perfect your sem testing: step back and look at what you should be measuring by ted ives
TRANSCRIPT
#SMX #22B @TedIves
• WhoamI?• Whydowetest?• Howshouldwetest?• Whatshouldwemeasure?• Whereshouldwemeasure?• Whenshouldwetest?
Agenda
#SMX #22B @TedIves
• AnonlinemarketingagencyoutofRhodeIsland(fourpeople)
• Wecouplebig-agencyresultsw/small-agencyattention– Leverageoffshoreresources
– Constantlydocumentandsystematize
– Delivertremendousvalue&focusonclient-specificanalysesfor“movingthedial”
• PaidSearch,SEO,Analytics,Remarketing/Display,LinkedIn
• Clients:
Who am I?
#SMX #22B @TedIves
BUT…wecanonlytestwhatwecanmeasure!Thisquestioniseverything:Howdocustomersbuy?
Why do We Test? Answer: To get these to go up.
http://searchengineland.com/consider-the-ends-to-justify-the-means-in-ppc-117740http://searchengineland.com/ad-testing-are-you-using-the-wrong-success-metrics-123556
#SMX #22B @TedIves
• Writemultipleadsanddropthemintoadgroups• Measureresultsandalsodocross-adgrouptestingwithAdAlysis• UseUnbounceand/orOptimizelyforLandingPageTesting• Don’tforget…asuccessfultestmayactuallycause*worse*performance(!)• Whenyou’redonetesting,testsomethingelse!• Butunderstandthatmuchoftestinginvolves….waitingaroundforresults.
How Should We Test?
#SMX #22B @TedIves
• Ifyou’redoingalotoftesting,justgetAdAlysis;itdoesitforyouautomatically.
• Severalofthemostpopularadteststatisticscalculatorsonthewebareabsolutely*incorrect*.
• Usethisinstead–it‘saccurate:http://www.cardinalpath.com/resources/tools/ppc-ad-testing-tool/#results
• Ifyoureallyneedtodomulti-adgrouptestingoflandingpagesacrossmultiplelandingpagevariations,oneeasywaytocheatisbysimplyappending#ontheendofyourfinalURLstorepresenteachvariation.Makessummarizingeasier:foo.com/asdf.htm#v1foo.com/qrst.html#v2
What should we measure? (i.e. How can you reliably check statistical significance?)
#SMX #22B @TedIves
To test, first you need a willing subject…
Website:h*p://www.historicalbricks.comDesignedandimplementedby:h*p://lucemsolu<ons.com
#SMX #22B @TedIves
Manycustomersalsocomeinsimplyviaphone.Whatweneededtodowastomeasurebothformsubmissionsandphonecalls,somehow.Ourtheorywas…ifengagementishigherwiththeungatedpage…perhapsthesepeoplecancomebackandconvertlater…maybeevenviaphone.
Then we realized….the key question is: Where should we measure?
#SMX #22B @TedIves
• YoucaneasilygetlimitedphonetrackinginAdWordsbysimplyaddingcallextensions.
• Thetrickis,gettingthephonenumberonyoursitetochangebasedonthephonenumberthepersonsawinthesearchadinyourcallextension.
• AdWordscanswapoutthephonenumber;butitrequiressomeknowledgeofjavascript.
• CallRail,bycontrast,makesitveryeasy:• Aniceone-linelibraryyoucallthatdoesalltheheavyliftingforswappingoutthenumber
• GoogleAnalyticsintegration
• Wordpressplugin
• Youcantrackorganic(channel,notkeywords)anddirectinadditiontopaidsearchkeywords.
• Justbecareful;iflocalsearchisimportanttoyou,don’tuseittotrackorganictraffic(NAPimplications)
Call Tracking
#SMX #22B @TedIves
When should we test?
• Oneperspective:100%ofyourspendshouldalwaysbeundertest• Butthismeans50%ofthetime,50%ofitisgoingtodo*worse*.Hmm…
• Thisarguesforsomesmallerportionofspendbeingundertest(!)
• Anotherperspective:howmuchtestingcanyoumanage?• Havingatooltomanagethetestingforyoumakesitmucheasiertorunandinterpretlotsoftests
• Whatdoestherestofthepanelthink?
#SMX #22B @TedIves
• Areyouevenmeasuringtherightthings?• Considerphonetracking
• Ifbuyingcycleincludesextensiveresearch,considerungatedlandingpages
• Takeintoaccountengagementaswellasconversions
• Don’tforgettorunattributionanalysesinGoogleAnalytics
• Makeiteasytoconvert!• Askearly,andoften
• Floatingcontactforms
• Makephoneprominent
• Checkout:http://tedives.com/
Takeaways