mrc ad verification supplement- enhanced content level ... ad verification... · 2 content level...

30
© Copyright Media Rating Council, Inc. All rights reserved. MRC Supplement to IAB Guidelines for the Conduct of Ad Verification: Enhanced Content Level Context and Brand Safety September 2018 FINAL Sponsoring associations: Media Rating Council (MRC) American Association of Advertising Agencies (4A’s) Association of National Advertisers (ANA) Interactive Advertising Bureau (IAB U.S.)

Upload: others

Post on 08-Jul-2020

7 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved.

MRCSupplementtoIABGuidelinesfortheConductofAdVerification:EnhancedContentLevelContextandBrandSafety

September2018FINAL

Sponsoringassociations:MediaRatingCouncil(MRC)AmericanAssociationofAdvertisingAgencies(4A’s)AssociationofNationalAdvertisers(ANA)InteractiveAdvertisingBureau(IABU.S.)

Page 2: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 2

TableofContents1 ExecutiveSummary ...................................................................................................... 31.1 AbouttheDevelopmentofthisSupplement ....................................................................................... 6

2 ContentLevelGranularity ............................................................................................. 63 SocialMediaandUGC ................................................................................................... 83.1 DataFreshnessandQuality .................................................................................................................. 93.2 CurationofUGC .................................................................................................................................. 103.3 Scalability ............................................................................................................................................ 113.4 MachineLearning ............................................................................................................................... 113.5 HumanIntervention ........................................................................................................................... 133.6 MonetizationControls ....................................................................................................................... 133.7 TaxonomyandCategorization ........................................................................................................... 14

4 Adjacency ................................................................................................................... 165 Metadata ................................................................................................................... 176 MobileApplications .................................................................................................... 207 IllegalSourcesandIndustryCommunication ................................................................ 218 EnhancedBuyerTools ................................................................................................. 229 RequiredReporting ..................................................................................................... 2210 UseofSamples ......................................................................................................... 2311 GlossaryofTerms ..................................................................................................... 2312 References ............................................................................................................... 2513 SupportingAssociationsandParticipatingOrganizations ............................................ 2514 ContactUs ................................................................................................................ 27AppendixA:SectionSummaries ........................................................................................ 28

Page 3: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 3

1 ExecutiveSummaryTheInteractiveAdvertisingBureau(IAB)andMediaRatingCouncil(MRC)producedtheGuidelinesforConductofAdVerification(theAdVerificationGuidelines)inFebruary2012,whichidentifiedacommonsetofacceptablemethodsandpracticesforconductingdigitaladverificationservices.Adverificationisaservicethatofferstechnologytoensurethatdigitaladsappearonintendedsitesandreachthetargetedaudience.Theoriginalprimaryservicelinesofadverificationare:(1)SiteContext,(2)Geo-Targeting,(3)AdPlacement,(4)CompetitiveSeparationand(5)FraudDetection.IncludedwithintheSiteContextservicelineistheconceptofBrandSafetyor,practicesandtoolstoensurethatanadwillnotappearinaContextthatcandamageanadvertiser’sbrand.Historically,BrandSafetyandtargetinghavebeenaddressedatadomain,siteorUniformResourceLocator(URL)levelwithContextclassificationandadalertingorblockingbasedonavoidanceortargetcategories.ForpurposesofthisSupplementContextreferstocertainattributesofaURL,site,mobileapplicationorpieceofcontentthatareusedasdeterminantsastowhetherthedestinationwhereadigitaladistobeservedrepresentsanappropriateenvironmentinwhichtheadshouldappear,asdeterminedbyparameterssetbytheadvertiser.BrandSafetyreferstopracticesandtoolstoensurethatadigitaladwillnotappearadjacenttoorinaContextthatcandamageanadvertiser’sbrand.SincethedevelopmentoftheIAB/MRCAdVerificationGuidelines,thedigitaladvertisingecosystemhasbecomemuchmorecomplexandwiththiscomplexity,ContextandBrandSafetyissueshavebecomemorenuancedandchallenging.CompoundingthiscomplexityisthemonetizationofSocialMediaPlatforms(referredtoasplatformsthroughoutthisdocument)andUserGeneratedContent(UGC),whichhasintroducedgreatvolumeandscaleaswellasreducedcontentcontrolwhencomparedtoprofessionallycuratedpublishercontent.TheMRCproducedSocialMediaMeasurementGuidelinesinNovember2015,whichhelpeddefinekeytermsforsocialmediaandestablishguidanceofmeasurementofsocialmediaactivity,audienceandadexposureonSocialMediaPlatforms.However,todate,theimpactofsocialmediamonetizationonContextandBrandSafetyhaslargelynotbeenaddressedinexistingmeasurementguidance.WhileexistingIABandMRCadmeasurementandverificationguidanceshouldcontinuetobeappliedtoSocialMediaPlatformsandUGC,thisSupplementisintendedtopromoteenhancedContextandBrandSafetyguidanceacrosstheseplatformsandcontentaswellastobeusedinconjunctionwithexistingguidanceaspartofMRCaccreditationauditing.Specifically,thisSupplementseekstoachievethefollowingobjectives:

• Establishguidanceformoregranular(contentlevel)ContextandBrandSafetymeasurementandreportingincludingvideoanddisplaycontentwithinadomain,site,platform,mobileapplicationorURL.

Page 4: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 4

• EstablishguidanceforContextclassificationandBrandSafetymeasurementofsocialmediaandUGCincluding:

o DataFreshnessandQuality;o CurationofUGC;o Scalability;o MachineLearning;o HumanIntervention;o MonetizationControls;ando TaxonomyandCategorization

• Establishguidancefordefiningandmeasuringdomain,site,platform,mobileapplicationorpageAdjacencyinordertoenableContextandBrandSafetymechanismswithinthisparameter.

• Establishguidanceforingestion,useandconsiderationofcontentMetadata(suchascontentcharacteristics,userandexposureactivity,commentsandotherrelevantdata)forContextandBrandSafetyconsideration.

• EstablishguidanceforContextandBrandSafetyconsiderationswithinmobileapplications.

• Establishguidanceforidentification,monitoringandtreatmentofcontentfromillegalsourcesordomainsandIndustrycommunicationregardingknownbadsourcesandBrandSafetymatters.

• Establishguidanceforenhancedbuyertoolsforbothtargetingandavoidance.• EstablishguidanceforreportingofcontentlevelContextandBrandSafetyoutcomes.• EstablishguidanceforuseofsamplesinContextandBrandSafetymeasurement.

SocialMediaPlatformsandlargedigitalpublishersmayhaveadvancedtechniquestoidentifyandcategorizecontentforpurposesofContextandBrandSafety,howevertheindependentapproachofthird-partyverificationvendorsremainsofvitalimportance.Approachestocontentleveladverificationwillneedtobeacollaborativeapproach.Platformsandpublishersarestronglyencouragedtoeitherallowdirectthirdpartymeasurementatacontentlevel,ortoactivelyprovidevalidatedfirst-partydataatagranularleveltoenableindependentContextandBrandSafetymeasurement,oneofwhichwillberequiredforcontentlevelaccreditationofthird-partyadverificationsolutions.Further,thecollectionandtransmissionofdataenablingcontentlevelContextandBrandSafetydeterminationsisstronglyencouragedtobesubjectedtoauditandindependentvalidation.ItisimportanttonotewhenconsideringthevariousaspectsofContextandBrandSafetyinthisSupplementthatdifferentbuyersmaynotonlyhavedifferenttargetsandavoidancecategories,butalsodifferenttolerancelevelsforriskdependingonbrandsuitability.Ultimately,adverificationorganizationsseekingcontentlevelaccreditationmustprovideconsistentanddiscretereportingtousersinordertoenablecustomizedtargetingandavoidanceandtoallowuserstostipulatewhichlevelsmeettheneedsofaparticularcampaignorbrand.

Page 5: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 5

Withtheissuanceofthisguidance,totheextentapplicable,MRC’saccreditationconsiderationsofContextandBrandSafetyadverificationfunctionswillbeappliedatthediscretedomain,siteorURL(property)levelandcontentlevelforbothdisplayandvideoforbothmeasurementandreportinggranularity.Domain,siteorURL(andevenmobileapplication)leveladverificationfunctionalityandreportingshouldcontinuetofollowtheguidanceintheoriginal2012IAB/MRCAdVerificationGuidelines,butcontentleveladverificationfunctionalityandreportingshouldapplythoseguidelinesaswellastheguidancewithinthisSupplement.ItisneitherexpectednorrequiredthatadverificationorganizationsdevelopContextandBrandSafetycapabilitiesinallareasdiscussedinthisSupplementinordertoachieveaccreditation.Forexample,adverificationorganizationsmaynotdevelopcapabilitiesfororchoosetosubmitforaccreditationaspectssuchasmobilein-applicationBrandSafetyfunctionalityorAdjacency(discussedanddefinedlaterinthisdocument).However,totheextentthesefunctionalitiesexistandareassessedforcompliance,allapplicableaspectsofcorrespondingsectionsofthisdocumentwillberequired.Further,adverificationorganizationsmustclearlydisclosespecificcapabilitiesandlimitationsandMRC’saccreditationconsiderationsanddecisionsrelatedtoContextandBrandSafetyadverificationfunctionswillbemadewithrespecttospecificcapabilities.Itispermissibleforadverificationorganizationswithcontentmeasurementgranularitytoreportavoidanceandverificationatthepropertylevelwithproperdisclosure.SuchpropertylevelreportingwithcontentlevelmeasurementgranularitywillbedistinguishedbyMRCfrompropertylevelreportingwithoutdiscretecorrespondingcontentmeasurement.Finally,MRCmayrequiresomeoverallminimumlevelofcapabilityorfunctionalityinordertoconsiderextendingaccreditationtoadverificationorganizationsforContextandBrandSafetyaccreditationatthecontentlevelandthiswillbedeterminedaspartofscopingdiscussionsbetweenMRC,ourauditorsandauditedadverificationorganizations.ThefollowingaretheanticipatedMRCBrandSafetyandContextaccreditationdistinctions:

MRCBrandSafetyandContextAccreditationDistinctionsMeasurement/ReportingGranularityLevel Property/Property Content/Property Content/Content

PrivacyConsiderationsforThisDocumentAdverificationorganizationsseekingMRCaccreditationarerequiredtoadheretorelevantMRCMinimumStandardsrelatedtorespondentanonymityandprivacyaswellasprivacyregulations.Whileplatformsandpublishersareencouragedtopassdetailedcollectedfirst-partydataincludingUGC(preandpost-processing)supportingadverificationmeasurementtothird-partyverificationorganizationstoenableindependentmeasurementincludingAdjacency,obfuscatedortruncateddatamaybemaintainedorpassedtosatisfythisrequirement,shouldtherebePersonalIdentifyingInformation(PII)orprivacyconcerns.Thisshouldbeavailableina

Page 6: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 6

transparentmannertoaccreditationauditorsandatadetailedleveltoallowgranularmeasurementreprocessingofreporteddatawherenecessary.Differentmetric/transactiontypesandvaryingrisksassociatedwithtransactiontypesshouldbeconsidered.PIIlegalrestrictionsmaydictateeliminatingoneormoreofcollectedfieldsfromexternallytransmittedrecordsoralteringthecontentoffieldsforidentityprotectionpurposes.Suchrestrictionsmaystillallowforalternativelevelsofdatatransmissionthatarestillsufficienttosupportgranularprocessingofadverificationdata.Inthesecasesdeviationsshouldbesupportedbytheplatform’soradverificationorganization’sprivacypolicyinrelationtoprivacyregulationsandrequirementsandshouldbeavailableforreviewbyauditors.

1.1 AbouttheDevelopmentofthisSupplementTheguidancecontainedinthisSupplementoriginatedfromaprojectledbytheMediaRatingCouncil(MRC)andsponsoredbytheAmericanAssociationofAdvertisingAgencies(4A’s),theAssociationofNationalAdvertisers(ANA)andtheInteractiveAdvertisingBureau(IABU.S.).Thisguidancewasdevelopedwiththeparticipationofacoreadvisorygroupofbuyer-sideorganizations,Industryassociations,largedigitalenterprisesandverificationvendors,aswellasalargegroupofDigitalmediacontentproviders,advertisingagencies,advertisers,vendors/consultants,measurementorganizationsandotherinterestedorganizations.Thisguidancewasalsoreviewedandapprovedbymajorbuyer-sidetradeorganizations(4As,ANA)andtheirconstituentsandthereafterprovidedtothepublicthroughaformalperiodofpubliccommentpriortoadoption.ThefinalguidanceispublishedandavailableontheMRCsiteandwillbere-assessedperiodicallytoensureitremainsapplicableovertime.

2 ContentLevelGranularityThird-partyadverificationfunctionshavetraditionallybeenfocusedondomain,site,URLorapplication(including“channels”withindomains)levelinformation(i.e.,determiningthesafetyorclassificationofaURLbasedlargelyontextanalysisofthecontent)asawhole.Forpurposesofthisdocument,thiswillbereferredtoaspropertylevel.Thisfocuswasnecessarytoenableadservingtostandardpages;however,thecurrentenvironmentdictatesContextandBrandSafetyguidancebeputinplacethataccountsforadservingwithincontentaswellastoURLsormobileapplicationswithdynamiccontent(contentthatisfrequentlyaddedorchangedincludingbyusersoutsidethecontrolofthepublisher)andchannelswherebypropertylevelclassificationsmaynotcompletelyordirectlycorrelatewithspecificpiecesofcontentpresent.Suchcontentlevelmeasurementdepthmustextendbeyondkeywordanalysisandincludeconsiderationofimages,video,textandaudiocontent.Adverificationorganizations(third-partymeasurers,platformsorpublishers)mustdistinguishContextandBrandSafetyfunctionalitybetweenpropertyandcontentlevelgranularityordepthofmeasurementandreporting.Ifitisexpectedthataspecificdomain/siteorgroupingof

Page 7: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 7

URLs/applicationswillexhibitthesametaxonomicattributesasmateriallyallofthecontentwithinthemonanongoingbasisasdemonstratedthroughperiodicanalysisandempiricalsupport,ContextandBrandSafetyclassificationsmaybeconsistentlyappliedacrossallcontentwithinthegrouping.However,totheextentthenatureofthecontent(displayorvideo)withinthepropertycouldmateriallydifferfromitsspecificpropertylevelclassification,ContextandBrandsafetydecisionsmustalsobemadeatthediscretecontentlevel.However,thisshouldbeinadditiontocontinuingtoprovideproperty-levelclassifications.Contentgranularityisparticularlyimportantwhencontentmaybeconsidered“unsafe”orcategorizedinaparticularavoidancecategorywherethegeneralpropertylevelclassificationisotherwiseconsideredsafeoracceptable.However,propertiesthatareconsideredunsafeorareclassifiedinaspecificavoidancecategoryshouldcontinuetobeclassifiedinthismannerevenifsafeoracceptablecontentappearsonthem.ThisincludesdynamicpublishercuratedcontentaswellasUGConSocialMediaPlatforms.Itispermissibleforadverificationorganizationswithcontentmeasurementgranularitytoreportavoidanceandverificationatthepropertylevelwithproperdisclosure.SuchpropertylevelreportingwithcontentlevelmeasurementgranularitywillbedistinguishedbyMRCfrompropertylevelreportingwithoutdiscretecorrespondingcontentmeasurement.Further,itispermissibleforadverificationorganizationstoclassifycontentorpropertiesbasedonURLkeywordswithoutcontentlevelmeasurement,howeverthismustbeclearlydisclosedanddoesnotrepresentcontentlevelmeasurementgranularityorreporting.Withtheissuanceofthisguidance,totheextentapplicable,MRC’saccreditationconsiderationsofContextandBrandSafetyadverificationfunctionswillbeappliedatthepropertyandcontentlevelforbothdisplayandvideo.Propertyleveladverificationfunctionalityshouldcontinuetofollowtheguidanceintheoriginal2012IAB/MRCAdVerificationGuidelines,butcontentleveladverificationfunctionalityshouldapplythoseguidelinesaswellastheguidancewithinthisSupplement.WhiletheadverificationtoolsandfunctionalityusedforeachrespectivelevelofContextandBrandSafetymaybesimilar,theywillneedtobeapplieddifferentiallyinconjunctionwithadditionaltoolsdiscussedthroughoutthisSupplementinordertobeeffectiveforcontentlevelgranularity.Adverificationorganizationsmustactivelydisclosecapabilitiesandlimitationsofmeasurementrelatedtopropertyvs.contentlevelgranularitytousersalongwithguidanceforproperuseofmeasurementoutputconsideringanymateriallimitationsofmeasurement.Further,itisneitherexpectednorrequiredthatadverificationorganizationsdevelopContextandBrandSafetycapabilitiesinallareasdiscussedinthisSupplementinordertoachieveaccreditation.

Page 8: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 8

3 SocialMediaandUGCGeneral ThecurrentdigitaladvertisingenvironmentdictatesContextandBrandSafetyguidancealsobeputinplacefordynamicUserGeneratedContent(UGC)onSocialMediaPlatformsincludingnewsfeedenvironments.Timingandscalepresentchallengesinlargedigitalenterprises,especiallyforvideo,andtheneedstoassessUGCandpostthatcontenttimelycreateadditionaldiligencechallenges.Thevolumeofcontentlikelynecessitatesautomation(includingMachineLearning),whichcanbesubjecttoerror,misinterpretationandambiguity.Arobustsetofrisk-basedcriteriaisnecessarytodrivefocusedHumanIntervention.ThepresenceofUGCfurtherpresentsnumerouschallengesforContextandBrandSafetymeasurementincludingreducedcontrolofcontentwhencomparedtopublisher-curatedcontent.TheintroductionofcommentssuchasinaThreadedMessagemaypresentadditionaldatapointstoanalyzeUGC,butmayalsochangetheContextandBrandSafetyclassificationoftheThreadedMessagewhenconsideredcollectively.Thevolumeanddynamicnatureofsuchcontentpresentsscalabilityissuesformeasurement.TheabilitytoscalemeasurementatagranularlevelandtoenforcedatafreshnessrulesiscriticaltoeffectivemeasurementofcontentforpurposesofContextandBrandSafety.AdverificationorganizationsseekingcontentlevelaccreditationmustincludeconsiderationofUGCandtheuniquenatureofplatformspecificaspectsofUGCindatacollectionmethodologiesandensuresystemsarecomprehensiveenoughtohandlescaleandperiodicrefreshing.Seefurtherguidanceonscalabilitylaterinthissection.Additionally,thecategorizationofthecontentitselfmaydifferfromtheuseractivityassociatedtoitinaThreadedMessageandtheoccurrenceofcertainkeywordsincommentsmayassistadverificationorganizationswithcontentclassification.WhilecommentsshouldbeconsideredandincludedincontentlevelContextandBrandSafetyfunctionality,adverificationorganizationsshouldalsotakecaretoappropriatelyconsidertheweightcommentscarryonclassificationsaswellasthefrequencyofcertainkeywordsthatmaybeclassifieddifferentlythanthecontentitself.ThespecificconsiderationandinclusionofusercommentsincontentlevelContextandBrandSafetydeterminationsmustbeempiricallysupportedanddisclosed,butrepresentsanimportantsignalthatshouldbeconsidered.Adverificationorganizationsmayutilizeexternalthirdpartieswithadditionalexpertiseincontentclassification(suchasprofessionaleditorialpersonnelandservicesandacademicresources)tofurthermitigatetheriskthatgeneralpersonnelorsurfacelevelreviewmaynotproperlyidentifyorcategorizecontent.Suchinvolvementofexpertexternalthirdpartiesisencouragedandifused,shouldbeleveragedtocalibrateMachineLearningparametersaswellastofocusHumanInterventionasdiscussedbelow.Useofthirdpartiesmustbedisclosed(althoughthespecificthirdpartiesneednotbedisclosed)andincorporatedintoauditorindependentvalidationofadverificationfunctionality.

Page 9: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 9

TheabilityofthirdpartyvendorstoaccuratelymeasureactivityoncertainplatformsmaybelimitedincertainsituationswithoutdirecttaggingorSoftwareDevelopmentKit(SDK)integrationinmobileapplications.Asaresult,platformsandpublishersarestronglyencouragedtoeitherallowdirectthirdpartymeasurementatacontentlevel(inclusiveofMetadatadiscussedbelow),ortoactivelyprovidevalidatedfirst-partydataatagranularleveltoenableindependentContextandBrandSafetymeasurement,oneofwhichwillberequiredforcontentlevelaccreditationofthird-partyadverificationsolutions.Further,thecollectionandtransmissionofdataenablingcontentlevelContextandBrandSafetydeterminationsisstronglyencouragedtobesubjectedtoauditandindependentvalidation.TotheextentthatathirdpartyvendorisnotabletodirectlymeasurecertainaspectsofcontentwithinapropertyforContextandBrandSafety,thevendormaylistthepropertyorportionsofthepropertyasuncategorizedorunknownwithcleardisclosureofcapabilitiesorlimitationswithrespecttospecificpropertyorcontentlevelgranularitytoenableavoidanceifausersochooses.Further,inthesesituations,propertiesthatallowUGCmaybegenericallyclassifiedassuch(andbedistinguishedasunknownwithregardtoBrandSafety)toenableausertochoosewhetherornottoplaceadsintheseenvironmentsincaseswherecontentleveldecisionscannotbemade.However,adverificationorganizationswillnotbeconsideredtooffercontentlevelContextandBrandSafetygranularityforpropertiesthatarelabeledgenericallyoruncategorizedwherethesemeasurementlimitationsexist.Incertainpre-bidsituations,DemandSidePlatforms(DSPs)mayhaveamaterialimpactonContextandBrandSafetydataanddeterminations.Inthesecases,theguidancerelatedtodatatransparency,qualityandvalidationcontainedthroughoutthissectionshouldbeappliedtoDSPsandconsideredbyadverificationorganizationsutilizingDSPdata.

3.1 DataFreshnessandQualityAsdiscussedintheAdVerificationGuidelines,ascontentevolvesinaveryfastanddynamicmanner,freshdatausedtoenableadverificationmeasurementisabsolutelynecessaryforaccurateadverification.Forexample,theeffectivenessofContextandBrandSafetymeasurementthatutilizesdynamicUGCmaydiminishovertimeifdataisnotperiodicallyrefreshedandrecollected.Therefore,thetimingoftheverificationserviceandunderlyingdataassumptionsmustbedisclosedandadverificationservicesmusthavestateddatarefreshpolicies.Atimestamporactiveandcleardisclosuresofdatalifeandcollectionpoliciesisrecommendedtoinformusersabouttheageoftheservice’smostrecentassessment.GiventhedynamicnatureofcontentonSocialMediaPlatforms,UGCandsurroundinguseractivity,itisessentialthatthefrequencyofdatacollectionandrefreshpolicies(includingmaximumdatalifeor“timetolive”policies)employedbyanadverificationorganizationforcontentlevelclassificationsareadaptedtothespecificplatformbeingmeasured(consideringscale,useractivityandpresenceofUGCaswellaspublisherpoliciesandcontrols)andthatthesepoliciesarebasedonempiricalsupportthatisperiodicallyanalyzed.Thesepoliciesmay

Page 10: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 10

alsobedynamictoaccountforcontent,suchasnewscontent,whichmayinitiallybedynamicandthenmorestaticovertime.Further,adverificationorganizationsareencouragedtoestablishminimumreportingthresholds(arequiredminimumnumberofdatapointsorobservationsbeforereportingContextandBrandSafetydeterminations),collectionperiodsorotherbaselineguardrailsbeforemakingContextandBrandSafetydeterminationsforspecificcontent.Thismayconsistofaminimumsetofparametersordatafieldsfordeterminationsbasedonstaticdataoraminimumnumberofobservationsifbasedindynamicdata.Minimumbaselineguardrailscanbebasedonintendedminimumstatisticalreliabilitysuchasconfidenceintervalsormeaningfuldeviationsindataandmayvarydependingontolerablerisk(asstipulatedbyusersorasdefinedforaparticularchannelorcategory).Forexample,premiumchannelsdiscussedbelowmaycarryahigherlevelofperceivedsafetyandnecessitatehigherreliabilityandminimumreportingthresholdsduetolowerexpectedadvertiserrisktolerance.Contentnotmeetingbaselineguardrailsmustbeclassifiedasunknownornotyetclassifiedandbuyersshouldhavetheabilitytopreventadservingtounknownorunclassifiedcontent.Collectionspoliciesandfrequenciesaswellasbaselineguardrailsemployedmustbedisclosedtousers.

3.2 CurationofUGCAsdiscussedabove,UGCpresentsnumerouschallengesforcontextualandBrandSafetymeasurementincludingreducedcontrolofcontentwhencomparedtopublishercuratedcontent,whichmayleadtodifferencesbetweenURL/applicationandcontentclassifications.InordertomitigatesomeoftherisksassociatedwithUGC,publishersorplatformsmaycurateUGCbyselectingandvettingspecificpiecesofcontentandgroupingthemintocategories,URLsorchannelswithincommonsubjectmatter,ContextandevenBrandSafetyclassifications(suchasforchildrenorcertainagegroups).Incertainsituations,curationmightbebasedonautomatedfunctionality,butmostlikelyrequiresmanualandhumanreview.Further,suchcurationmightbebasedonpopularityoruseractivitynotnecessarilycorrelatedtoBrandSafety,suchas“premium”pagesorchannelsreservedforcontentcontributorsanduserswiththegreatestnumberoffollowers,subscribers,views,likesorotheractivity.Whilepublishersandplatformsmayconductcurationprimarilyforuserexperience,theyarealsoencouragedtoconsiderdoingsoinordertoassistbuyersandthirdpartymeasurersinmakingContextandBrandSafetydeterminationsandtoexposesuchcurationprocessestoauditingandindependentvalidation.However,whileadverificationorganizationsmayconsiderandleveragethiscurationwhenmeasuringContextandBrandSafety,itshouldnotberelieduponunlessindependentlyauditedandverified.GiventhedynamicnatureofUGCanduseractivityinaThreadedMessage,evencuratedcontentrequiresdiscreteandrobustmeasurementforeffectiveContextandBrandSafetydeterminations.

Page 11: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 11

Further,premiumpagesorchannelscuratedbasedonpopularityoruseractivityalonemaynotnecessarilyrendercontentwithinthemBrandSafeandlikelyrequireenhancedmeasurementfocus,morerobusthumanreviewandperiodicmonitoringduetothefactthatthiscontentmaybemorehighlymonetizedandtrafficked.Whileuseractivitysuchasnumberoffollowers,subscribers,viewsandlikescanbeusedasameanstocontrolmonetizationitshouldnotbesolelyrelieduponforContextorBrandSafetydeterminations.

3.3 ScalabilityAsdiscussedabove,thevolumeanddynamicnatureofcontentonSocialMediaPlatformsandotherlargedigitalenterprisespresentsscalabilityissuesformeasurement.TheabilitytoscalemeasurementatagranularlevelandtoenforcedatafreshnessrulesiscriticaltoeffectivemeasurementofcontentforpurposesofContextandBrandSafety.Adverificationorganizationsmustensuresystemsarecomprehensiveenoughtohandlescaleandperiodicrefreshing.IncompleteorcorruptdatacanleadtoinaccurateorinsufficientContextandBrandSafetydeterminations.Asaresult,inadditiontouseraccess,programchangeanddisasterrecovery/businesscontinuitycontrolsthatarerequiredaspartofMRCaccreditationaudits,adverificationorganizationsshouldalsoemployrobustloadbalancingaswellascapacityanduptimemonitoringprocedurestoensureadequatesystemresourcesareavailablefordatacollectionandbackendprocessingoflargedatavolumes.Further,adverificationorganizationsareencouragedtoutilizebatchprocessingandtoconductperiodicdatacorruptionandlatencycheckstoensuredatacollectioniscompleteandtoidentify,investigateandlimitdataloss.Suchcontrolsshouldbeappliedtosystemsthatinclude,butarenotlimitedto,MachineLearningfunctionalityaswellasingestionandprocessingofMetadata(eachisdiscussedingreaterdetailinthisSupplement).

3.4 MachineLearningThescaleandarrayofdatafieldsinvolvedincontentlevelContextandBrandSafetyfunctionalitylikelynecessitatessomeformofautomationorMachineLearningutilizedbyadverificationorganizationswherebymodelsoralgorithmsareusedtoautomaticallyingest,analyzeandclassifycontentforpurposesofContextandBrandSafety.MachineLearningconsistsofutilizingdatasetsincludingvisualandaudiodatapoints(suchasinvisualandaudiallearning)totrainandevaluatemodelsoralgorithmssuchthattheycanpredictoutcomesonanautomatedbasis.Itiscriticalthatrigorousdataqualityandanalysisproceduresareappliedtotheprocessestoselecttrainingdata,selectparametersusedinthemodel,preparedata,selectthemodel,trainandevaluatethemodelaswellastoperiodicallytuneit.Inaddition,theMRCInvalidTrafficAddendumestablishesrequirementsfordetectionandfiltrationofinvalidtraffic(IVT)andadditionalIVTguidancepromulgatedbytheMRCestablishesbestpracticesforpre-bidapproachestoIVT(whicharelikelyhighlyrelevantinadverificationfunctionality).SuchfiltrationshouldbeappliedtodatasetsusedforMachineLearningtotheextentapplicabletoensureIVTdoesnotintroducebiasesintheclassificationdecisionsor

Page 12: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 12

otherwiseobfuscatedatatrendsthatmaysignalaneedforfurtheranalysissuchasuserviewingandskippingbehaviorthatmaydirectfurtheranalysisofaspecificsegmentoflong-formvideothatmaybedilutedbythepresenceofinvalidtrafficwithnosuchbehavior.Further,thepresenceandvolumeofIVTwithinapropertyorassociatedwithaparticularpieceofcontentmaybeasignalthatinandofitselfassistsinContextandBrandSafetydeterminationsoratleastallowsanorganizationtofurtherfocusresources.TheimpactofIVTonContextandBrandSafetydeterminationsmustbeconsideredwhendesigningMachineLearningfunctionality.WhileIVTshouldbefilteredfromadvertisingmeasurement,thepresence(orlackthereof)ofIVTmaynotdirectlyhaveabearingontheContextandBrandSafetyclassificationofaparticularpieceofcontent,althoughadverificationorganizationsmayconsideralertinguserstopropertieswithhighlevelsofIVTforpotentialavoidance.TheabilityofMachineLearningtoaccuratelypredictoutcomesiscorrelatedtothesizeandqualityofthedataunderlyingit.AdverificationorganizationsmustestablishminimumdatasamplesizesandqualitythresholdswhenselectingtrainingandevaluationdatasetsforMachineLearning.Totheextentreducedperformanceoraccuracyisexpectedincertaindataconditions(suchasinsmallersamplesizes,shortercollectionperiods,missingorlowerqualitydata,etc.),thismustbeactivelydisclosedtousersviaestimatesoferrorusingstatisticalmethodsorobservederror.Further,thedataselectedandtheparametersorfieldsusedshouldberelevanttotheuseofthemodel(empiricalsupportmustexistthatestablishesarelationshipbetweendataorparametersusedandContextorBrandSafetydeterminations).Whenpreparingtrainingandevaluationdata,whichmustconsistofdistinctandmutuallyexclusivedatasets,robustdataqualityproceduresshouldbeappliedtoclean,normalizeanddeduplicatedataaswellastoaccountfororadjustdataimbalancesorbiases.Biasesintrainingandevaluationdatashouldbereducedtotheextentpossibleanddisclosedwherematerial.Additionally,themodeloralgorithmselectedforspecificMachineLearningfunctionalitiesmustbeappropriatefortheintendeduseandperiodicallyupdatedorrefreshed.UseofMachineLearningmodelsmustbelogicalanddefensiblebasedonempiricalsupportthatisdocumentedandmadeavailableforaudit.Suchsupportmustbeperiodicallyvalidatedandupdated.Trainingandevaluationofthemodelwilllikelyresultinweightingorbiasadjustmentaswellasparametertuning,andthismustalsooccuronaperiodicbasis.WhileMachineLearningwillhelpmitigatecertainchallengesdiscussedthroughoutthisSupplementrelatedtoscalabilityanddynamiccontent,itshouldbedeployedinconjunctionwithfocusedHumanIntervention.WeightsandbiasadjustmentsresultingfromtheperiodictrainingandevaluationofMachineLearningmodelsshouldbeusedtofocussuchHumanIntervention.Inotherwords,knownbiasesorweaknessesinthemodelorforspecificdataconditionsshouldinformproceduresthatincludemanualhumanreviewofcontentforContextandBrandSafetyclassifications.AdditionalguidanceforHumanInterventionisdiscussedlaterinthisSupplement.

Page 13: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 13

Machinelearningmethodologyandproceduresmustbeactivelydisclosedtousersatanon-technicallevel,includingrelevantdatasetsizes,modelsusedanddatasourceswhileensuringpropersafeguardsagainstreverseengineeringofBrandSafetytechniquesemployed.Thelevelofrelianceonmachinelearningversushumaninterventionormanualreviewmustbegenerallydisclosed.

3.5 HumanInterventionAsdiscussedabove,MachineLearningshouldbedeployedinconjunctionwithfocusedHumanInterventiondefinedforpurposesofthisSupplementasmanualreviewofcontentbyadverificationorganizationpersonnelforthepurposeofContextandBrandSafetyclassifications.WeightsandbiasadjustmentsresultingfromtheperiodictrainingandevaluationofMachineLearningmodelsshouldbeusedtofocussuchHumanIntervention.EvenwhereMachineLearningisnotutilized,arobustsetofrisk-basedcriteriaisnecessarytodrivefocusedHumanIntervention.Suchrisk-basedcriteriashouldincludeconsiderationsofdataconditionsorcontenttypesthataremoredifficulttomeasureorcategorize,thepresenceofmonetizedUGCanddynamiccontent,andcontentwithhighertrafficincludingpremiumpagesorchannelscuratedbasedonpopularity.ThepresenceandlevelsofIVTshouldalsobeconsideredwhendeterminingthefocusofHumanIntervention.AsHumanInterventioninvolvesmanualreviewofcontent,considerationshouldbegiventotimeconstraintsandscalability.Inconjunctionwithrisk-basedconsiderations,HumanInterventionpoliciesshouldbedesignedwithcontentvolumeandtheneedfortimelyreviewandclassificationinmindinordertoensurestaffinglevelsareadequatetomeetworkloadneedsbasedonreportingtimingpoliciesandvolume.ThepersonnelperformingHumanInterventionmustbeadequatelytrainedandsupervised.TheresultsofHumanInterventionmustbeperiodicallyreviewedandformallydocumented.Allpersonnel(includingsupervisors)mustbefurnishedwithdetailedinstructionsandmanualscoveringallstepsoftheirwork.PersonnelperformingHumanInterventionmustbeperiodicallyassessedforperformance.Lowerperformingpersonnelshouldbere-trainedandtheirworkinproductionenvironmentsshouldbelimited.TheresultsofHumanInterventionshouldbeusedtoperiodicallyvalidatetheresultsofMachineLearningandtoupdateevaluationdatasets.ItisexpectedthatHumanInterventionbeusedasanongoingandcontinuousqualitycontrol,butthatresultsareusedtoupdateMachineLearningasfrequentlyasisfeasibleandatleastseveraltimesannually.

3.6 MonetizationControlsPublishersorSocialMediaplatformsmayapplymonetizationcontrols,orthresholdsabovewhichcontentwillbemonetized(andbelowwhichitwillnotbe).Thesethresholdscanbebasedontheageorrecencyofcontentandtheactivitysurroundingitsuchasinteractions,views,subscribersorviewtime.SuchcontrolsmayhelpmitigatetheriskofincompleteorinaccurateContextorBrandSafetyclassificationsinthattheyextendthetimebeforewhich

Page 14: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 14

contentismonetizedtoallowmorethoroughreviewandalargeraccumulationofdatatoanalyze.Publishersandplatformsareencouragedtoutilizesuchcontrolsandtodisclosethemtousers.Further,likeHumanInterventiondiscussedabove,thesemonetizationcontrolsshouldbedrivenbyarobustsetofrisk-basedcriteria.Additionally,publishers,adnetworks,exchangesorSocialMediaPlatformsshouldconsiderestablishingpoliciestolimitorpreventmonetizationofcontentforsourcesthathavebeenfoundtoviolatecontentpolicies,specificcontentthathasbeendeterminedtobeinviolationofcontrolsorotherBrandSafetycontrolsandthesepoliciesshouldbeappliedataplatformlevel.Thesepoliciesifpresent,must,beclearlydefinedanddisclosedtousersandsourcesthatcancontributecontent.InaccordancewiththeIAB/MRCAdVerificationGuidelines,adverificationprocessesmustbetransparenttocustomersandsellerorganizations,inthatbothknowthegeneralcriteriabeingusedforevaluationaswellasanycontentorsourcesofcontentsuchasusersorpublishersthatareexcluded(detailsofspecificproprietarymethodsmaybeomitted).ThepresenceandlevelsofIVTshouldalsobeconsideredwhendeterminingwhichparameterstouseformonetizationcontrolsandshouldbeutilizedonapost-filtrationbasis.Theabilityforaspecificparametertobemanipulatedthroughinvalidtrafficshoulddecreaserelianceonitforpurposesofmonetizationcontrols.Likecurationofcontentdiscussedabove,publishersandplatformsareencouragedtoexposemonetizationcontrolstoauditingandindependentvalidation.Again,whileadverificationorganizationsmayconsiderandleveragemonetizationcontrolswhenmeasuringContextandBrandSafety,theyshouldnotberelieduponunlessindependentlyauditedandverified.Finally,publishers,platformsandadverificationorganizationsareencouragedtoenablebuyerstocustomizemonetizationcontrolsbasedoncontextualorBrandSafetyneeds.EnhancedbuyertoolsarediscussedlaterinthisSupplement.Whileuseractivitysuchasnumberoffollowers,subscribers,viewsandlikescanbeusedasameanstocontrolmonetization,itcannotbesolelyrelieduponforContextorBrandSafetydeterminations.

3.7 TaxonomyandCategorizationTheIABNetworkandExchanges“ContextualTaxonomy”(nowtheIABTechLabContenttaxonomy)wereincorporatedintotheAdVerificationGuidelinesbyreferenceasalistingofappropriatecontentcategoriesforsites(althoughrecentupdatestothistaxonomyhavedeprecatedsomepredecessoravoidancecategories).Additionally,theIAB/MRCAdVerificationGuidelinesincludealistingofgeneraltypesofpotentialavoidancecategoriesanditisexpectedthatadverificationorganizationsmaymaintainmoredetailedsub-categoriesforeachofthesepotentialavoidancecategories,whichwillhelpfurtherrefineanddifferentiatetheirservice.WhileadverificationorganizationsareencouragedtousecommonIndustrysourcesforgenreclassificationandsegmentation,theymayalsouseinternallydevelopedorproprietarysources

Page 15: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 15

ifempiricallysupportedwithauditableevidencewithcleardisclosure.Thesourceandtaxonomyusedingenreclassification(includingcustomproprietarysources)forpurposesofContextandBrandSafetyclassificationsmustbedisclosedtousersandperiodicallyupdated.Suchdisclosuresshouldconsiderprotectionofintellectualpropertyandtheriskofreverseengineeringthatcouldpotentiallyresult.Further,dynamicUGConSocialMediaPlatformsmayrequiregranularanddiscreteavoidancecategoriesandpresentsfurtherchallengesforaccurateclassification,butthenatureofthiscontentmayalsointroducesubjectivityregardingtheclassificationofaparticularpieceofcontent.Forexample,videocontentmaycontainsegmentswithdifferingcategorizationsandadverificationorganizationsareencouragedtoprovidegranularsegmentcategorizationdatainadditiontotop-levelclassificationswherethismayoccur,suchasinlong-formdigitalvideo(especiallyforepisodicandrepurposedTVcontentwhereexistingcategorizationpracticesmayalreadyexist).ThisSupplementdoesnotaimtoproscribespecifictaxonomiesorcategoriesbeyondthosereferencedabove,butinsteadtohighlighttheuniquecategorizationchallengesUGConSocialMediaPlatformspresents,andtorequireadverificationorganizationstodisclosethemethodologyanddecision-makinghierarchyunderlyingcategorizationstousers.Additionally,totheextentthatcustomcriteria,partialormulticategorizationforaspecificpieceofcontent,ordegreesofcategorization(suchascontentratingsorscores)areused,theymustalsobeactivelydisclosedtousersaswellasempiricallysupportedandperiodicallyupdated.MRCstronglyencouragesandsupportsmechanismsandeffortstopromoteconsistentBrandSafetydefinitionsandminimumbaselinessuchasthatproposedbythe4A’sAdvertisingProtectionBureau(APB)aswellasotherbuyerorganizations.ThisSupplementrecommendsestablishingaminimumlevelofcontentestablishedandmaintainedbyindustrygroupsthatexcludescertainbaselinecategories.Thisminimumcanbeofferedasacollectivereportingorblockinggroupconsideredunsafeorineligibleformonetization(“thefloor”)eitherasadefaultcontrolforplatformmonetizationoraselectedbyadvertisers.ThisfloorshouldbebasedonexclusionofcertaincategoriesrootedintheIndustry’s“dirtydozen”categorylist.Specifically,the4A’sAdvertisingAssuranceBrandSafetyFloorFrameworkissuedSeptember2018includesbaselinedefinitionsofafloorwithin13categories.Thesecategorieslikelyrequireongoingconsiderationanddefinitionandthecategorizationofapieceofcontent(forexample,whatconstitutesanobscenity)maydifferfromthepointofviewofdifferentadverificationusers.The4A’sAdvertisingAssuranceBrandSafetyFloorFrameworkaimstocodifyaspectsoftheabovethatarelesssubjectiveandcanbeagreedtobepartofaconsistentexclusioncategorytoestablishthefloor.Theexpectationisthatplatformsandpublisherswillenactcontrolstominimizethepresenceofcontentinthesecategories.Wherecontentinthesecategoriesispresentadvertisersmaychoosetoforgotheagreeduponfloorandallowadvertisingtooneoralloftheabovecategories.

Page 16: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 16

Further,the4A’sAPBwillincludeongoingeffortstoreviewexistingtaxonomiesanddeterminehowconsistencycanbecreated.Thisincludesreconcilingdefinitionsofcommoncategorydifferencesbetweenvendors,whileallowingforcustomsettingsthatvendorsapplytoaddressspecificclientsBrandSuitabilityneeds(suchasviatheAPB’sBrandSuitabilityFramework).MRCfurtherrecommendsandsupportsIndustryeffortstostandardizeandmodernizetaxonomiesandcategoriesacrossplatformsconsideringUGCandcontentgranularity,andweintendtoparticipateinsucheffortsonanongoingbasis.

4 AdjacencyForthepurposeofthisSupplement,AdjacencyisdefinedasthedegreeofphysicaldistanceinwhichadsareplacedinrelationtospecificcontentthatisclassifiedforContextandBrandSafetypurposeswithinadomain,site,platform,mobileapplicationorpage.Asdiscussedabove,historicallyBrandSafetyandtargetinghavebeenaddressedatapropertylevelwithContextclassificationandadalertingorblockingbasedonavoidanceortargetcategories.However,thenatureofSocialMediaPlatformsandvariousnewsfeedenvironmentsrequiremorediscreteproximitymeasurementforbothtargetingandavoidanceincludingadjacencyinnewsfeed,continuousscrollorswipe(verticalorhorizontal)orcontinuousplayenvironments.WhilemeasurementofAdjacencyisnotrequired,itisencouragedandanaccreditableaspectofcontentlevelContextandBrandSafetyfunctionality.Thefollowingguidancemustbeappliedtoadverificationorganizationselectingtoofferadjacencymeasurement.Adjacencyandmeasurementofitwilldifferbetweenplatforms,URLsandmobileapplicationsaswillauser’sdeterminationofwhetheraspecificpieceofcontentisnearorfarfromanadplacement.Tothatend,publishersandplatformsareencouragedtoestablishmeasurementunitsindeterminingadjacencyattheplatform,URLormobileapplicationlevel,whereapplicablethatcanbemeasured,reportedandverifiedbythirdparties.Forexample,themeasurementunitwithinaverticalorhorizontalnewsfeedenvironmentcouldbeadiscretepane,tile,screenorotherunit.Themeasurementunitshouldreflectthemeaningfulsegmentationofcontentwithinaplatform,pageormobileapplicationandmustbeclearlydisclosedaspartofadjacencyreporting,anddocumentedandsupported.Asabaseline,adjacencymeasurementunitsshouldbediscretepiecesofcontent(suchasanimage,video,song,adetc.)andallmeasurementbasesmustbemeaningfultocontentandadconsumptionandempiricallysupported.Adverificationorganizationsareencouragedtoconsidertheintersectionoftimeandadjacencyviatemporalthresholdsappliedtoadjacency.Theimpactofadjacentcontentmaydiminishduringlongerusersessionssuchasinsituationswhenusersleavecontentandreturntoitafterthepassageoftime.Adjacencyshouldbemeasuredandreportedonthebasisofmeasurementunitdistanceandmustaccountforscrollingandswipingcapabilitiesinmobilenewsfeedenvironments.Adjacencycouldbereportedintermsofabsolutevalueofmeasurementunits(i.e.,numberofunitsplacementproceeds/isaboveorfollows/isbelowspecificcontentbeingmeasured).The

Page 17: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 17

basisforreportingmustbeclearlydisclosed.Placementsthatarezeromeasurementunitsfromspecificcontentbeingmeasuredorthatareabletobesimultaneouslyviewable(50%ofpixelsonscreenconsistentwithMRCViewabilityGuidelines)withspecificcontentbeingmeasuredshouldbeconsidereddirectlyadjacent.Otherwise,adjacencyrangesorscoresmaybereportedandmustbeclearlydefined.Invideoplaylistenvironments,adjacencyshouldbemeasuredintermsofthenumberofvideounitsthatplaybeforeandafterthespecificcontentbeingmeasuredincludingcontentthatautomaticallybeginsplayingwithoutadditionaluserinteraction.Considerationmustalsobegiventothepresenceandcategorizationofthumbnailimagesaccompanyingorrelatedtovideocontent,especiallywheretheseimagesmaybemisleadingordifferfromthenatureoftheunderlyingvideocontent.AsdiscussedearlierinthisSupplement,publishersorSocialMediaplatformsmayapplymonetizationcontrolsandthisshouldincludeadjacencyconsiderations(aminimumdistanceornumberofmeasurementunitsbelowwhichadswillnotbeplaced).Suchminimumadjacencyrulesmayvarybycontent.Again,publishersandplatformsareencouragedtoutilizesuchcontrolsandtodisclosethemtousers,butwhileadverificationorganizationsmayconsiderandleveragemonetizationcontrolswhenmeasuringContextandBrandSafety,theymustnotberelieduponunlessindependentlyauditedandverified.Thirdpartyadverificationvendorsmayhavetorelyonsignalsrelatedtoadjacency(suchasunits,scoresofplacementsorsession-levelcontentinformation)fromplatformsorpublisherswherethethird-partyisrestrictedfromdirectlymeasuringadjacencyeitherduetomeasurementpoliciesortechnicalchallenges.Further,inthesesituations,onlytheadserver,oftentimestheplatformitself,canallowblocking,coveringorotherwiseavoidancebasedonadjacency.Asaresult,platformsandpublishersarestronglyencouragedtoeitherallowdirectthirdpartymeasurementatacontentlevelinclusiveofadjacencysignals,ortoactivelyprovidevalidatedfirst-partydataatagranularleveltoenableindependentContextandBrandSafetymeasurementalertingandreporting,oneofwhichwillberequiredforcontentleveladjacencyaccreditationofthird-partyadverificationsolutions.Further,thecollectionandtransmissionofdataenablingadjacencydeterminationsisstronglyencouragedtobesubjectedtoauditandindependentvalidation.Publishers,platformsandadverificationorganizationsareencouragedtoenablebuyerstocustomizeadjacencybasedoncontextualorBrandSafetyneeds.EnhancedbuyertoolsarediscussedlaterinthisSupplement.

5 MetadataVariousadverificationtechniquesandprocessesdiscussedthroughoutthisSupplementmayutilizepublisherorplatformMetadata,ordatathatprovidesinformationaboutthemakeupofadomain,site,URLorapplicationincludingspecificcontentdetails.Metadatatypicallyincludesdescriptivedatasuchasageneraldescription,authorandkeywords,structuraldatathat

Page 18: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 18

includespageandcontentorganization,andadministrativedatasuchaswhenafilewascreated,bywhom,contenttypeandothertechnicalinformation.Thismayincludecontentsizeandlength,appearancecharacteristics(suchascolorandshape),transcribedaudioforvideocontentandassociateduserinteractionattributessuchascomments,views,exposureandengagementdataornumberofsubscribers.Metadatashouldalsoincludedescriptorsofthumbnailimagesaccompanyingorrelatedvideocontent,especiallywheretheseimagesmaybemisleadingdifferfromthenatureoftheunderlyingvideocontent.AsdiscussedearlierinthisSupplement,platformsandpublishersarestronglyencouragedtoeitherallowdirectthirdpartymeasurementatacontentlevel(inclusiveofMetadata),ortoactivelyprovidevalidatedfirst-partydataatagranularleveltoenableindependentContextandBrandSafetymeasurement,oneofwhichwillberequiredforaccreditationofcontentlevelthird-partyadverificationsolutions.ThecollectionandtransmissionofMetadataenablingcontentlevelContextandBrandSafetydeterminationsisstronglyencouragedtobesubjectedtoauditandindependentvalidation.AsdiscussedintheMachineLearningsectionofthisSupplement,certainMetadataparametersmaybemoredirectlycorrelatedwithContextandBrandSafetydeterminationsthanothersandthiscorrelationmayvaryovertimeorwiththesizeoftheobserveddataset.InadditiontoguidanceaboverelatedtoscalabilityandMachineLearning(whichshouldbeappliedtoMetadata),adverificationorganizationsutilizingMetadatamustconsiderlimitationsofthisdatawhenusingittomakeContextandBrandSafetydeterminations.Further,certainMetadataisstaticandnotsubjecttochange,whileotherMetadata,suchasthatcorrespondingtouserinteractionandexposuredata(whichmaybeusedformonetizationcontrols)isdynamic.AsdiscussedintheAdVerificationGuidelinesandabove,freshdataisabsolutelynecessaryforaccurateadverification.DatafreshnessrulesarecriticaltoeffectivemeasurementofcontentforpurposesofContextandBrandSafety.MetadatausedforContextandBrandSafetydeterminationsmayincludeuserreportingorflaggingofobjectionablecontent.Thisparametercanbeavaluableinputintoadverificationfunctionality,butduetothesubjectivenatureofsuchreporting,shouldbeappropriatelyweightedinthedecisionmakingprocessbasedonreviewandrobustqualitycontrol.AutomatedanalysisofMetadataalonewithoutadditionaldatasourcesorconsiderationofhumanandmanualprocessesislikelyinsufficientforcertaincontentlevelContextandBrandSafetydeterminations(suchasvideocontent)andadverificationorganizationsareencouragedtoutilizeotherdataandtechniquesinadditiontoMetadata.ThedegreeofrelianceonMetadatainContextandBrandSafetydeterminationsbyanadverificationorganizationmustbedisclosedtousers.Suchdisclosuresshouldconsiderprotectionofintellectualpropertyandtheriskofreverseengineeringthatcouldpotentiallyresult.Use-casesandexamplesoftoolsusedareencouragedtobeapartoftechnicaldisclosuresmadebytheadverificationorganization.ThespecificMetadatacollectionmethodsandcontentrecognition

Page 19: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 19

processesusedbytheadverificationorganizationmustbedisclosedatahighlevelandmayinclude,butarenotlimitedto:

• ApplicationProgrammingInterfaces(APIs)• DirectMeasurement• MachineLearning• OfflineExport• PartnerData• Scraping/Crawling(wherepermissiblebytermsandconditions)• Third-PartyTracking• MetaTags

AsdiscussedintheAdVerificationGuidelines,adverificationservicesmaybeintegratedintoadcampaignsusingsomeformoftrackingasset(adtags,beacons,etc.).Additionally,passive(non-integrated)analysistoolssuchasautomatedcrawlersorspidersmaybeused.Thelimitationsassociatedwiththesemethodsmustbehighlightedindisclosures–forexample,thenon-randomnatureofobservations,non-censusapproaches,ortheimpactontheabilitytoprojectresults.Datacollectionmethodsmustbesubjecttointernaltestingandvalidationuponoriginalimplementationaswellasperiodicinternalverificationtesting.ThistestingshouldincludethevalidityoftheContextandBrandSafetymeasures.DatacollectionmethodsanddisclosureswillbethesubjectofintenseproceduralverificationifMRCaccreditationissoughtbythemeasurementorganization.TheMetadatasourcemustbeclearlydefinedandtheextentofMetadatausemustbeincludedintheadverificationorganization’sdisclosures,forexample:

• Whattypeofcontentiscapturedfromwhatplatform;• Usercoverageformeasuredplatforms;• Collectionmethod(direct,API,offlineexport,partner,third-party,scraping/crawling,

etc.);• APItypewhereapplicable(public,closedclient,platformorfirehose);• Depthofcontentcollection(scraping)applied;• Storagemethods,timing;• Natureofmachine(orhuman)analysisprocessesemployed;• Timeperiodsanalyzed;• Editsordataadjustmentsapplied;• Natureofspecialhandlingofambiguouscases;• Errorcorrectionandreissueprocedures;and• ContextandBrandSafetyaccuracy.

Page 20: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 20

Differentlanguagesandlanguage-relatednuancesshouldbeconsideredwhenanalyzingmetadataincludingdialectsandprocessestoingestandanalyzemetadatashouldbedesignedtoaccountfordifferentormultiplelanguagespresentandwhereapplicable.

6 MobileApplicationsWhiletheguidanceprovidedintheAdVerificationGuidelinesandwithinthisSupplementshouldbefullyappliedwithinmobileapplications,theremaybeuniqueContextandBrandSafetyconsiderationswithintheseenvironments.Therisksandtechniquesrelevanttomobilein-appenvironmentsrelatedtoContextandBrandSafetyexhibitdifferentialcharacteristicswhencomparedtothoseemployedindesktopormobilewebenvironments.Moreover,measurementassetsutilizedindesktopormobilewebenvironmentssuchascrawlers,JavaScript,cookiesorFlashmaynotbeavailableorfunctionalwithinmobileapplications.InadditiontospecificconsiderationsdetailedintheIAB/MMA/MRCMobileApplicationAdvertisingMeasurementGuidelines,adverificationorganizationsmeasuringContextandBrandSafetyshouldapplyincrementalanddifferentialconsiderationtomobilein-application(in-app)environmentsandtheymustdosofordistinctmobilein-appcontentlevelBrandSafetyaccreditation.Suchconsiderationmustbefocusedonmeaningfuldifferenceswithinmobileapplicationsregardinginitialandongoingriskassessments,userbehavior,differencesincontent,heuristicsandsignalsusedtomakeContextandBrandSafetydeterminationsandongoingdataanalysis/benchmarking.Inadditiontoadverificationfunctionalityatanapplicationlevel,adverificationorganizationsseekingaccreditationformobileapplicationsmustapplycontentlevelgranularity(seeguidanceearlierintheSupplement)discretelywithinmobileapplicationsandnotbasedsolelyondescriptionsofanapplication,applicationstoreinformationortraffictocommonlyownedwebproperties.Whileclassificationoravoidanceofapplicationsbasedontheirdescriptionsorgeneralcontentarepermissible(applevelakintopropertylevel)withcleardisclosure,thiswouldnotbeconsideredcontentlevelin-applicationmeasurementgranularityforpurposesofthisSupplement.MRCintendstoconsiderBrandSafetyaccreditationdiscretelybetweendesktop/mobilewebandmobilein-appenvironments.MRCrecognizesthatdiscretecontentlevelContextandBrandSafetyfunctionalitywithinmobileapplicationsmaybeaspirationalasmanyapplicationsdonotsupportAPIfunctionalityandwhileitisnotrequiredthatadverificationorganizationsdevelopmobilein-appcapabilitiesinthisarea,itishighlyencouragedandrequiredinorderforMRCtograntspecificcontentlevelin-appaccreditationforthisfunctionality.Alternatively,applicationswithoutAPIsupportenablingcontentlevelmeasurementgranularitymaybeclassifiedas“unknown”forBrandSafetypurposes.TheMRCbelievestheremaybevalueinthedevelopmentofacommonandopenstandardAPIorSDKbytheindustrythatcanbeusedbyallpartiestoenablein-appContextandBrandSafetymeasurementviaadverificationorganizations.Weencouragesuchdevelopmentand

Page 21: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 21

wouldsupportandparticipateinanopenandsinglesourcetechnicalstandard.Itshouldbenotedthatasofthedateoftheissuanceofthisdocument,theMRCisnotawareofanyactiveinitiativestoaddressacommonindustryAPIorSDKandassuch,itislikelythatcontentleveladverificationmeasurementsolutionsinapplicationswillbecustomorspecifictocertainadverificationorganization’splatformandpublisherintegrationsfortheforeseeablefuture.

7 IllegalSourcesandIndustryCommunicationAdverificationorganizationsmustmakereasonableattemptstoconsiderlegalrestrictionsrelatedtoprohibitionofcontentandthesourcesofthatcontenttopreventmonetizationofillegalcontent.Specifically,certaintypesofcontentorsourcesofcontentmaybeexpresslyprohibitedbylegalrestrictionssuchascontentpiracyinvolvingcopyrightviolationsandcounterfeiting,ortheremaybeotherrequirementstoremoveillegalcontentwithinaspecifiedtimeperiod(suchasGermany’sNetworkEnforcementAct).Publishersandplatformshavearesponsibilityforsafeguardingtheirpropertiesfromillegalentitiesandsourcesandshouldhaveaqualificationprocesstomakesuretheyaredealingwithalegitimateentityandlegallypermissiblecontentthatisappliedacrosstheplatform,aswellaspoliciestocomplywithrequiredenforcementperiods,evenpriortoapplyingadditionalBrandSafetycontrols.Whilenon-publisherthird-partymeasurementorganizationsaregenerallynotincontrolofcontentsource,theymustbeawareoflegalrestrictionsandincludeconsiderationsofthisinContentandBrandSafetymeasurementanddetermination.ProceduresrelatedtodeterminingthelegalityofsourcesandcontentshouldincludeinitialqualificationusingIndustryandlocalsourcesofknownillegalentities,aswellasongoingevaluationlinkedwithadverificationresultsandperiodicinternalauditingofcontentsources.Adverificationorganizationsseekingaccreditationwillberequiredtoprovideevidenceofsourcevettingprocesseswhereapplicableduringaccreditationauditprocesses.Further,totheextentthatadverificationorganizationshaveidentifiedillegalorillegitimatesourcesthatareeithernotincludedinIndustryorlocalsourcesofknownillegalentities,orthataredisguisedasotherlegalandlegitimateentities,processesshouldbeputinplacetoroutinelycommunicatethesesourcestolegalauthorities,oversightbodiesandtheIndustryatlarge.Suchcommunicationsshouldbeconductedwithsensitivitytoprotectionofintellectualpropertyandtheriskofreverseengineeringthatcouldpotentiallyresult.Decisionstoforgocommunicationbyadverificationorganizationsmustbesupportedbyauditableevidenceofsuchrisk.MRCstronglyencouragesandsupportsmechanismsandeffortstocommunicaterelevantBrandSafetyinformationsuchasobservedmonetizationof“unsafe”contentacrosstheIndustryandpubliclysuchasthatcurrentlyproposedwithinthe4A’sAdvertisingProtectionBureau(APB).ThismayincludeopenandtransparentlistingofContextandBrandSafetycategorizationsforpropertiesaswellascontentURLsviaemergingtechnologicalmeanssuchasblockchain.

Page 22: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 22

8 EnhancedBuyerToolsThisSupplementdetailsvarioustoolsverificationorganizationsmayutilizetoclassifyandcategorizecontentforcontextandBrandSafetydeterminations,aswellasproactiveapproachesthatmaybetakenbypublishersorplatformstorestrictadplacementwiththeultimategoalofmeetinganadvertiser’sspecificationsonprotectingitsbrand.Theseapproachesincludecuration,monetizationcontrolsandgranularcategorization.However,thenatureofdynamicUGConSocialMediaPlatformsmayintroduceadditionalsubjectivityregardingtheclassificationofaparticularpieceofcontent.Additionally,differentbuyersmaynotonlyhavedifferenttargetsandavoidancecategories,butalsodifferenttolerancelevelsforriskdependingonbrandsuitability.Whileaminimumbaselineofmonetizationcontrolisencouraged,ultimatelyadverificationorganizationsseekingcontentlevelBrandSafetyaccreditationmustprovideconsistentanddiscretereportingtousersinordertoenablecustomizedtargetingandavoidance.Beyondcurrentcategorybasedblockingandalerting,adverificationorganizationsmayenable:adprevention(pre-bid);blockingandalerting(post-campaign)onthebasisofMetadata;discretecontentcharacteristicsincludingtheageofcontentsubscriberstoachannel,thenumberofviewsandinteractionstoenablecustommonetizationcontrolsorflagging/reportingrelatedtothecontent,adjacency,etc.;andadjacencyparameterstoallowuserstostipulatewhichlevelsmeettheneedsofaparticularcampaignorbrand.Thebasisofmeasurementunderlyingtheseenhancedbuyertoolsmustbeclearlydisclosed,alongwiththemethodologyusedtocollect,process,editandreportsuchmeasurement.Further,anylimitationsintheunderlyingmeasurementorknownbiasesmustalsobeclearlydisclosedtoenableuserstoutilizethesetoolsinthepropercontext.Adverificationorganizationsshouldestablishempiricallysupportedandclearlydisclosedminimumrequirementsforreportingwiththeselimitationsandbiasesinmind.

9 RequiredReportingResultsofadverification(includingContextandBrandSafety)shouldbereportedeitheronthebasisoftheamountoftrafficthatisblockedorclassifiedinrelevantcategories(insituationswhentheadverificationfunctionisoutsidetheadchain)oronthebasisoftrafficnetofContextandavoidancecategoriesoutsideofthosestipulatedbytheuser(whenpartoftheadservingfunctionality).Reportingcanbeattheimpression(orotherrelevantmetric)level,withincampaignsbutmayalsoconsistofinventoryorcontentclassification.Reportingmayinclude;(1)whetherresultsareprojectabletoacampaignornot,dependingonverificationdatacollectionmethod;(2)thereliabilityorstandarderroraroundprojectableresultswheresamplesareused;and(3)likelihoodoffalsepositiveresultsbyservicelinebasedonthepriorexperienceoftheverificationservice.Additionally,knowntechnicallimitationsofverificationservices,ifany,mustbeexplained.Beyondearlyconfirmationandset-uperrordetectionprocedures,performancecommunicationtoadverificationserviceusersmustbeongoingviaperiodicreportsorthroughanautomatedreportingdashboard.

Page 23: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 23

Itispermissibleforadverificationorganizationswithcontentmeasurementgranularitytoreportavoidanceandverificationatthepropertylevelwithproperdisclosure.SuchpropertylevelreportingwithcontentlevelmeasurementgranularitywillbedistinguishedbyMRCfrompropertylevelreportingwithoutdiscretecorrespondingcontentmeasurement.TotheextentthatathirdpartyvendorisnotabletodirectlymeasurecertainaspectsofcontentwithinapropertyforContextandBrandSafety,thevendormaylistthepropertyasuncategorizedorunknownwithcleardisclosureofcapabilitiesorlimitationswithrespecttospecificpropertyorcontentlevelgranularitytoenableavoidanceifausersochooses.Further,inthesesituations,propertiesthatallowUGCmaybegenericallyclassifiedassuch(andbedistinguishedasunknownwithregardtoBrandSafety)toenableausertochoosewhetherornottoplaceadsintheseenvironmentsincaseswherecontentleveldecisionscannotbemade.However,adverificationorganizationswillnotbeconsideredtooffercontentlevelContextandBrandSafetygranularityforpropertiesthatarelabeledgenericallyoruncategorizedwherethesemeasurementlimitationsexist.Finally,itispermissibleforadverificationorganizationstoclassifycontentorpropertiesbasedonURLkeywordswithoutcontentlevelmeasurement,howeverthismustbeclearlydisclosedanddoesnotrepresentcontentlevelmeasurementgranularityorreporting.Further,suchURLlevelreportingislikelyunreliableandlimitedforBrandSafetypurposes.

10 UseofSamplesVarioustechniquesandcontrolsarediscussedthroughoutthisSupplement,andadverificationorganizationsareencouragedtousemorethanoneofthesetechniquesandcontrolsinconjunctionwitheachotherinordertoachieveahighdegreeofconfidenceinContextandBrandSafetydeterminations.Adverificationorganizationsareencouragedtoadoptcompletecensusmeasurementusingdeterministictechniqueswherepossible.However,wheresamplesareusedinmakingContextandBrandSafetydeterminations(suchaswhenusingprobabilisticorpre-bidapproaches),suchsamplesmustbeconstructedwithaconfidencelevelof99.7%(within3standarddeviations).ConfidencelevelsusedandstandarderrorsurroundingContextandBrandSafetydeterminationsmustbeactivelydisclosedtousers.Forsample-basedmeasurementofanykind,theadverificationorganizationmustbediligentaboutensuringvalidprojectionsaremadeandthatthesampleisrepresentativeofthepopulationtargetedformeasurement.Methodsforweightingoradjustingdatatoensureprojectabilitymustbesupportedbyempiricalstudy,andtheseempiricalstudiesmustbeupdatedperiodically.Standarderrorsaroundsample-basedprojectionsmustbedisclosed.

11 GlossaryofTermsAdjacency:ThedegreeofphysicaldistanceinwhichadsareplacedinrelationtospecificcontentthatisclassifiedforContextandBrandSafetypurposeswithinadomain,site,platform,mobileapplicationorpage.

Page 24: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 24

AdVerificationOrganization:ForthepurposeofthisSupplement,referstoanyorganization,beitathirdpartymeasurementvendor,apublisheroraSocialMediaPlatform,providingAdVerificationservices(measurementand/orreporting)includingContextandBrandSafety.BrandSafety:PracticesandtoolstoensurethatadigitaladwillnotappearadjacenttoorinaContextthatcandamageanadvertiser’sbrand.Context:ReferstocertainattributesofaURL,site,mobileapplicationorpieceofcontentthatareusedasdeterminantsastowhetherthedestinationwhereadigitaladistobeservedrepresentsanappropriateenvironmentinwhichtheadshouldappear,asdeterminedbyparameterssetbytheadvertiser.Curation:Selectingandvettingspecificpiecesofcontentandgroupingthemintocategories,pagesorchannelswithincommonsubjectmatter,ContextandevenBrandSafetyclassifications(suchasforchildrenorcertainagegroups).Incertainsituations,curationmightbebasedonautomatedfunctionality,butmostlikelyrequiresmanualandhumanreview.HumanIntervention:ManualreviewofcontentbyadverificationorganizationpersonnelforthepurposeofContextandBrandSafetyclassificationsMachineLearning:Useofmodelsoralgorithmstoautomaticallyingest,analyzeandclassifycontentforpurposesofContextandBrandSafety.Metadata:Datathatprovidesinformationaboutthemakeupofadomain,site,URLorapplicationincludingspecificcontentdetails.Metadatatypicallyincludesdescriptivedatasuchasageneraldescription,authorandkeywords,structuraldatathatincludespageandcontentorganizationandadministrativedatasuchaswhenafilewascreated,bywhom,contenttypeandothertechnicalinformation.Thismayincludecontentsizeandlength,appearancecharacteristics(suchascolorandshape),transcribedaudioforvideocontentandassociateduserinteractionattributes(suchascomments,views,exposureandengagementdataorsubscribership).SocialMediaPlatform:Amediavehicle,site(asawholeorinpart),app,widgetorothermediamechanismthathasasocialorientation;specifically,thatcapturesusercommentaryandfacilitatessharingofinformationamongusersofadefinednetwork–i.e.,representingachannelforsharinginformation,opinionsorexperiencesfromusersofthatplatform–orallowscross-usercollaboration.ThreadedMessage:Asetofcommunications,entriesoreventsthatarerelatedtoacommonpieceoforiginalcontent.UserGeneratedContent(UGC):Contentthatisentered,copied-to,posted(orotherwisecreated)byusersofaSocialMediaPlatformforsharingwithothersonthatPlatform.

Page 25: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 25

12 ReferencesReferenceListofRelevantPreviouslyReleasedMRCStandardsIndustryGuidelines:MRCMinimumStandardsforMediaRatingResearch:http://mediaratingcouncil.org/MRC%20Minimum%20Standards%20-%20December%202011.pdfMRCInvalidTrafficDetectionandFiltrationGuidelinesAddendum:http://mediaratingcouncil.org/101515_IVT%20Addendum%20FINAL%20(Version%201.0).pdfMRCSocialMediaMeasurementGuidelines:http://mediaratingcouncil.org/MRC%20Social%20Measurement%20Guidelines%20v1.0%20Final.pdfMRCDesktopandMobileViewabilityGuidelines:http://mediaratingcouncil.org/081815%20Viewable%20Ad%20Impression%20Guideline_v2.0_Final.pdfhttp://mediaratingcouncil.org/062816%20Mobile%20Viewable%20Guidelines%20Final.pdfIAB/MRCGuidelinesfortheConductofAdVerification:https://www.iab.com/wp-content/uploads/2015/06/Ad-Verification-Guideline-for-the-Conduct-of.pdfIAB/MMA/MRCMobileApplicationAdvertisingMeasurementGuidelines:https://www.iab.com/wp-content/uploads/2017/11/Mobile-In-App-Measurement-Guidelines-MMTF-Final-v1.1.pdf

13 SupportingAssociationsandParticipatingOrganizationsAbouttheMediaRatingCouncil(MRC)TheMediaRatingCouncilisanon-profitIndustryassociationestablishedin1963comprisedofleadingtelevision,radio,printanddigitalmediacompanies,aswellasadvertisers,advertisingagenciesandtradeassociations,whosegoalistoensuremeasurementservicesthatarevalid,reliableandeffective.MeasurementservicesdesiringMRCaccreditationarerequiredtodisclosetotheircustomersallmethodologicalaspectsoftheirservice;complywiththeMRCMinimumStandardsforMediaRatingResearchaswellasotherapplicableIndustrymeasurementguidelines;andsubmittoMRC-designedauditstoauthenticateandilluminatetheirprocedures.Inaddition,theMRCmembershipactivelypursuesresearchissuestheyconsiderprioritiesinanefforttoimprovethequalityofresearchinthemarketplace.Currentlyapproximately110researchproductsareauditedbytheMRC.AdditionalinformationaboutMRCcanbefoundatwww.mediaratingcouncil.org

Page 26: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 26

AbouttheAmericanAssociationofAdvertisingAgencies(4A’s)Foundedin1917,the4A’swasestablishedtopromote,advanceanddefendtheinterestsofourmemberagencies,theiremployeesandtheindustryatlarge.Theorganizationserves700+memberagenciesacross1,300offices,whichcontrolmorethan85%oftotalU.S.advertisingspend.Astheleadingtradeorganizationformarketingcommunicationagencies,the4A’spurposeistohelpempoweritsmemberstodeliverinsightfulcreativitythatdrivescommerce,andinfluencescultureallwhilemovingtheindustryforward.Theorganizationprovidescommunity,leadership,advocacy,guidanceandbest-in-classtrainingthathelpenableagenciestoinnovate,evolveandgrow.4A’sBenefitsdivisioninsuresmorethan160,000employeesanditsD.C.officeadvocatesforpoliciesthatbestsupportathrivingadvertisingindustry.The4A’sFoundationfuelsarobustdiversitypipelineoftalentforitsmembersandthemarketingandmediaindustry,fosteringthenextgenerationofleaders.Theorganizationisdedicatedto,andvestedin,ourmembers’successjustastheyarededicatedtohelpingbrandscreate,distribute,andmeasureeffectiveandinsightfuladvertisingandmarketing.Visitthe4A’sathttp://www.aaaa.org.AbouttheAssociationofNationalAdvertisers(ANA)TheANA(AssociationofNationalAdvertisers)makesadifferenceforindividuals,brands,andtheindustrybydrivinggrowth,advancingtheinterestsofmarketers,andpromotingandprotectingthewell-beingofthemarketingcommunity.Foundedin1910,theANAprovidesleadershipthatadvancesmarketingexcellenceandshapesthefutureoftheindustry.TheANA’smembershipincludesmorethan1,000companieswith15,000brandsthatcollectivelyspendorsupportmorethan$400billioninmarketingandadvertisingannually.Themembershipiscomprisedofmorethan750client-sidemarketersand300marketingserviceproviders,whichincludeleadingagencies,lawfirms,suppliers,consultants,andvendors.FurtherenrichingtheecosystemistheworkofthenonprofitANAEducationalFoundation(AEF),whichhasthemissionofenhancingtheunderstandingofadvertisingandmarketingwithintheacademicandmarketingcommunities.AbouttheInteractiveAdvertisingBureau(IAB)TheInteractiveAdvertisingBureau(IAB)empowersthemediaandmarketingindustriestothriveinthedigitaleconomy.Itsmembershipiscomprisedofmorethan650leadingmediaandtechnologycompaniesthatareresponsibleforselling,delivering,andoptimizingdigitaladvertisingormarketingcampaigns.Thetradegroupfieldscriticalresearchoninteractiveadvertising,whilealsoeducatingbrands,agencies,andthewiderbusinesscommunityontheimportanceofdigitalmarketing.InaffiliationwiththeIABTechLab,itdevelopstechnicalstandardsandbestpractices.IABandtheIABEducationFoundationarecommittedtoprofessionaldevelopmentandelevatingtheknowledge,skills,expertise,anddiversityoftheworkforceacrosstheindustry.ThroughtheworkofitspublicpolicyofficeinWashington,D.C.,IABadvocatesforitsmembersandpromotesthevalueoftheinteractiveadvertisingindustrytolegislatorsandpolicymakers.Thereare43IABslicensedtooperateinnationsaroundtheworldandoneregionalIAB,inEurope.Foundedin1996,theIABisheadquarteredinNewYorkCityandhasaSanFranciscooffice.

Page 27: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 27

ParticipatingWorkingGroupOrganizations:

4As Google NewsweekACA Grapeshot OleCommunicationsANA GroupM OpenSlateAnnalect HorizonMedia P&GApple IAB PepsiCoAT&TAdWords IABCanada PfizerBankofAmerica IntegralAdScience PinterestBuzzfeed ISBA PublicisMediaCIBC ITNNetworks SIPIcomScore JPMorganChase SnapCrownMedia LinkedIn TAGDeloitte&Touche MediaManagementInc. TwitterDigitalContentNext Mediaoptimise UnileverDoubleVerify MMA VABEY Moat/Oracle VizioFacebook NAB WWEFOXNetworksGroup NCMMedia

14 ContactUsMRC:RonPinelli,VPDigitalResearchandStandards212-972-0300rpinelli@mediaratingcouncil.org

Page 28: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved.

AppendixA:SectionSummariesTorecap,themajoraspectsofeachSection1-8include(butarenotlimitedto)thefollowing:Section2(ContentGranularity)Summary(foradverificationorganizationsseekingcontentlevelBrandSafetyaccreditation)

• ContextandBrandSafetyfunctionalitymustincludecontentlevelgranularity• DynamiccontentandUGCmustbediscretelyconsidered• Differencesinpropertyandcontentlevelmustbeconsidered• Contentlevelgranularityshouldbeinadditiontocontinuingtoprovidediscrete

propertyclassifications• Adverificationorganizationswithcontentmeasurementgranularitymayreport

avoidanceandverificationatthepropertylevelwithproperdisclosure• AdverificationorganizationsmayclassifycontentorpropertiesbasedonURLkeywords

withoutcontentlevelmeasurement,howeverthisdoesnotrepresentcontentlevelmeasurementgranularityorreporting

• MRC’saccreditationprocesswillevaluateanddistinguishbetweenlevelsormeasurementandreportinggranularity.

• CapabilitiesandlimitationsateachlevelmustbedisclosedwithguidanceforuseSection3(SocialMediaandUGC)Summary

• SocialMediaPlatformcontentandUGCnecessitateuniqueanddiscreteconsiderations• Dynamiccontentmustdrivedatafreshnessandminimumbaselinedataguardrails• SocialMediaPlatformsandpublishersarestronglyencouragedtoeitherallowdirect

thirdpartymeasurementatacontentlevelortoactivelyprovidefirst-partydatatothird-partymeasurers

• Freshdataisabsolutelynecessaryforaccurateadverification• Minimumreportingthresholdsshouldbeemployed• Curationofcontentisencouragedandcanbeleveraged,butnotreliedonifunaudited• Curationbasedonpopularityorexposurenecessitatesenhancedfocusandmonitoring• Volumerequiresconsiderationofabilitytoscalemeasurementandminimizedowntime• MachineLearningmustbeempiricallysupportedandbasedonqualitydata• Thelevelofrelianceonmachinelearningversushumaninterventionormanualreview

mustbegenerallydisclosed• Models/algorithmsunderlyingMachineLearningmustbeappropriateandsupported• MachineLearningshouldbedeployedinconjunctionwithfocusedHumanIntervention• HumanInterventionshouldbeinformedbyMachineLearningandarisk-basedapproach• PersonnelinvolvedinHumanInterventionmustbeadequatelytrainedandsupervised• HumanInterventionresultsshouldcontinuallyvalidateandinformMachineLearning• MonetizationcontrolsareencouragedtomakeContextandBrandSafetymanageable

Page 29: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 29

• IVTshouldbeconsideredinContextandBrandSafetyprocessesandfunctionality• Proprietarytaxonomiesmustbesupportedanddisclosed• MRCstronglyencouragesandsupportsmechanismsandeffortstopromoteconsistent

BrandSafetydefinitionsandminimumbaselinessuchasthatproposedbythe4A’sAdvertisingProtectionBureau(APB)

• Aminimumlevelofcontentthatexcludescertainbaselinecategoriesofferedasacollectivereportingorblockinggroupconsideredunsafeorineligibleformonetization(“thefloor”)isencouragedeitherasadefaultcontroloraselectedbyadvertiserssuchasthe4A’sAPBAdvertisingAssuranceBrandSafetyFloorFramework

• Advertisersmaychoosetowaivethefloorandallowadvertisingtooneoralloftheincludedcategories

• The4A’sAPBwillincludeongoingeffortsincludingreconcilingdefinitionsofcommoncategorydifferencesbetweenvendors,whileallowingforcustomsettingsthatvendorsapplytoaddressspecificclientsBrandSuitabilityneedssuchasviatheAPB’sBrandSuitabilityFramework

• MRCfurtherrecommendsandsupportsIndustryeffortstostandardizeandmodernizetaxonomiesandcategoriesacrossplatformsconsideringUGCandcontentgranularity,andweintendtoparticipateinsucheffortsonanongoingbasis

Section4(Adjacency)Summary(foradverificationorganizationsmeasuringandreportingadjacency)

• Adjacencyshouldbemeasuredbasedonmeaningfulunitsandshouldbeplatformspecific

• Adjacencyalsoappliestovideoplaylistenvironments• Placementsthatfallzeromeasurementunitsfromspecificcontentbeingmeasuredor

thatareabletobesimultaneouslyviewable(50%ofpixels)areconsidereddirectlyadjacent

• Adjacencyrangesorscoresmaybereportedandmustbeclearlydefined• Platformsandpublishersareencouragedtoenablethird-partyadjacencymeasurement

orotherwiseprovidevalidatedfirst-partysignalstovendorsSection5(Metadata)Summary

• SocialMediaPlatformsandpublishersarestronglyencouragedtoallowdirectthirdpartymeasurementatacontentlevelortoprovidevalidatedfirst-partydata

• ThecollectionandtransmissionofMetadatamustbeauditedandvalidated• AdverificationorganizationsutilizingMetadatamustconsiderlimitationsofthisdata• Userreportingorflaggingofobjectionablecontentshouldbeappropriatelyweighted

basedonreviewandrobustqualitycontrol• AutomatedanalysisofMetadataalonewithoutadditionaldatasourcesorconsideration

ofhumanandmanualprocessesislikelyinsufficientforcertaincontentlevelContextandBrandSafetymeasurement

Page 30: MRC Ad Verification Supplement- Enhanced Content Level ... Ad Verification... · 2 Content Level Granularity Third-party ad verification functions have traditionally been focused

MRCAdVerificationSupplementFINAL

©CopyrightMediaRatingCouncil,Inc.Allrightsreserved. 30

• ThedegreeofrelianceonMetadatamustbedisclosedtousers• TheMetadatacollectionmethodmustbefullydisclosedandsubjecttointernaltesting• TheMetadatamustshouldbeclearlydefinedandtheextentofMetadatauseshouldbe

includedindisclosures• Processestoingestandanalyzemetadatashouldbedesignedtoaccountfordifferentor

multiplelanguagesSection6(MobileApplications)Summary(foradverificationorganizationsseekingcontentlevelBrandSafetyaccreditationformobileapplications)

• TheguidanceprovidedintheAdVerificationGuidelinesandwithinthisSupplementshouldbefullyappliedwithinmobileapplications

• SpecificconsiderationsdetailedintheIAB/MMA/MRCMobileApplicationAdvertisingMeasurementGuidelinesshouldbeapplied

• Adverificationorganizationsmustapplyincrementalanddifferentialconsiderationtomobilein-appenvironments

• Contentlevelgranularitymustbeapplieddiscretelywithinmobileapplications• MRCintendstoconsiderBrandSafetyaccreditationdiscretelybetweendesktop/mobile

webandmobilein-appenvironmentsSection7(IllegalSourcesandIndustryCommunication)Summary

• Adverificationorganizationsmustconsiderlegalrestrictionsrelatedtoprohibitionofcontentandthesourcesofthatcontentaswellasenforcementandremovalperiods

• Publishersandplatformsshouldhaveaqualificationprocesstoensuretheyaredealingwithalegitimateentityandlegallypermissiblecontentappliedacrosstheplatform

• ProceduresshouldincludeinitialqualificationusingIndustryandlocalsourcesofknownillegalentities,aswellasongoingevaluationlinkedwithadverificationresults

• Processesshouldbeputinplacetoroutinelycommunicatefindingsrelatedtoillegalsourcestoauthorities,oversightbodiesandtheIndustryatlarge

• MRCstronglyencouragesandsupportsmechanismsandeffortstocommunicaterelevantBrandSafetyinformationacrosstheIndustryandpublicly

Section8(EnhancedBuyerTools)Summary(foradverificationorganizationsseekingcontentlevelBrandSafetyaccreditation)

• Curation,monetizationcontrolsandgranularcategorizationmustbeconsistentlyanddiscretelyreportedtousersinordertoenablecustomizedtargetingandavoidance

• Adverificationorganizationsmayenableadprevention,blockingandalertingonthebasisofMetadata,discretecontentcharacteristicsandadjacencycriteria

• Thebasisofmeasurement,methodologyandlimitationsunderlyingenhancedbuyertoolsmustbeclearlydisclosed