©2017ProgrammaticMechanics
Page1
TheFourProblemsSolvedbyRTBReal-time bidding (RTB) or what has become known as programmatic technology allowsinstantaneous decision making with regard to which ad will show on a website. Instead ofdeciding in advance which assets the adserverwill display,now thedecision canbemade as the user loads the website page.Theadserverrunsanauctiontosell theadspace in real time, forming an online adexchangedrivenbyaSellSidePlatform(SSP).Not surprisingly, this technology has seenstrong adoption and continues to thrivetoday.
Whilethetechnologywasdevelopedinearly2007 it wasn’t until later that the marketbegan adopting at scale. The technology isnow used by much of the market and theamount of money spent through RTBcontinues to increase each yea, as seen inFigure1.1
RTB technology provides solutions tonumerousproblemsthatplaguepublishersandadvertiserstryingtobuyandselldigitaladspace.Onthepublisher’sside,RTBtechnologysolvesissuesofforecastingandunsoldinventory.Ontheadvertiser’sside,RTBmanagesissuesofdistrustandcontrol.
ForecastingIssues
As display advertising became a ‘must buy’ for digital agencies, all major advertisers beganpurchasing inventory. The most prestigious brands started cutting up-front deals with theirfavoritepublishers.Allofthe leadingonlinepublishersestablished ‘yieldmanagement’teamswhoweretaskedwithensuringtheratescreatedenoughdemandtofillthesupply.Theseteamswouldusedatafromtheadserver,aswellasotherthirdpartyapplications,todeterminehowmany impressions were available in each section. They would then price out the inventory(availableimpressions)intheinterestofmaximizingrevenue.Somepublisherspulledtalentfromtheairlineindustry,astheywerethemostadvancedinsupply/demandanalysisandinventorymanagement.
Theyieldmanagementteamwouldusehistoricaldatatodeterminehowmanyvisitorsoneachpage.Usingspreadsheets,advanceddatabasetechniques,orthirdpartysoftware,theseteams
Figure1:USRTBDigitalDisplayAdSpend
©2017ProgrammaticMechanics
Page2
wouldestimatehowmanyadimpressionswouldbeavailableforsaleinthefuture.Salesteamswould,inturn,approachthemarketandselltheinventory,inpieces,toanyadvertiserwillingto
pay the premium prices. To avoidconstantly providing ‘make-goods’, orfuture inventory due to under delivery,the yield management team wouldforecast 5%-10% less than the exactamount they anticipated. The inventorywould vary by day and the forecastedinventory vs. sold inventory would looklikethechartinFigure2.
Unpredictability, an inherentcharacteristic of the Internet, creates afar larger loss of revenue than shownabove.TheInternetprovidesthemostupto theminute informationpossible, andwhenmajoreventshappenwebsitescansee 10x the traffic they had forecasted.Thesupplyspikecanoccuronaweatherwebsite as a northeaster approaches, anewswebsiteduringanaturaldisaster,agossipwebsiteafterthesuddendeathofamajorcelebrity.Theadvertisersdonotwant‘unevendelivery’,meaningthepublishercannotdeliveralloftheimpressionsfora12-weekcampaigninonesingleday.Thismeansthatwhenaspikeof10xoccurs,eveniftheyieldteamanticipatesthisbased on the morning numbers, they cannot deliver more than 2x or 3x the amount theypredicted.ThemonetizationofthatdaywouldlooklikeFigure3totheleft:
Depending on circumstance,this could equate to lostrevenuesanywherefromtensof thousands of dollars toseveralmillion, alldue to theinability to monetizeinventoryinonesingleday.
-
500
1,000
1,500
2,000
2,500
Mon
Tues
Wed
Thurs
Fri
Sat
Sun
Mon
Tues
Wed
Thurs
Fri
Sat
Sun
Impressio
ns(inthou
sand
s)
Time->
ActualDeliveryvs.ForecastedAvailability
Actual Forecast
Allareainbetweenactualdeliveryandtheforecastareareasoflostrevenuewhere'houseads'orself-promotingadsservedinsteadofapayingadvertiser
- 2,000
4,000
6,000
8,000
10,000
12,000
Impressio
ns(inthou
sand
s)
Time->
ActualDeliveryvs.ForecastedAvailability
Actual Forecast
Figure2:ActualDeliveryvsForecastedAvailabilityChart1
Figure3:ActualDeliveryvsForecastedAvailabilityChart2
©2017ProgrammaticMechanics
Page3
UnsoldForecastedInventory
Premiumpublisherswhocanselltheabilitytoreachaspecificnicheaudiencesareoftenabletosettheirpricescorrectlyastosellthrough100%oftheirinventory.Otherpublishershavetheopposite problem. The largest internet publishers, or portals to the internet, amassed suchenormoustraffictheycannotsellalloftheiradspaceateffectiverates.Thelargestportalssawthreebillionvisitsadayattimes.Threeadsonapageperdayequatesto9,000,000,000availableeachday.Iforganizationswithtoomuchinventorycoulddetermineamethodtosellthelong-tailor‘remnant’inventoryatevena$0.01CPMtheywouldmakemillionsofdollarsamonthinextrarevenue.Whileremnantsolutionswereavailable,theystillcouldnothandlethemajorityofthisunsoldinventory.
Distrust
WhenAdNetworksstartedgainingpopularitycirca2000,theyallrepresenteduniquepublishersandcouldshowdifferentiationwiththeir‘sitelists’ordomainsinwhichadswouldserve.Thebiggestpublishersbeganrunningtheiradsacrossseveraldifferentnetworks.ThisstyleofyieldmanagementwouldbecomethecoreoftheInternetpublishingbusiness.Byrotatingtheads,thenetworksbeganhavinganissueofdifferentiation,asthesamesitewouldbeavailableacrossavarietyofnetworks.Additionally,manyoftheseadnetworkswererestrictedtosellingblindlyorundisclosedinordernottounderminethepublishers’relationshipswithmarketers.
Toachievethescalenecessary,theybegantoavoiddisclosingtobuyersthedeliveryreportsbydomain.Theywouldinsisttheycouldnotsharehowthedeliveryvariedacrossthewebsitesonthe ‘site list’. This would allow them to ‘dump’ a lot of inventory on a low-quality website.FacebookappssuchasFarmville,websitessuchasmyclassmates.comanddictionary.comareclassic locationswhere youwill see network ads in volumes,which are not conveyed to theagency or more importantly the advertiser. These networks could also adjust their marginmethodologyandchargeadvertisersmarginsashighas70%-80%.
Somenetworksbeganusingotherformsoftechnologytoshowdifferentiation.Networksclaimeduniquenessby“focusingonaspecialaudience”or“providingscale”or“usingcontextualsearchtechnology,”butuntiltheadvancementoftheprogrammaticRTBexchangetherewasbasicallynodifferencebetweentheseorganizationsastheyallworkedwiththesamepublishers.
Astheseproblemsleakedoutofthebackroomsofthesenetworksandintotheearsofmediabuyers, there was a clear need for greater transparency. By handling the auction for theadvertising space on an impression-by-impression basis, the RTB exchange offered atechnologicalsolution.
©2017ProgrammaticMechanics
Page4
Control
Thenetworkswerethefirsttoimplementanddevelopadservingtechnologyandcontinuedtodevelop in the interestof theirownbottom lines.Thebestadvances in targetingweremadeavailabletothepublishers.Publisherscouldusetheuptodateadserverstocontrolwhattypeofadswouldshowonacertainsectionofawebsite,inaspecificgeographicallocation,orataparticular time. While advertisers could request these targeting parameters, there was nocontrolontheirsideofhowtheadswereserved.Someoneonthepublisherornetworksideneededtohandlethetrafficking.
Asthelackofcontrolbecameagrowingissueontheadvertiserside,theRTBtechnologybecameavailable.Programmaticpushedthe‘control’intothehandsofthe“buy-side”orwhatisoftenreferredtoDemand-side.Thisallowedtheadvertisertomanagehowtheywanttotargettheiradsacrosspublishers(inbulk,becoming‘exchanges’).Thiswasaclearbenefitforallinvolved,asthenewfoundconfidenceledtolargerbudgets.
TheRTBexchangeprovidedsolutionstothesefour issues,buttheagencieswe’reset intheirways and comfortable with the accustomed methods. Changing the way an entire industrybehaves takes time and takes ‘buy-in’ from big players early. How could the networks andpublishers ensure the advertisers would test this technology that offered ‘control’ and‘transparency’intopricingandtargeting?Theycreatedanauctionenvironmentpromisingthebuyerstheywouldpayonlya‘penny’morethanthenexthighestbidderforeachimpression.ThisSecondPriceAuctionmethodcreateda ‘perfect’marketallowing thenextgenerationofmediatoemerge:ProgrammaticMedia.
References1 http://www.iab.net/about_the_iab/recent_press_releases/press_release_archive/1996_pr_archive
2(or nodes, or boxes, or servers - all terms which refer to a computer, usually without a monitor and in a building with no windows somewhere relatively unpopulated) 3 Pemberton, Steve. http://homepages.cwi.nl/~steven/Talks/2011/05-07-steven-visualisation/,CWI and W3C, Amsterdam, © 2011 4 http://www.google.com/patents/US20070192356.pdf 5 http://adage.com/article/digital/real-time-bidding-account-25-display-ad-spending-2015-emarketer/238300/