iot and deep learning in retail: the hyper-relevant, competitive … · 2017-09-11 · iot and deep...
TRANSCRIPT
IoTandDeepLearninginRetail:thehyper-relevant,competitiveretailerByPremCouture,CEO,ShareMyInsight,
Inapreviousposting,IdiscussedhowIoTconnectedstoresareabletocombineliveshopperjourneyandproductdatawithPOS,loyalty,socialmediaandotherdatasets.Also,howapplyingmachinelearningenablesrealtimeinsightsthatcantransformthecustomerexperience,enablecustomercentricmerchandisingandstreamlineoperations.Inthisposting,IwouldlikesharemythoughtsandexperienceonhowIoTinretailcanpowerbricksandmortarstorestocompeteinanomni-channelworldbybecominghyper-relevantacrossallcustomertouchpoints.
ASurgingWaveofDisruptionandOpportunityAspreviouslynoted,classicretailstrategiesandmethodologiesfordiscoveringandengagingcustomersareincreasinglyunmanageable,duetorapidlyevolvingcustomerinterestsandbehaviorpatternsandasevidencedby:
• theexponentialgrowthintheamountofshopperjourneys:fromresearchtopurchasetofulfillmentandcustomersupport,thenumberofpossiblejourneyshasgrownfrom40toamazeofmorethan800(Cisco)andfurtherincreasesovertime.
• theexpandingnumberofdatapoints(beyondspendanddemographics)andtherapidchangeinconsumerinterestsismakingthetraditionalrulesapproachtodataminingcustomerstobelessandlessmeaningful.
• theincreasingdemandforseamlessshoppingwithgreaterchoicesandlowerpricesacrossonline,in-store,andmobileplatforms,iscreatinga‘digitaldivide’betweenconsumerexpectationsandretailers’abilitytodeliver.
InnovationattheHeartoftheNewRetailRealityIfasensornetworkrepresentsournervoussystemandaDeepLearningplatformisourbrain,thenthepartthatmanagesretailprocessesfromsupplychaintomerchandisingandcustomercommunicationsissimilartothewayweengageandlearnfromtheenvironmentaroundus.Enablingcustomerstomakeeasyandcostefficientdecisionsfromawidearrayofchoicesiswhataconnectedretailerpreciselybecauseitcontinuouslylearnsandadaptstonewinformation.Someofthekeytechnologyadvancesthatmaketheabovepossibleinclude:
1. Retailsensordevicesthatactinasensorfusionmodeandlivestreamshopperandproductdatatocloudplatforms.
2. AIandDeepLearning:advancesinGPUacceleratedcomputingpowerenablesDeepLearningalgorithmstofindpatternsinlargeanddisparatedatasetsandtotransformdataintoinsight.
3. Storediagnosticscandetecthowproductplacement,brands,rangeassortment,pricing,personnelandstorelocationaffectshopperbehaviorandpurchasingdecisions.
4. Dynamic,automatedprocessescantriggeratkeymomentsonthepurchasedecisionpathandengagecustomersonthepreferredcommunicationchannel.
5. AnewevolutioninCRMmanageshyper-relevantandcontextualcustomerinteractions,deliversmoreefficientengagementsandoffersimmediatecustomersavings.
ProductivitySavingsforbothRetailerandCustomerAconnectedretailercanrealizeproductivitygainsacrossanumberbusinessareas,fromsupplychaintomerchandisingtomarketingactivities;further,helpresolveissueswhichretailershavebeenstrugglingwithforanumberofyears.Hereafewkeyareaswhereproductivitygainsaremostvisibleinaconnectedenvironment:StoreInventoryEfficienciesRetailersandFMCGpartnershavelongknownthatincorrectproductplacements,poorshelfmaintenanceandoutofstockconditionsallcontributetosignificantlossesinrevenues.RetailersandFMCGtackletheproblembyutilizingfieldmarketingagenciestoperiodicallycheckforcompliancewiththeagreedrange,shelfshareintheproductcategory,shareofcompetitors’shelfandpriceforeachitem.Noteworthyisthatatypicalcategoryauditsamplesonly2%to5%ofallstorelocationsatafrequencyof1timeperweekoreveryotherweek.AccordingtoECRwhenbuyerscan'tfindtheproducttheyarelookingforinitsusualplace,9%ofclientschooseanalternativeproduct,ordonotmakeapurchase.Outofstockisestimatedtocostaretailerapproximately4%ofsalesinlostrevenues.Incontrast,anIoTpoweredstorewithefficient,batterypoweredcamerasthatsendsproductimagestothecloudforproductrecognition,canprovideongoinginformationonconditionsandpredictwhenshelvesneedreplenishment.
ProductandInventorymanagementisoneofthekeyareaswhereIoTandDeepLearningcanmakeabigdifferencebymonitoringproductsandsignalingwhenerrorsoccurandreplenishmentactionsneedtobetaken,resultinginachievablegains:
• 2monthlyvisitsperstorebyafieldmarketingrepresentativeatayearlycostofapproximately$1,500percategory/storecanbesaved
• merchandiseplacementerrorsacrossallIoTconnectedstorescanbereducedby50%ormore
• timelystockreplenishmentcanreducelostsalesfromoutofstockproductsby1%-2%
CustomerCentricMerchandisingThe‘onesizefitsall’planogramdeployedacrossallstoresfailstoconsiderthatconsumersandtheirshoppingbehaviordiffersbypointofsaleandmanyotherfactors.
Didmovingthebakerysectiontothefrontofthestoreresultincustomersspendingmoretimeinthestore?Didmovingthewinesectionnexttothecheesecountercreatemorecrossshoppingbetweenthose2categories?Dowehavejusttherightamountofsalespeopleintheshoedepartmentatpeakshoppingtimesand,ifnot,areweloosingsales?SensorfusionandDeepLearningcanprovidealevelofdiagnosticsandinsightsthatuncoverwhichvariablesareworkingtogethertoinfluencehowshoppersmakepurchasingdecisions.Further,suggestplanogramsandproductassortmentsthattargetshopperpreferencesduringtheirshoppingjourney,aswellasoptimizingpricingstrategiesandforecastingdemandforbettercustomerservice.Bycontinuouslydetectingshopperjourneysacrossmerchandisezonesandapplyinglearningalgorithms,analyticscanpinpointareasofassortmentoptimization,rangelocalizationandbetterproductvisibility,resultinginashopperjourneybasedstorelayoutwithimprovedshoppingmetricsandreturnoneverysquaremeterofshoppingarea.
Basedonlivestoreexamples,herearesomeofthecapabilitiesandefficiencygainsobtainedfromimplementingtrackingsensorsinshoppingareas:
• Monitoringofkeymetricsineveryshoppingzone,withclearvisibilityintoover/underperformingzones
• Measuringtheeffectsonshoppingbehaviorbeforeandaftermerchandisechangesareputintoeffect,resultinginengineeredstorelayoutplansthatincreasetraffictopoorlyvisitedzonesbyupto3%
• Reducingtimefrictioninservicezonesbydetectingcongestionandalertingtheneedforadditionalpersonnel,resultinginincreasedsalesconversionsof1-3%
• IncreasingReturnonSpaceinspecificstorezones/categoriesbymorethan2%byflaggingtheneedforspacere-allocationandrangeplanning
• Fasterreactiontimetochangesinshoppingconditionsandidentifyingprobablecausese.g.Promoareatrafficdecreasedby15%becauseoflowinventoryconditionsandtheneedtoreplenishstock
Hyper-RelevantEngagements,byDesignInfluencingcustomersbygettinginsidetheirmindsduringthepurchasejourneyrepresentsanongoingchallengeformarketers.
Withincurrentmeans,marketingdepartmentpersonnelsuperviseopportunitiesforengagingcustomersandcreatetargetedmarketingcampaignsbasedontheirbestjudgment.Inaddition,usecommunicationchannelsthatareunabletoreachthecustomeratthemomentofmakingapurchasedecision.Thesetypesoflimitationsmean,forexample,thatawineoffermayreachacustomeronlyafterashoppingtripandwhenhomedrinkingwineatdinner.However,IoTenabledretailersthatarepoweredbyDeepLearninganalyticsareinapositiontodeliverrealtimesavingsduringtheshoppinglifecycle.DeepLearningistheenginethatprovideshypercontextualandrelevantinteractionsexactlyattherightmoment,therebyaddingalayerofefficiencythatishighlyvaluedbytheconsumer
PersonalizedadtriggeredandsentwhenJane,aluxurycategoryshopperwho‘Likes’LouisVuittononFacebook,waslocatedintheLVhandbagdepartment
Withanabilitytoknowwhatcustomersarelookingforandneedtoknowatagivenmomentduringtheshoppingjourney,marketerscanachieveanunparalleledlevelof‘responsetoconversion’metrics:
• Increasedcross-shoppingbetweenzonesby8-15%
• Increasedbasketsizeby1.25%intargetedcustomergroups
• Increasedvisitrepeatrate,shoppingfrequencyby1.5%
• Increasedsalesonpromoitemsupto4%
• Increasedsalesconversionsbyfloorpersonnelupto35%
• Moreinteractionsinserviceareasbetweenstorepersonnelandcustomersby15%
ConclusionBytestinganddeployingIoTandDeepLearningforbricksandmortarstores,retailersareabletoevolvetheirbusinessinachallengingnewenvironment.
Gettingitrightentailsknowingyourcustomersinamuchdifferentwaythaneverbefore,meetingtheirexpectationsastheychangeovertimeandbecominghyper-relevantacrossalltouchpoints.
Intrinsictosuccessisbecomingmorecostefficientonalloperationsandfindingtherightbalancebetweenpricing,productassortmentandcustomerservices-allofwhichdependsonadigitalizedphysicalenvironmentcapableofdetectingandadaptingtoconditionsastheychange.
AfewwordsaboutmyselfAstheCEOandprincipalarchitectatShareMyInsight(SMI),Ihavebeeninvolvedoverthepast10yearsindevelopingproprietarytechnologiesandapplicationsforbigdataanalyticsandstatisticalmodelsonconsumerbehavior.InthelastfewyearsIhaveseenretailersincreasinglystruggletocreatemeaningfulandrelevantcustomerengagements,largelyduetotraditionalstatisticalmethodsthatarebecomingobsolete.IbelievethatsensorfusionandDeepLearningtechnologiesarenowreadytoreplacetraditionalrulebasedmodels,enablinganewtypeofshoppingexperiencethatwillbenefitconsumers,brandsandretailers.Mycurrentfocusisonthedesigntoproductioncycleofavarietyofin-storesensorsthatlivestreamdatatotheSMImachinelearningplatformfordetecting,identifyingandputtingintoactioninformationforstoreoperations,merchandising,marketingandcustomercommunications.Iworkwitharangeofpartners,fromconsultantstomarketresearch,trademarketingandadagencies,tosolutionprovidersandintegrators.Feelfreetocontactmeatpcouture@cyscom.com