Google Clickstream Whitepaper

Beyond last click: Understanding your consumers’ online path to purchase

Insights and learnings from Google’s clickstream research

Contents

• ExecutiveSummary

• Introduction

• Methodology

• OnlineResearchDuration

• PurchasePaths

• TheRoleofSearch

• Brandvs.GenericSearches

• Conclusions

• NextSteps

Executive SummaryThispapersummarisesthekeyfindingsandinsightsfromGoogle’sresearchintotheimportantareaofclickstreamattribution.Itwillgiveyouthefactstoanswerkeyquestionssuchas:howlongisthetypicalpathtopurchase,howmanysiteswillbevisitedonaverageandhowdoessearchcontributetoconsumerdecisionmaking?Wehopethatthispaperwillencouragemarketingleaderstomovebeyonda‘lastclickwins’modelandtakeafact-basedapproachtounderstandingtheirconsumers’pathtopurchase.Keyfindingsinclude:

• Researchjourneysarelong and can last a month or more insomecategories.• Oneinthreeconversionsoccur30days after the online researchbegan.• 70%ofpurchasersusesearchatsomepointintheirresearchjourney.• Searchpathstopurchaseoftenincludebothbrandandgeneric terms and 48%ofpurchasersswitchbetweentermsatsomepoint intheirjourney.• Thewayshoppersresearchandpurchasedifferentproductsonline does notneatlyfollow traditional marketingfunnelmodels.• Theamountofresearchconductedonlineisnotclearlylinkedwiththe averageordervalueoftheproductsinquestion.

IntroductionAccordingtotheConnectedKingdomReport1,theInternetcontributed£100billiontotheUKeconomyin2009,andlastyear62%ofUKconsumersboughtgoodsorservicesonline.TheonlineadvertisingmarketcontinuestogrowandthelatestPwC/IABfiguresshowitisnowworthover£4billionintheUK2.Thesheernumberofpeopleengagedinonlineresearchinsomecategoriesisstaggering.Forexample,60%oftheUKinternetpopulationresearchedtravelproductsonlineina3monthperiod3,andhalfofalltravelintheUKisbookedonline1.Intheapparelsector,wherethemajorityofsalesstillgothroughretailstores,therewereover22millionpeopleresearchingapparelonlineintheUKinthe3monthperiodweanalysed2.

Inamarketwhere£1inevery£4ofmarketingbudgetsisspentonline,understandingshoppers’onlinejourneyswillhelpensureonlinemarketingactivityandadvertisingmeasurementpracticesreflectrealconsumerbehaviourpatterns.Tothisend,GoogleandNielsenpartneredtoconductclickstreamanalysesofonlineresearchandpurchasebehaviourinmultiplesectorsintheUK.

Theobjectivesoftheseclickstreamanalyseswastoprovidespecificinsightsinto:

• Whatconsumersresearchonline• Howlongtheyspendresearching• Whichsitestheyvisit• Howtheynavigatebetweensites• Whattheysearchfor

1. Boston Consulting Group / Google - Connected Kingdom report, October 2010. 2.IAB/PwCasspendfiguresfor2010.3. Custom Google / Nielsen clickstream research, 2010.

MethodologyTheclickstreamanalysescommissionedbyGooglearebasedondatafromNielsen’sNetViewonlineconsumerpanel–thesourceofUKOMdata.Thefirststepinastudyistoidentifyasub-setofNetViewpanellistsbydefiningcertainbehavioursweareinterestedin,forexampleresearchingtravelonline,andthenexamineallothercategoryrelatedactivityinagivenperiod.

Thenweisolateatimeperiodofanalysis–usuallyathreemonthperiod.Thislengthwaschosenaspreviousresearchinthisareaindicateditwaslongenoughtocaptureentirejourneyperiodsbutshortenoughtoisolatetherelevantactivity.GiventheelongatednatureofthehousepurchasingprocessintheUK,thepropertystudywasrunoveralonger6monthtimeperiod.

OnceacategoryandtimeperiodhasbeenchosenwethenworkwithNielsentoclassifythesitesandsearchtermsintorelevantsub-categories.Foragivencategory(suchascarinsuranceortravel)weidentifyalltherelevantsitesthatfallunderthatcategory.Thendifferenttypesofsiteswithinacategory,e.g.aggregatororadvicesites,aresub-classified.WealsoanalysewebsiteURLstructurestoidentifyconsumerswhoperformedanactionofinterest,e.g.placinganiteminabasket,fillinginanapplication,orbuyingaflight.Finally,allcategory-relevantsearchactivityisidentifiedandclassifiedaseitherbranded(e.g.“Vodafone”,“Nokia”),generic(e.g.“phonecontract’’,“mobilephone”),oracombinationofboth.

Oncewehaveclassifiedsitesandsearchterms,wesegmenttheaudienceintomutuallyexclusivegroups:researchers,estimators,andpurchasers.Researchersarepeoplewhovisitcontent-relatedpagesbutdon’tgetaquoteortransact.Estimatorsarepeoplewhoplaceitemsinbasketsorgetquotesonlinebutdonotmakeapurchase.Purchasersarepeoplewhotransactorperformanotherconversionevent,asdefinedinfigure1.Forthemajorityofthispaperwefocusonpurchasers.

AsweareusingNielsen’sopted-inconsumerpanelalldataprotectionlawshavebeenmetandnopersonallyidentifiableinformationhasbeenusedinthesestudies.

Thevalueinusingapanel-basedanalysis,comparedwithwebsiteanalyticsoradservertrackingdata,isthatwegetamorecompletepictureoftheonlinejourneyacrossthewholecategoryecosystem,ratherthanjustthroughthelensofanindividualadvertiser’stouchpointswiththeirshoppers.

*Samplesizereferstoanyuserwhovisitedcontentorsearchedontermsrelatedtothatcategoryinthe definedperiod.

Figure 1: Studies included in this paper

Travel

Apparel

Mobile Phones

Energy

Car Insurance

Loans

Property

15,453

12,831

6,713

4,729

3,796

1,700

3,722

Online travel purchase

Online apparel purchase

Online mobile phone purchase

Online utilities purchase

Online insurance purchase

Online loan purchase

Download (PDF, schedule etc), Action (email friend etc) or contact pages

UK Study Sample size* Definition of purchase

Online Research Duration

Figure 2: Average journey length in days per purchase

Base:Allpurchasersasdefinedineachcategory

MobilePhones

23

Loans

9

Apparel

2730

25

20

15

10

5

0

Energy

20

Travel

24

CarInsurance

12

1 in 3 conversions takes place 30 days after the research began

Oneofthefirstthingswelearnedfromtheclickstreamresearchisthatconsumerstypicallyspendmuchlongerresearchingpurchasesonlinethanmightbeconventionallyassumedfromexaminingadserveroranalyticspathdata.Peopleshoppingforclothesorshoesonlinespend,onaverage,almostamonthshoppingaroundandmakinguptheirmindbeforemakingapurchase.Travel,withallofitsfantasyshoppingpotentialandmyriadofproviders,aggregatorsandadvicesitesinvolvesjourneyswithanaveragelengthof24days.

Advertiserscandeterminewhattheirownconversionjourneyslooklikebyusingconversiontrackingtoolswhichallowthemtomakeinformeddecisionsontheircookiewindows.Ideallysuchdecisionswouldstarttogetcategoryspecific–theattributionwindowfordifferentcategoriesshouldreflectdifferentcustomerjourneylengths.

Figure 3: Cumulative percent of purchases for apparel, travel and energy


80-89Days

50-59Days

25-29Days

10-14Days

70-79Days

40-49Days

20-24Days

6-9Days

60-69Days

30-39Days

15-19Days

1-5Days

<24Hours

100%

70%

80%

90%

50%

60%

40%

30%

20%

10%

0

%ofp

eoplewho

purchased

Timetopurchase

ApparelPurchasersTravelPurchasersEnergyPurchasers

Thoughtheaveragepathtopurchaseforallverticalsisunder1month,wefindthatoneinthreeconversionsoccuratleast30daysaftertheinitialresearchbegan.Intravel,morethanonequarterofpurchasesoccurafter30days.Thistrendisrepeatedinretailwhereonethirdoftransactionsoccurafter30days.Incontrast,whenpeopleresearchonlineforagasorelectricityproviderabouthalf(54%)ofthemconvertwithin24hours,comparedwithonlyaround20%ofwithinthedaypurchasesintravelandapparelshopping.

Moredaysinmarketusuallyaddsuptomoreactivityrackeduponline–intermsofbothhoursspentandnumberofsitesvisited,asillustratedinFigures4&5.

Figure 5: Site visits before a purchase

Figure 4: Average research time in hours per purchase



MobilePhones

01:29

Loans

00:29

Apparel

03:17

03:00

03:30

02:30

02:00

01:30

01:00

00:30

00:00

CarInsurance

00:50

Travel

01:43

Energy

00:43

Travel

Apparel

Mobile Phones

Energy

Car Insurance

Loans

Overall Averages

21.6

11.4

9.4

7.4

5.6

4.2

9.9

9.4

2.9

4.1

3.4

3.9

2.8

4.4

2.3

3.9

2.3

2.2

1.4

1.5

2.2

CategoryAverage site

visits per purchase

Number of different sites visited

per purchase

Average number of visits

to each site

Purchase Paths

Figure 6: Travel path

Journey paths can be very long and complex

Inordertobringthesevariousmetricstolife,ithelpstolookatactualuserpaths,andhowpeoplesearchandresearchontheirwaytoapurchase.

Forexample,abovewehaveanindividual’sonlinejourneyforahighinvolvementpurchase:travel.Thisparticularjourneylasted28days,andincludedmanygenerickeywords(genericinthiscontextmeanstravelrelatedtermsnotincludinganybrandsforcompaniesinthetravelsector).TherangeofsitesvisitedincludeaggregatorssuchasTripAdvisor,aswellastouroperatorssuchasFirstChoiceandThomson,manyofwhomprovidedquotes–butFirstChoicegetsthebusinessintheendafterallthatshoppingaround.Thefinalpurchaseismadeaftertheindividualsearchesonceagainfor“corfuholidays”,whichwasthequerythatbegantheirjourney.

Quotation Purchase

FirstChoice

Searches

Searches

Sites Visited

Sites Visited

DAY

DAY

1 4 122 7

Bingmaps&localHoliday-Truths.com

TripAdvisor

TripAdvisor

Bingmaps&local

Corfuholidays Palacemonrepos

Travelzoo

AfricanSafariClub

TravelzooFirst Choice

TUI Thomson

TUI ThomsonOlympic Holidays

TUI ThomsonSunshine.co.uk

BluebayescapeBluebayescape

TripAdvisorHoliday-Truths.com

TravelLibraryGoogleMaps

16 2722 28

TripAdvisorKallistoresort,

corfuKallistoresort,

corfu

Kontokalibayresort&spaKontokalibayresort&spaKontokalibayresort&spa

Bustoursincorfu

First ChoiceOlympic Holidays

TUI ThompsonTravel Republic

Travelzoo

Travelzoo

TUI ThomsonBooking.com

First choice

Corfuholidays

Figure 7: Audience movement in the travel market before a purchase

Overallnavigationpatternsintravelremaincomplex,rightupuntilthemomentofpurchase.Figure7illustratesthattheonlineshopperisstillconsultingawiderangeofsourcesduringthefinalsessionwhenatransactionhappens.

Base:PurchasersintheTravelmarket.Note:Arrowsrepresentproportionofuserswhogofromonesitetypetoanotherinthepurchasesession.(Audiencemovementsunder7%notstated)

Other Search

Travel OTA

Air Travel

LodgingGroundTransport

TourOperator

3%5%

26%

2%

11% 8%

13%

4%

16%

10%

7%

9%

16%

13%

7%

11%

7%

17%8%

10%

8%

Startpointsinblack GoogleTraffic

19%

10%

12%

15%

Figure 8: Car insurance path

Figure 9: Audience movement in the car insurance market before a purchase

Incontrast,weseethatatypicalonlinecarinsurancepurchaseismuchshorterandmoredirect.Inthisexamplepath(figure8,below),thereareonlytwoactivedaysinthecategory,andtheindividualobtainsfewerquotespriortoengaginginaconversionevent.Inaggregatethenavigationpathsinsuranceshopperstake(figure9)appearmuchsimplerthanintravel.

Base: Purchasers in the Car Insurance market. Note:Arrowsrepresentproportionofuserswhogofromonesitetypetoanotherinthepurchasesession.(Audiencemovementsunder7%notstated)

Quotation Purchase

Searches

Sites Visited

DAY 1 15

CarinsuranceCarinsurance

GoCompareMoneySupermarket

Go CompareMoney Supermarket

Go CompareMoney Supermarket

Saga

Startpointsinblack GoogleTraffic

Direct InsurersAggregators

10%26%

25%

27%

40%

The Role of Search

Thisanalysisonlyshowscorrelationbetweensearching,clickingonsponsoredlinksandmakingapurchase,notcausation.However,thisdoesindicatethatpeoplewhoareinmarkettobuyaremorelikelytoclickonadstonavigatemoredirectlytorelevantshoppingrelatedcontent.

70% of consumers who purchase use search

Figure 10: Percent of consumers using search in their purchase journey

Base:Searcherswhopurchasedasdefinedineachcategory

Apparel

66%

Loans

56%

Travel

92%

MobilePhones

66%

Property

86%

76%

Energy CarInsurance

60%

100%

70%

80%

90%

50%

60%

40%

30%

20%

10%

0

Acrossthesectorsstudied,over70%ofconsumerswhopurchasedonlineperformedatleastonerelevantsearchduringtheobservationperiod.Individualswhobuyonlineare30%morelikely,onaverage,tohaveperformedasearchthanindividualswhoseonlineresearchdidnotendwithaconversion.Theseindividualsarealsomorelikely(by17%)tohaveclickedonasponsoredsearchad.

Figure 11: Average number of searches per purchase

Althoughapparelshoppersspendthelongestnumberofdaysdecidingtobuy,travelshoppersarebyfarandawaythemostprolificsearchersduringtheresearchprocess.Infact,1in10travelresearchersactuallysearchedmorethan50timesduringthe3monthobservationperiod.


Apparel

6

CarInsurance

3

Travel

20

25

20

15

10

5

0

Energy

5

MobilePhones

7

Loans

3


Brand vs. Generic Searches

Figure 12: Brand vs. generic searchers

Generics play a key role in the research process

Insomecategories,suchaspropertyortravel,usinggenericqueries(e.g.“propertyforsaleoxford”)isalmostubiquitous(seefigure12).Inothersectors,althoughthevolumeofgenericactivityisstillsubstantial,asignificantproportionofshoppershavegoodknowledgeofthesectorparticipantsfrompriorexperienceorextensiveadvertisingandnavigatesolelythroughbrandterms(e.g.“Vodafone”).Itshouldbenotedhoweverthatalthoughthereachofgenerickeywordsmaynotmatchtheaggregatereachofbrandterms,thereachofgenerickeywordswillusuallybemuchgreaterthanthereachofanyindividualbrand’ssearchterms,astheirshareofvoicewillonlybeafractionofthebrandedsearchtotal.

Property

Only Generic search terms

Brand and Genericsearch terms

Only Brand search terms

Apparel

Travel

Energy

Loans

MobilePhones

CarInsurance

0% 30%20%10% 40% 50% 60% 70% 80% 90% 100%

X:NofurthersearchBase: Purchasers in the ‘Apparel’ Market

Justaspeoplewhoconvertonlinearemorelikelytousesearchandsponsoredlinksthannon-converters,theyarealso19%morelikely,onaverage,tosearchongenerictermsthanthosewhoendupjustlookingand/orconvertingoffline.

Figure 13: Generic to branded paths

Itisoftenassumedthatconsumersstarttheironlinejourneywithgenericterms,andusebrandedtermsastheymoveclosertothepurchaseevent.However,theseclickstreamanalysesdonotfullybackupthisbelief.Thereisatendencyformoregenericstoappearearlierintheprocess,butmanypathsdonotfollowtheexpectedpattern.

X - 20%

G - 53%

G - 14%

Generic29%

Brand71%

B - 27%

B - 69%

G - 19%

G - 11%

G - 60%

G - 56%

B - 69%

B - 73%

B - 21%

B - 37%

Search

1st Search 2nd Search 3rd Search

X - 17%X - 8%

X - 16%

X - 12%

X - 19%

Shoppersoftenswitchbetweengenericandbrandkeywordsearchesatdifferentpointsintheirjourney.Forexample,figure13documentsfirstthreesearchesofapparelshoppers. 14%ofapparelshopperswhostartwithabrandedsearchfollowonwithagenericsearch,however27%ofpeoplewhostarttheirjourneywithagenericsearch,continueupwithabrandedsearch.Onaverageacrosscategories,48%ofpeopleswitchbetweenbrandandgenericatsomepointonthewaytoaconversion(seeFigure14).

Figure 14: Percent of consumers who use both brand and generic terms in their purchase journey

48% Average


90%

70%

80%

50%

60%

40%

30%

20%

10%

0

Loans

50%

CarInsurance

21%

Travel

79%

Apparel

39%

Property

59%56%

Energy MobilePhones

34%

Conclusions

Consumer behaviour is intensive, complex and highly personal

Toachievethemaximumnumberofprofitablesalesorconversionsadvertisersneedtounderstandthefullvalueofallconsumeronlinetouchpointsandinparticulartheroleof‘assist’clicksintheconsumerpathtoconversion.

Traditionalmarketingfunnelmodelsarelessusefulintheonlineenvironment,wheresmart,savvyconsumersareincontrol.TheNielsenclickstreamstudiesshowtheextenttowhichpeopleshoparound,andresearchovermultiplesessions,enteringmultiplesearchesacrossbothbrandedandgenerickeywords.Evidencefromthestudiescallsintoquestiontraditionalassumptionsabouttheroleofbrandedandgenerickeywordsinthefunnelastheyareoftenbothusedatallstagesintheonlinejourney.

Tomaximisesalesonlineadvertisersmusttakeadvantageofallconsumertouchpoints,notjustthosethatdrivebusinessonalastclickbasis.Multi-clickjourneysmatter,theydrivesalesvolumeandvalue:theyaccountfornearlyhalfallpaidsearchconversionsandtheyareworthmore,onaverageanadditional8%inbasketvalue4.Genericsearchactivitytendstohappenearlierintheconsumerjourney,andthereforeislikelytobeundervaluedbylastclickattributionmodels.

4. td Search weighted attribution paper, 2010

Next StepsRefineyourapproachtosearchtoaddressallstagesofthecustomerjourney.Usethefollowingprocesstohelpyouactivelymanageandbideffectivelyonbothconvertingandassist(sometimesknownas“pathway”)keywords:

Test and iterate

•Examineyourowncustomerjourneysandassessareasforpotential misattribution.UsetoolssuchasGoogle’sSearchFunnelsor DoubleClickExposuretoConversionreportingtohelp.

•Measurethesensitivityofkeywordperformancetomultipleattribution modelstoexposepotentiallyundervaluedkeywordsandidentifywhat needsfurthervalidation.

•Runexperimentsorregressionmodellingtovalidatewhatreallyworks.

•FeedtestresultsbackintotheattributionmodelsorCPAtargets.

•Continuetooptimise.

•Repeat!Thetestanditeratecycleshouldbearegularroutine.

Google Clickstream Whitepaper

Documents

Transcript of Google Clickstream Whitepaper