Digital Wallet

UseCases

UNDERSTAND • Customer • Createa360-degreeviewofcustomeracrossdemographics,productholdingand CUSTOMERS genome productusage.  Consolidatedatafrommultiplesourcestogetanintegratedviewofeverycustomer  Createfeatures/derivedvariablestomakeinferences oncustomersbehavior  Createaninteractivedashboardtovisualizecustomerbehaviorandviewrecommended products, interventions Age Balance Transaction Merchant Attribute Income Info s Demo Assets Transaction Location Prod Holding Liabilities Index Usage CUSTOMER Investments Usage Website Channel Product Merchant Up-Sell Cross-Sell Interest nt ATM Branch

ACQUIRE • Taste-led • Identify lifestyle characteristics of profitable wallet customer base to run acquisition CUSTOMERS Acquisition campaigns  Identify thebestcustomersbasisprofitability andofferusage Charlie’s Restaurant  Identify tasteattributes ofsuchcustomers. andBar  Providelistofdemographics,tasteattributesfor SegmentX targeting and list of merchants foracquisition Indicatesaffinity Age:25-30 | Location: Bayfront,Singapore offer Preferences: European cuisine | Bars | Sea-food Facebook/GoogleAdWordscampaignis runonSegment Xwith acquisition offer for Charlie’s Restaurant and Bar ACQUIRE • OrganicReferral • Identify right customers toreferandright incentive forreferral CUSTOMERS  Createclusters basedonNetworksize(L1/2/3),NetworkRangeandStrengthofRelationshipsbyleveraging multiple algorithms -K-Means,AffinityPropagation,HierarchicalClustering(Agglomerative/Divisive), DensityBased(DBSCAN, OPTICS), Distribution Based(GMM)  Identify right set of customers (CustomersgetCustomers)and employees (EmployeegetCustomer)totargetforreferral  Identify lifestyle tastes and recommend relevant offerstoincentivize referral

ACTIVATE • Activation • Activate new to wallet customers and revive dormant customers with relevant CUSTOMERS campaign recommendations  Createcustomercohorts–‘newtowallet’usersand‘vintageuserswhohavebecomedormant’  Identifykeydrivers ofactivationforeachcohort–demographics,location,walletbalance,featureusage, offer usage across vintages -3days, 7 days, 14days, etc  Targetcustomerswithrelevantmessagingonthe‘keydriver’throughalternatechannelslikeEmail,SMS

ENGAGE • Lifestyle • Recommendrelevantlifestylemerchantsandproductstodrivespends CUSTOMERS recommendations  Identify relevant merchants, products forevery customerbyleveraging maya.ai tastegraphanduser’sdigital interaction behavior, transaction data  Optimize the ‘rank order of preference’ for merchants and products for each individual customer basedoncontext-location, timeofday,dayofweek ENGAGE • Financialproduct • Identify customerswithpropensity totakeupfinancial products(loan,insurance, etc..) CUSTOMERS recommendations  Identifycustomerswhohavetakentargetproductwithoutprompting  Identify behavior for6monthspriortotakinguptheproduct  Matchcustomers in theportfolio whoexhibit similar behavior based onidentified characteristics andrecommendproducts  Alternatively, createX-sell models forthevarious products offered andrankordertheoneswithhighpropensityscore

RETAIN • Silent Attrition • Identify dropinengagementlevels,triggerstoidentifyrighttimetointervene. Identify CUSTOMERS relevant promotions to win-back customers  Identify customercohortsusing transactional, demographic anddigital interaction data  Identify signals that indicate silent attrition (for e,g, drop in balance, decreasing frequency ofusage, poorapp rating)  LeverageAI/MLmodelslikedecisiontreeto identify customers in eachcohortatrisk ofchurn  Targetwithrelevantofferstowin-backcustomers

MANAGE • Credit Risk Scoring • Createriskscoretoestimateprobability ofdefault RISK  CreateaRiskScorebasedonapplicationdataforanewcustomer  CreateaRiskScorebasedonapplicationformdataandwallettransactionbehavior  Risk modelbuilt using:  WOEbinningandIVbasedfeaturesselection  EnsembleModelling MANAGE • KYCFraud • Identify frauds based onKYCchangedetectionandwalletusagebehavior RISK detection  DetectfraudsbasedonKYCchange,largeoutflows/cashoutsbasedongeolocation  Build sequencemodelstoidentifythesequenceofeventsresultinginFraud  Example:KYCChange->longdistancewithdrawal

MANAGE • AMLdetection • Detectunusualcustomerbehaviorandanomaliesinmoneymovement RISK  Createcohortsofcustomersbasedontransactionamount,inflowsource,outflowsource,timeoftransaction, location  Identify lookalikes for each segment MANAGE • Float optimization • Identify optimal level offloat to run wallet while enabling opportunity to earn interest RISK income  Createclustersofcustomersbasedonvarianceinfloat,inflow/outflowchannels,inflowandoutflowmix,number ofinteractions, min/maxdrop,tenureetc.  Predict wallet balance andtransactionbehaviorforcustomersineachcluster  Estimate the right float % needed to gain time and earn interest

UNDERSTAND • Merchant • Createa360-degreeviewofmerchantacrosslocation,operations,inventory,and MERCHANTS genome transactions.  Consolidatedatafrommultiplesourcestogetanenrichedviewofeverymerchant  Merchantsassignedtooneormoregroupsbasedonattributesoflocation,operations,inventory,and transactions.  Categories Inventory  Products  Day, Time Location Transactions  Volume  Value  Avgticket size  Region  City  Postal Code  Latitude,Longitude  #Employees Operations  Years since Inception  Legal Status

ACQUIRE • Territory • Identify potential merchants toacquire in everygeo- MERCHANTS Management location HyperLocalMarketSizing&Targets HyperLocal‘TastePrints’ Sizing  Mobile-based territory management of key metrics with built-in daily feedback  Equipterritory managersto plan, acquire& managecustomer/merchants toreach100% Inclusion goals Inactive  A/BTesting Framework&Models forDriving Active Monthly Active Users (Customers&Merchants)  DailySalesPerformance Trackingfor The territory management console shows Opportunity basedontaste: merchants &customersby territory managers current penetration across merchants, agents and teams and customers  2000customersin theareahave anaffinity for transacting at the restaurant  ML-basedterritory-level pricingsimulator to  $4,500is thepotentialvalue that can be measure, modify &drive revenuesacross generated fromthese customersin amonth territories

ENGAGE • Customer • Empowermerchantswithinsightsoncustomersandenablethemtotargetthose MERCHANTS Targeting specific customers  Identify and target right customer segment totargetformerchantsbyleveraging maya.aitastegraphanduser’sdigitalinteraction behavior,transaction data  Optimize the ‘rank order of preference’ for merchants and products for each individual customer basedoncontext-location, timeofday,dayofweek ENGAGE • Financialproduct • Identify merchantswithpropensity totakeupfinancial products (loan,insurance, etc..) MERCHANTS recommendations  Identifymerchantswhohavetakentargetproductwithoutprompting  Identify behavior for6monthsprior totakinguptheproduct  Matchmerchantsintheportfoliowhoexhibitsimilarbehaviorbased on identified characteristics andrecommendproducts  Alternatively, createX-sell models forthevarious products offered and rankordertheoneswithhighpropensityscore

RETAIN • Silent Attrition • Identify dropinengagementlevels,triggerstoidentifyrighttimetointervene. Identify MERCHANTS relevant promotions to win-back merchants  Identify merchantcohortsusing transactional, demographic anddigital interaction data  Identify signals that indicate silent attrition (for e,g, drop in value, decreasing frequency ofusage, poorapprating)  LeverageAI/MLmodelslikedecisiontreeto identify merchants in eachcohortatrisk ofchurn  Alert area sales/ RMs to potential at-risk merchants  Win-backmerchantsthroughbetter offer/product matching andhighervisibility