforIndividualArtistsandTracks......................................................................................85. Conclusions......................................................................................................................116. Appendix:TheAnalysisProcess.......................................................................................136.1. ProRataModel.................................................................................................................................136.2. UserCentricModel...........................................................................................................................13
3(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
1. INTRODUCTION
1.1. BackgroundAccordingtotheIFPI,digitalrevenuesmadeup50%ofthetotalrevenuesofrecordedmusicbusinessongloballevelin2016.Theshareofstreamingofthedigitalrevenueswas59%anditisgrowingfast.InsomeforerunningterritorieslikeNordiccountriestheshareofstreaminghasalreadyexceededthelevelof80%oftheoverallrevenuesoftherecordedmusicbusiness.1Therefore,ithasbecomeveryimportanttostudycloserthecurrentdistributionmethodsofthesubscription-basedstreamingservices.Accordingtoourknowledgeallmajoraudiostreamingservicesarecurrentlyusingtheproratamodelbutvariousrightholdergroupshaveexpressedneedstostudyalsoalternativemethodssuchastheso-calledusercentricmodel.
1.2. AimsTheaimofthestudyistocomparethecurrentproratasystemandthealternativeusercentricsystemwithrealdatafromamajorstreamingserviceinordertoseehowthemethodsaffectdistribution.Thehypothesesofthestudywere:
1. Thecurrentproratasystemfavoursonlyfewtop-levelrightholders.2. Theconsumerswhopaymonthlysubscriptionfeesdonotknowwhichartists,worksand
otherrightholdersgettheirmoneyregardlessoftheirlisteninghabits.3. Thecurrentdistributionsystemdoesnottakeintoaccountthedifferencesbetweenthe
durationsofthevariouskindsofmusicalworksbutaretreatedequally.
1.3. CreditsThestudywascarriedoutduringApril-November2017.Thestudywasdoneintwophases.Thefirstpartwasthestatisticalanalysis,whichwasdonebyDr.PradeepDurgamofAaltoUniversity.Thesecondpart,writingofthereportandsomefurtheranalysiswasdonebyConsultant,Dr.JariMuikkuofDigitalMediaFinland.
1http://www.ifpi.org/downloads/GMR2017.pdf
4(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
2. DISTRIBUTIONMODELS
2.1. GeneralAllmajormusicaudiostreamingservicessuchasSpotifyandAppleusecurrentlyproratamodelintheirdistributiontotherightholders.Analternativemethod,usercentric,hasbeenatopicofinternationaldebateforawhilebut,sofar,noneofthemostimportantstreamingserviceshaveusedit.Therearesomeannouncedon-goingusercentricpilotprojectslikeinFrancebetweenDeezerandSACEM,but,atthetimeofwritingthisreport,theresultsoftheseprojectswerenotpubliclyavailable.ThemostextensiveacademicstudyonthistopicistheresearchprojectbyUniversityofOslo(2014)2.AsimilarkindofstudywasdonealsoinDenmarkin2014,anditusedsimilarmethodsinordertomakethesetwostudiescomparable3.BothstudieswerebasedonnationalreportsfromNorwegianstreamingserviceWiMP(laterTidal).Theoverallincomeofamusicaudiostreamingservice,whichconsistsmainlyofsubscriptionfeesand/oradvertisingrevenues,issharedbetweenthreemaingroups:GROUP RIGHTHOLDERS SHARE(ESTIMATEDTOTAL)Recordings
Recordcompaniesandotherproducers4Performers5
55-60%
Recordedworks ComposersLyricistsArrangersMusicPublishers
10-15%
Streamingservices --- 30%Incaseofsubscription-basedstreamingservicesthemostcommonmonthlyfeeforanindividualconsumeris€9.99permonth.Serviceshavealsovariouskindsofofferings,specialpricesfor
2https://www.hf.uio.no/imv/forskning/prosjekter/skyogscene/publikasjoner/usercentric-cloudsandconcerts-report.pdf3https://koda.dk/fileadmin/user_upload/docs/Analysis_Music-Streaming-In-Denmark_2014.pdf4Otherproducers=e.g.productioncompaniesandartiststhemselves.Distributorsandaggregatorsdonotusuallyownanyrights,buttheygettheirsharefromrecordcompanies’revenues.5Thisrefersmainlytotheartists,whogetroyaltiesfromrecordcompanies.
5(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
students,familypackages,andsoon,butinthisstudy,weusethe€9.99permonthpersubscriberasthecalculationbasis.Ifwedeductservices’share(30%),theremainingsharableamountofmoneyperuserpermonthis€6.993.ThissumincludesVAT,butwedonotdiscusstheVATissuesofthecross-borderInternetservicesinthiscontext.Inthisstudy,wedonotanalysethefore-mentionedsharesorhoweachgroupdividesitssharebetweentherespectiveparties.Forthesakeofclarity,weconsiderallroyaltiespaidbythestreamingservicesas100%incomeforallfore-mentionedrightholders.Inthisstudy,theterm“track”meansasinglerecording,whichusuallyconsistsofonerecordedmusicalwork.
2.2. ProRataModel6Intheproratamodel,allrevenuespermonthareputtogetherandsharedbetweenindividualtracksaccordingtothenumberoftotalstreams.7Theformulaoftheproratacalculationmethodforanindividualtrackisasfollows:
Totalno.ofstreamsofTrackA_________________________XTotalrevenueTotalno.ofstreams
Forexample,ifTrackAgets100streams,thetotalnumberofstreamsduringamonthisonemillionandthetotalrevenueafterservice’sdeductionis500,000euros,theshareofTrackAiscalculatedasfollows:
100÷1,000,000=0.00010.0001x€500,000=€50
Thesumof€50isdividedbetweentherightholdersofTrackA(asdescribedinchapter2.1).
6Thenumberspresentedinchapters2.2and2.3aretheoreticalanddonotrefertotheSpotifyreportmaterialusedinthisstudy.Theyareonlyexamplestoclarifythecalculationmethods.7Itshouldbenotedthatthiscalculationformulaisasimplifiedmodel,whichdoesnottakeintoaccountthemoredetailedcalculationmethods,whichareusedinrealityasdefinedincontractsbetweenthevariousparties.
6(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
2.3. UserCentricModelIntheusercentricmodel,thecalculationisbasedonthelisteninghabitsofeachindividualuser.Theformulaoftheusercentriccalculationmethodforanindividualtrackisasfollows:
No.ofstreamsofTrackAlistenedbyUser1____________________________________ XTotalrevenuefromUser1No.oftotalstreamslistenedbyUser1
Forexample,ifthetotalnumberofstreamslistenedbyUser1duringamonthis100,theshareofTrackAoutofthese100streamsis25,andtherevenuefromUser1is€6.99,theshareofTrackAiscalculatedasfollows:
25÷100=0.250.25x€6.993=€1.75
Thesumof€1.75isdividedbetweentherightholdersofTrackA(asdescribedinchapter2.1).
3. RESEARCHMATERIALTheresearchmaterialconsistedofaconfidentialusagereport,whichwaskindlyprovidedbySpotify.Thematerialwasavailableonlytotheresearcher,Dr.PradeepDurgamofAaltoUniversity.ThematerialconsistedofusageinformationonFinlandduringthemonthofMarch2016.ItincludedallSpotifyPremiumsubscribers(anonymised)inFinlandandallthestreamstheyhadlistenedduringthisperiodoftime.Thereportconsistedofmorethaneightmillionindividualstreams.8
8TheanalysisprocessstepshavebeendescribedindetailintheAppendix.
7(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
ThedatareceivedfromSpotifywasarrangedaccordingtothefollowingparameters:
1. UserID:UniqueIDsperuser92. TrackID:AutomaticallygeneratedtrackIDs3. TrackTitle:Titleofthetrack4. TrackArtists:ArtistoftheTrack5. AlbumTitle:Albumtitlefromwhichthetrackwasplayed6. StreamCount:Numberoftimesthetrackwasstreamed
Theresearchertookarandomsampleof10,000trackstorepresentthewholematerial.Thissampleincludedtracksfrom4493individualartistsand22,496streams(individuallisteningtimes),listenedby8051userIDs.
4. ANALYSISANDRESULTSTheanalysisispresentedintwosections.Inthefirstpart,wepresenttheresultsofthestatisticalanalysis.Inthesecondpart,wepresenttheanalysisfromthepointofviewofindividualtracksandartists.
4.1. TheResultsoftheStatisticalAnalysisThestatisticalanalysisoftheresearchmaterialwasdonebyusingthePearsonTwo-TailedCorrelationAnalysismethod.Theresultswereasfollows:
Correlations
RevenueDiffUser-Pro
ConsolidatedStreamCount
RevenueDifference:UserCentricvs.ProRata
PearsonCorrelation 1 -.769**Sig.(2-tailed) .000
N 4493 4493ConsolidatedStreamCount PearsonCorrelation -.769** 1
Sig.(2-tailed) .000 N 4493 4493
**=Correlationissignificantatthe0.01level(two-tailed)N=Numberofindividualartists
ThetableshowsthatthereisahighcorrelationbetweentheStreamCountandtheRevenueDifferencebetweentheUserCentricandProRatamodels.
9Theuser-datawastotallyanonymized.
8(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
Thenegativesign-.769meansthatthevariables,StreamCountandRevenueDifference,aremovinginoppositedirections.ThismeansthatastheoverallStreamCountdecreases,therevenuedifferencebetweentheUserCentricandProRatamodelsincreases.ThisindicatesthatUserCentricrevenuefortrackswithsmallernumberofstreamsgetsbiggerasthetotalnumberofstreamsperusergetssmaller.Therefore,UserCentricrevenuefortrackswithsmallernumberofstreamswillincreasewhenthetotalstreamcountperuserislower.Thetrackswithsmallernumberofstreamswillearnmorefromeachuser,whichindicatesthattrackswithsmallernumberofstreamswillearnrelativelymoreinUserCentricmodelthaninthecurrentProRatamodel.
4.2. TheResultsforIndividualArtistsandTracksTheresultsofthemoredetailedanalysisweredecidedtobearrangedaccordingtotheartists10.The4493individualartists11,whowereincludedinthesampleof10,000tracks12,weredividedintothreemaingroupsaccordingtothetotalnumberofstreams.Thepercentagefiguresinthefollowingdiagramrefertotheirshareofthetotalnumberofartists,andthestreamcounttothetotalnumberofstreamsperartistwithinthesample(canincludeonlyonetrackorseveraldifferenttracks):
10Itshouldbenotedthateventhoughweusehereartistsastheleadingparameteroftheanalysis,theanalysisconcernsallrightholdersofeachindividualtrack.11Asthereportmaterialdidnotincludeinformationonthenationalityoftheartists,wewerenotabletocomparetheresultsbetweennationalandinternationalartists.12Dr.Durgamdidalsoanadditionaltestcomputerrunwithasampleof50,000trackswithalimitednumberofartistsinordertoverifytheresultsofthesmallersample.Theresultsconfirmedthattheoverallresultswerethesameasinthesampleof10,000tracks.Themaintrendisthatasthesizeofthesamplegrows,therelativesizeandthestreamcountoftheBasic-TierremainsthesamebutthedifferencebetweenthebasicandTop-Tierstreamcountsgrowsexponentially.
9(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
Inallfore-mentionedgroupsthedifferencesinthefinancialresultsbetweentheprorataandtheusercentricmodelswerequitedramaticfromthepointofviewofanindividualartist.Intheproratamodelthesituationisquiteclear.Themoreyougetstreams,thebiggerisyourshareofthetotalrevenues.TheshareoftheTop-Tierartistsofthetotalrevenuewas9.9%accordingtotheproratasystemwhereasonly5.6%accordingtotheusercentricsystem.NoneoftheTop-Tierartistsgotmorerevenuesintheusercentricdistributionmodel.Thedifferencebetweenthetwodistributionmodelsinthetop-tiershowsthattheproratamodelfavoursthefewtopartistswhenthecalculationisbasedonlyonthetotalnumberofstreamsperartist.However,intheMid-andBasic-Tiersthedifferencesbetweentheprorataandusercentricmodelsforindividualartistswerepercent-wisehightobothdirections.Inotherwords,artistscouldgeteithermuchmoreorlessintheusercentricmodeldependingonthecasewithinthesamplereportmaterial.Intheusercentricmodeleachartist’ssharedependson,firstly,thetotaloverallnumberofstreamsorwhetheritislowerorhigheraspresentedinchapter4.1.Secondly,itdependsontheindividuallistener’shabitsorwhetherhe/shelistensonlycertainartists,howmanytimescertaintracksarelistened,howthelistenedstreamsspreadamongvariousartists,andsoon.
Top-Tier:~0.4%StreamCount:
100+
Mid-Tier:~9.6%StreamCount:10-99
Basic-Tier:~90%StreamCount:1-9
10(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
Theonlyartistgroup,whichgotalmostthesamefinancialresultsinbothmodels,wasartists,whoarepopularinmanyconsumersegmentsandhavealargelistenerbase.Thisandallotherobservationsseemedtoconcernbothdomesticandinternationalartists13.Popularartists,whowouldhavegotmuchlessmoneyintheusercentricmodel,aremostlikelylistenedbyasmallernumberofsubscriberswholistentheirtracksmanytimes.Popularartists,whoarelistenedbyalargenumberofindividuallistenersandwholistenthemfewertimes,arelessaffectedbythechangesbetweenthetwomodels.Thisseemedtorelatealso,toacertaindegree,todifferentmusicalgenres14.Incaseofmostartists,whogetonlyfewstreams,thechangesarepercent-wisedramaticbetweenthetwomodels.Intheusercentricmodeltherevenuesaredirectlyaffectedbyeachsubscriber’slisteninghabits.Ifasubscriberlistensonlyacertainartisteitherfewtimesoralotoftimes,his/hermoneygoesonlytothisartistandtherespectivetracks.However,ifhis/herlisteningisspreadamongmanyartists,therevenueisalsospreadaccordinglyandaffectseachartist’sandtheothertrack’srightholders’revenuesinanegativeway.Theextremealternativescanbedemonstratedinthefollowingway:
• Ifasubscriberlistensonlyonetrackonceduringamonth,thewhole€6.933wouldgo,accordingtotheusercentricmodel,tothattrackanditsrightholders.
• However,intheproratamodelthismoneywouldbejustaddeduptotheoverallpot,whichwouldbedividedaccordingtothenumberofstreamsofeachtrack,andtherightholdersofthatonetrackwouldgetonlyafractionofthesameone-track-once-a-month-user’smoney.
Furthermore,ifthelisteninghabitsofvarioussubscribergroupsarecumulated,theoveralleffectscanberemarkableintheusercentricmodel.Accordingtothestatisticalanalysisofthesample,thechangesof50%ormoretobothdirectionsbetweenthetwomodelsarenotrare.TheoverallshareoftheTop-Tierartistsisquitedifferentinthetwomodels.Comparedwiththeproratamodel,intheusercentricmodelalmosthalfoftheTop-Tier’srevenueswouldhavebeenspreadamongtheartistsandtheotherrightholderswithfewerstreams(Mid-TierandBasic-Tier).Allinall,themostimportantpointincomparingthetwomodelsistherelationbetweenthetotalnumberofstreamsandthelisteninghabitsofasinglesubscriberorsubscribergroups.13Itmustbeemphasizedthatthiswasnotanalysedstatisticallyasthedatadidnotincludeinformationonthenationalityoftheartists.14Thispointofviewwasnotanalysedstatisticallyasthedatadidnotincludeanygenreclassification.
11(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
Intheproratamodelonlythenumberofstreamscount,whereasintheusercentricmodelthetotalnumberofstreamspersubscriber,aswellashowtheyarespreadamongvarioustracksandartists.Therefore,theusercentricmodelislesspredictable.Itisalsoclearthattheproratamodelfavoursthefewtop-tierartistswhogetbiggestamountsofstreams.Thisanalysiscouldalsohavebeendonebasedonworksinsteadofartists.Inthatcasetherewouldhavebeendifferentkindsofvariationssuchasseveraldifferentrecordingsofaworkbyseveraldifferentartists,workscreatedbyseveralcomposersandlyricists,andsoon.However,weassumethattheresultsofthework-basedanalysiswouldhavemostprobablybeenquitesimilar.Itcouldalsobestudiedifsubscriberschoosetrackstheylistenbasedonanartist,aworkoracombinationofthem,orsimplyprefertouseready-orself-madeplaylists.Subscriberscanalsorelyoneachservice’salgorithms,whichlearnthetasteofeachuserandofferrecommendationsorplaylistsbasedontheirlisteninghabits.Itshouldalsobenotedthatthecurrentproratadistributionsystemdoesnottakeintoaccountthebigdifferencesbetweenthedurationsofthevariouskindsofmusicalworks.Forexample,eachplayofathree-minutepopsongistreatedindistributioninthesamewayasaplayofa15-minuteclassicalmusicwork.Inthisstudy,itwasnotpossibletoanalysethis,astheresearchmaterialdidnotincludeinformationonthedurationoftheindividualtracks.
5. CONCLUSIONSThestudyhasdemonstratedthebasicdifferencesandcorrelationsbetweentheprorataandtheusercentricdistributionmodels.Thebasictrendisthatastheoverallstreamcountdecreases,therevenuedifferencebetweentheusercentricandtheproratamodelsincreases.Theproratafavoursartistsandtracks,whichgetthebiggestamountofplayedstreamsregardlessiftheyarecreatedbyalargenumberofuserswithfewplaysorasmallernumberofuserswhohaveplayedthemrepeatedly.Theusercentricmodelfavoursartistswithsmallernumberofstreams,especiallywhentheoverallstreamcountissmaller.However,itshouldbeemphasizedthatthepositivefinancialeffectisnotautomaticinallusercentriccasesbuttheresultmayaswellbetheopposite.Theresultsdependonthecumulativeeffectsofbothindividualandusergroups’listeninghabits.
12(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
Inanycase,theusercentricmodelgivesmoredirectpowertouserstotargetthemoneytheypayfortheservicetotheartistsortrackstheyfavourcomparedwiththeproratamodel,whichisnottransparentfromtheirpointofview.
13(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
6. APPENDIX:THEANALYSISPROCESSThereportconsistedofmorethaneightmillionindividualstreams.Thedatawasreceivedinatextformat.Itwasconvertedintoanexcelformat(.xsls)andSPSSformat(.sav)foranalysis.TheanalysisofthereportingmaterialprovidedbySpotifywasdonebyexecutingthefollowingsteps:
6.1. ProRataModelStep1:Togetthetotalsubscriptionfeecollectedfromtheusersinthe10,000tracks,thejustUserIDandtheStreamCountwereextracted.Step2:TheUserIDwasconsolidatedwithStreamCountthroughtheconsolidationfunctionunderthedatatab.Aftertheconsolidation,thenumberofuniqueUserIDswasreducedto8051andthestreamcountaddedtoeachUserID.Step3:TheTrackArtistandStreamCountwereextractedfromtheoriginaldatafileandcopiedtoanewexcelsheet.TrackArtistandstreamcountwereconsolidatedthroughtheconsolidationfunctionunderthedatatabintheexcelsheet.Afterthisiswaspossibletocountthetotalamountcollectedandthetotalamountfordistribution:
Collected: 8051x€9.99=€80,429Fordistribution: €80,429x0.7=€56,300(30%deductionforSpotify)
Step4:Thedistributionwascalculatedforeachartistaccordingtotheformulapresentedinchapter2.2.
6.2. UserCentricModelStep1:TheconsolidatedUserIDandconsolidatedStreamCountwereextracted.Step2:Therevenueperstreamforanindividualuserfeewascalculatedasfollows:
Revenuereceivedfromeachuser=€9.99Deduction(30%)forSpotifyFinalrevenuereceivedfromperuser=0.7x€9.99=€6.993
14(14)
ProRataandUserCentricDistributionModels30.11.2017
DIGITALMEDIAFINLANDOY|digitalmedia.fi
Revenueperstreamforeachindividual’suserfee=Revenuereceivedfromeachuser÷consolidatedstreamcount
Theresultwasrevenueperstreamforeachuser.Step3:TherevenueperstreamforeachuserandtheconsolidatedUserID(count8051)wasthenplacedalongsidewiththeUserIDandTrackArtists,whichhadtheoriginalcountof10,000.ThentherevenueperstreamwasplacednexttotheindividualTrackArtist(withcount10,000)andUserID(withcount10,000).AftertheSUMIFfunctionoftheexcelsheetthecountofcolumnofrevenueperstreamwas10,000.Step4:Revenueperstreamandperuserwascalculatedwiththecountof10,000.Step5:ThentheUserCentricRevenueperArtistwascalculatedbyusingSUMIFfunctiononcolumnTrackArtist(count10,000)andrevenueperstreamthatwascalculatedearlier.Theresultwasanewcolumnofusercentricrevenueforallartists.Step6:Theartistswererepeatedandtheamounttheyearnedwererepeated.Thereforeartistsandtheirrevenueswereconsolidatedatthesametimeassumminguptheirnumberofstreams.Step7:Afterconsolidation,theProRataRevenueswereplacedalongsidewithUserCentricRevenues.ThentheRevenuedifferencewascalculated.Step8:AfterhavingtheStreamCount,ProRataRevenue,User-CentricRevenueandthedifferencebetweenUserCentricandProRataRevenuesIBMSPSSwasused.ThecorrelationfunctionwasusedinordertounderstandthecorrelationbetweenStreamCountandtheRevenueDifference.