Professional Documents
Culture Documents
2
Why Event Based Marketing?
IBM Opportunity Detect enables event-based marketing
TraditionalBatchCampaigns Illtellpeopleaboutit
Ihaveanofferforproducts whenIchooseto
Offer
Customerinitiated,behaviordriven
Illpresentanoffernow,
Ihaveacustomerwithachangeinbehavior
whentheyneedit
thatsuggestsaneed
Offer
Offer
Offer
Offer
4
Leading practices deliver better company results
1.8x higher
The 3-yr Gross Profit Growth
top
20 % 3.4x higher
achieve: 3-yr Net Income Growth
2.4x higher
3-yr Stock Price Growth
*SOURCE: The State of Marketing 2013: IBMs Global Survey of Marketers, IBM Center for Applied Insights
5
Challenges in the market Telco customers perspective
Thetelcoindustryisfacedwithseveralchallengesmakingitnecessaryto
innovate andchangethewaywemanageourcustomers
Increasingcompetition
6
Why we need Event Based Marketing?
Differ from others with excellent customer experience management
Sales Approach through
the channels
What should I sell? Evolution Model of Customer Management
Products/services differentiated
based on individual current behavior + EVENT
in real time ORIENTATION
Products/services differentiated
based on individuals + CUSTOMER
ORIENTATION
Products/services not
differentiated all for all + PRODUCT
ORIENTATION
Marketing Approach through
MASS MARKETING SEGMENT INDIVIDUAL the clients
MARKETING MARKETING
Who should I sell to?
Impact and acquire large Cross sell to high value Develop and communicate highly relevant
amounts of prospects segments based on and personal offers using advanced
through mass campaigns segmentation variables segmentation capabilities
7
Results: Reaping the benefits of understanding customer behavior
Buildinglogicisintuitive,Campaigndeploymentisfast
Reduces time-to-market
Enablemessagingtocustomersassoonaseventsaredetected
Identifyopportunitiesimmediately
Increases conversion
Relevanceissignificantlyhigherforthecustomer
rates Conversionratesareincreased
8
Todays Business Environment
9
Customer Example: Large US Retail Bank
Customer
Replaced Harte Hanks Daily Deposit Builder solution with
Detect
Used an existing large Financial Investment ompany as a
reference site:
At the time Customer was able to move 10 new event triggers
into production per week.
Approximately 90 triggers in production
10
CustomerExample:Large US Retail Bank
% Increase
New Personal Checking
15%
Retail New Business Checking Sales
55%
New Direct Home Equity Production
29%
Commercial New Business Banking Loan
Production 33%
Deposit Outstandings 11%
Investment Loan Outstandings 55%
Banking Capital Market Fees 29%
Debt Capital Markets Revenue 30%
Wealth CIB New Business Revenue 19%
Institutional New Trust Business 11%
11
Terminology
Event Based Marketing
13
Gather and Respond Marketing versus Event Based Marketing
14
Traditional Approach
Makes Complains
Opens Makes About
Makes Medium Buys
Makes
Account Buys Medium Overdue
Small Purchase Inkjet
Large
Purchase Printer Ink Purchase Delivery
Purchase
Day 1 5 10 15 20 25 30
Act based on
accumulated
information
15
Event Based Marketing Approach
Complains
Opens Makes Makes Buys Makes Makes About
Account Small Medium Inkjet Large Buys Medium Overdue
PurchasePurchase Printer Purchase Ink Purchase Delivery
Day 1 5 10 15 20 25 30
16
Events versus Triggers
Event
anything that happens, especially something important or unusual
Trigger
an event, situation, etc. that causes the start of something, especially
something bad
or
anything, as an act or event, that serves as a stimulus and initiates or
precipitates a reaction or series of reactions
17
Guiding Principles
Trigger Matrix Principles
High Smart Advanced
Intelligence
Simple Complex
Low
Intelligence
Non Salary Credits
Credit Card Payment Due
Regular Cheque Deposits
Salary Decrease Small Business Payments
Payment to Movers
Salary Increase PIN Wrong
Card Captured at ATM
Paying Council Rates
Viewed Credit Card Page
Travel card Top Up Tax Bracket Change
Low Account Balance < 0
21
Characteristics of Triggers
Forward Looking Inactivity
Fire when a customer does X but then not Y within two weeks
Detect any of the patterns components
Fire when a customer does either X or Y or Z Backward Looking Inactivity
Fire when a customer does X without having done Y in the
previous two weeks
22
Getting Started
Makecustomer Dowhatyou
Focusonusecases goalsyourbusiness canwith
thatrequireatimely goals availabledata
response
Buildtoyour Domoreas
capacityto knowledgeof
Expandon
respond technologydeepens
successful
detectionpatterns
23 23
Technology
What Makes an Event Detection System
Real-Time Feed
Operational System
Database for
State &
Server 1 Outcome
Server 2
25
What Makes an Event Detection System
Real-Time Feed
Operational System
Database for
State &
Server 1 Outcome
Server 2
26
Event processing consists of
Transaction input
Event detection
Pattern detection
Action triggering
27
Wheres the Data Warehouse?
28
What is a State
29
Why SQL is not suited to EBM
30
Why SQL is not suited to EBM
The current weekly sum of Amounts exceeds the
average of weekly sums over the last 3 months by 2
standard deviations
31
Technology is only part of the solution
32
Data
Types of Data
34
Data Requirements
Other data attributes will enhance the variety of events that can be built
35
Typical Transaction File Types
36
How does it all work?
If a high-value customer who has been trading very actively over the last two weeks and has an increasing
account balance over the last year suddenly makes a $50,000 withdrawal from his account, then have his
broker contact him immediately
Trigger
Source Attribute Action
Events Patterns Target
Systems Filters Triggers
Systems
Increasing
Monthly Account
Over Last
Statements Balance
Year
Outbound CTI
Brokerage > 10 Trades in Sequence of High Value Broker System
Trade
Feed Last 2 Weeks All Three Customer Contact /Broker
Desktop
Opportunity Detect 37
I Get it. Now What?
Getting Started with an EBM Program
39
Workshop your Ideas
Start SIMPLE
Leverage Analytics to support ideas
Service events are often the easiest
have a positive effect on customers
have a positive effect on the business (hopefully)
Review trigger libraries from IBM and other sources
Align action with detection
a significant X. so what
Refer back to the Guiding Principles and develop your own!
Get guidance and support from an expert to help avoid the pitfalls
40
Event Detection Libraries
41
Build and Test
Building triggers can take minutes, hours or days depending on many factors
Steep learning curve but its NOT a technical challenge
Testing is crucial
Can take 3 5 times longer than build for complex triggers
Should be performed by people other than the event builder
Expected outcomes at every logic step must be well documented
Test on fake data first
Custom build data inputs to test around every logic step
Test on production data second
Nothing beats real world data to see what is going to happen when you flick the switch
Test with other events last
Does your event play nicely with others?
What happens to system performance and volume of triggers fired?
42
Marketing versus Customer Service
Emerging Imperatives of Marketing
Understandeachcustomerasanindividualandincontext
Innovateandscalepersonallyrelevantandrewarding
experiences
Designyourcultureandbrandsotheyareauthenticallyone
44
Current Trends in Event Based Marketing
45
How can I retain Max as a customer?
MaxtalkstohisbrotherFreddaily
FredmovestotheUS,MaxremainsinGermany
MaxnowcallsFredintheUS3timesaweek
Maxisatrisk!
Maxscurrentcellphoneplanisnot
optimalforcallingabroad
Maxscallingchargeswillrisesharplyand
dissatisfactionlooms
46
2015IBMCorporation
Marketing understands the what and why
MarketingmakesapersonalizedoffertoMax
Automaticallydetectsthechangeincallingbehaviorandmakes
anofferformoreoptimalplan
ButMaxignoresofferandisstillatrisk!
Maxdoesntrealizethesignificanceoftheoffer
Marketingofferrenewedandaccepted!
Marketingdetectslackofofferacceptanceand
renewstheoffertocoincidewithbillingcycle
TheofferisacceptedandMaxremainsa
satisfiedcustomer
47
How might you accomplish this with technology?
emailmobile
push
48
IBM Opportunity Detect Trend Components
Event DynamicOffer
detection Generation
Working in the application
emailmobile
push
Sarah captures from the CDR the minutes used on each of Maxs calls
Sarah then specifies in the Exceeded Standard Deviation Spike
component the conditions for a spike in call volume
When Max meets the conditions a notification is sent to the real-time
offer generation system
49
Example: Checking for jump in call volume using the Exceeds
Standard Deviation trend component
50
IBM Opportunity Detect Inactivity Detection
Event DynamicOffer
detection Generation
emailmobile
Working in the application push
51 51
Watching for events to not happen
Meetingconditionsforoffersets Timereventfiringindicatesoffernot
timer accepted
Acceptingofferstopstimer
5/13/2016 52
IBM Triggered Messages Opportunity Detect gateway to Interact
Event DynamicOffer
detection Generation
emailmobile
Working in the application push
Sarah makes the outcome of the trigger send a real-time second offer
53 53
Interact Triggered Messaging
54
Lets make sure Anne gets what she needs!
ExactlyWhatYouWantisamerchantthatsellsurbanclothingandhasbothphysical
storesandawebcommercesite.Anneregularlypurchasesitemsfromthestoreandonline.
Annevaluesherrewardcardandidentifiesherselfwhenmakingpurchases
AnnehasarecenthistoryofbuyingitemsOnSalebuthasjustpurchaseda
Springcoatinthestoreatfullprice.
Anneisnowbrowsingthewebsiteforaccessories
OpportunitytoincreaseAnnesdelight!
Apurchaseofamajorgarmentthatbreaks
apatternofOnSalepurchasescould
indicatethatAnneismakingawardrobe
changeandeagertocompleteit
55
Lets make sure Anne gets what she needs!
MarketingmakesapersonalizedoffertoAnne
Automaticallydetectsnewpurchasingbehaviorandinformstheonlineoffermanagement
systemwhichcanautomaticallycouplethatinformationwithbrowsingbehaviour
WhenAnnebeginstolookforaccessories,thecombinationofpurchaseandbrowsing
behaviorcreatesanofferthatleadsAnnetoaccessoriesthatcomplementhernewcoat.
Annefollowsthroughontheoffer!
TheacceptanceoftheofferdeepensAnnes
relationshipwiththemerchantand
improvesthelikelihoodofmakingmorefull
pricepurchases
56
How might you accomplish this with technology?
Onlinepattern
detection
Instoreor
website Eventdetection Realtime OnlineOffer
Generation
Customer
purchaseevents Onlineoffertailoredfor
customerbasedon
Offlinebehaviorpatterns
(realtimeorbatch)
Inlinebehaviorpatterns
Mobileorweb
Offer&Responsehistory
onlineoffer
Segment
57
Current trends in event detection
58
Events to deepen the relationship
59
Opportunity Detect Roadmap
10.0 Feature: Groupby aggregation
Aggregationsetting:by
timestampand
Example transaction flow transactiontype;
Amountsaresummed
Day 1: $25, Deposit
Day 1: $50, Withdrawal Exampleshows
aggregationthatoccurs
Day 1: $75, Deposit
basedonthetransaction
flowforDay1
Date Transaction Type Amount
Trend components execute complex queries on Containers and Selects that do time slicing of data
into historical and current periods; and then check if a growth or decline relationship exists between
these periods
Sequence pattern - detects events occurring in a particular temporal order
Regular expression matching
e.g., detects if an incoming bank transaction has payroll with a description field
Import/Export
a utility for transferring logic from one environment to another
62
Candidate features for 2017
Geo-location detection
Whatisit?
Awaytodetectgeolocationeventsandincorporatethemwithinthetriggersystem
Auserwillbeabletospecifyasetofnamedpolygonlocationsusinglatitudeandlongitudecoordinates
Fortransactionsthatincludelatitudeandlongitudeuserlocationinformation,theapplicationwillbeableto
determinelocationeventslike:1)useriswithxfeetofapolygon,2)useriswithinapolygon,and3)userisclosest
tolocationXYZ
Whataretheanticipatedbenefitsforourcustomers?
Customerscandetectcomplexbehaviorsthatincludelocationevents,like
CustomervisitedcompetitorstoreXtimesin3days
Customervisited3ofmystoresintwoweeks
CustomervisitedcompetitorstoreinageographyXmilesfromhome
Customervisitedastore3timesinonemonthandthenfailedtoenteranystoreinthenext2weeks
64
Example: Geo-location detection from telco
Competitor
Offer
65
UBX (Universal Behavior Exchange) integration
Whatisit?
1. OpportunityDetectpublishestriggeredbusinesseventstotheUBXeventbus;thusmakingitpossibleforany
UBXconsumerstoutilizedetectedcustomerbehaviors
2. PatterndetectionengineforUBXevents
1. UserdefinespatternsbasedonUBXevents
2. OpportunityDetectconsumestheseeventsatruntime,detectpatterns,andpublishesthedetected
patterns
3. ThesepatternsofUBXeventscarrymoreweightthanindividualevents,whichcouldberandom
Whataretheanticipatedbenefitsforourcustomers?
1. CustomerwillbeabletosenddetectedbusinesseventstoCampaign,Tealeaf,Engage,MobilePush,orany
applicationparticipatingintheUBXapplicationeventbus
2. CustomerwillbeabletouseOpportunityDetecttoobtainpatternsofUBXevents,like2Cartsover$50
abandonedinthelast10minutes
66
Example: Detecting patterns of UBX events