Transforming Customer Relationships and Experiences Through Predictive Analytics
South Florida Interactive Marketing Association
Today’s agenda
2
De Dawkins NA Sales Leader
IBM Predic2ve Customer Intelligence
Speaking to you today…
1. How Analy2cs can add value to six key use cases in the marke2ng lifecycle
2. Iden2fy basic predic2ve analy2cs techniques and concepts
3. Define an end to end data driven, advanced analy2cs powered customer engagement architecture
4. Review a real-‐life case study
This session will cover the following areas…
Leaders leverage big data and analytics for innovation in marketing and creating a superior customer experience
3
Source: 2014 IBM Innova2on Survey. IBM Ins2tute for Business Value in collabora2on with the Economist Intelligence Unit.
3
Predictive Analytics Leveraging technology and applied mathematics to learn from the past in order to predict the behavior of individuals and outcomes of events in order to drive better business decisions.
Acquire, Grow & Retain customers by improving customer interactions and relationships by harnessing all customer data
ACQUISITION
RETENTION
PERSONALIZATION
PROFITABLE GROWTH
To create a superior customer experience and effective marketing campaigns, you must start with a complete view of the customer
Transac?onal data • Orders • Transac2ons • Payment history • Usage history
Descrip?ve data • AVributes • Characteris2cs • Self-‐declared info • (Geo)demographics
AFtudinal data • Opinions • Preferences • Needs & Desires
Interac?on data • E-‐Mail / chat transcripts • Call center notes • Web Click-‐streams • In person dialogues
WHY?
WHAT?
HOW?
WHO?
6
A Living Customer Profile
Base Customer Profile Data What We Know
What They’ve Told Us
How They’ve Responded
What They Are Doing
How They Feel
Living Customer Profile (360°)
Transactional Data
Explicit Preferences and Permissions
Contact & Response Data
Behavioral Data
Social Insights
What They’ve Purchased
Predictive Customer Intelligence How will they Act
7
Predictive Analytics enables marketers to extract deep insights from data and better understand customers in order to send more relevant offers.
Consume greater amounts of data
VOLUME
Make sense of data more
quickly VELOCITY
Amalgamate more types of data
VARIETY
Examine and validate uncertain data
VERACITY
Data mining: The self-organizing use of algorithms to
interrogate data and uncover hidden patterns, associations, and key
predictors. Great for large data sets.
“Who are the most likely consumers of organic granola bars, and what else
do they typically buy?”
Statistical analysis: Tests hypotheses about your data to drive
confidence in business decisions
“I think 35-year old single women in urban metro areas are the largest
consumers of organic granola bars.”
8
Type Classification
Identify attributes causing likelihood of something occurring
Segmentation Find patterns and clusters of similar
things, and outliars
Association Discover associations, links, or
sequences in your data
Types of models
Rule deduction, Regression, Time Series, Decision, Trees, ANN, SVM,
KNN, ...
K-Means, Kohonen SOM, Correspondence Analysis, Anomaly
Detection, ....
Association, Sequence, Correspondence Analysis,......
Examples
§ What signals a customer leaving? § How many umbrellas will we sell in
the next three months in Chicago?
§ Who is likely to respond to a marketing campaign?
§ Which insurance claims should we investigate?
§ What products are purchased together?
§ What is the series of clicks on my web page that leads to a sale?
Use to
Build alerts for call centers to take corrective action on customers identified as at risk for going to a competitor.
Increase ROMI and reduce opt-out rate by reduce the number of people you market to by selecting only those most likely to respond.
Increase average sales by building campaigns and promotions that combine items offered or provide recommendations for purchase
Algorithms find the relevant data among the noise
9
Example models for customer analytics
• Propensity Modeling, Campaign Response Models, Product Affinity Models, Up Sell/Cross Sell Models – Knowing who is most likely to respond to a campaigns, offers or product recommendation increases campaign returns without increasing cost. It reduces customer fatigue by not bothering customer with unnecessary messaging.
• Churn Models – Knowing who is likely to attrite, cancel contracts or buy from competitors allows customer communication to be oriented to retaining the customer.
• Customer Value, Life-time Value – Knowing which customer are valuable or have the potential to be valuable changes the way markets will communicate to them and what incentives and programs should be aligned.
• Segmentation Models – Segmentation models cluster customers into homogenous groups for improving marketing tests and align offers based on common behaviors.
• Pricing Sensitivity – Insure marketing incentives are a aligned with customers sensitivity. Protect margin by not discounting products to customers that are not driven by price.
• Sentiment Analysis – Negative sentiment aligns with churn analysis above. Positive sentiment helps marketers which customer may become social advocates. § © 2015 IBM - Internal Use
10
Customers Contacted
Total Sales
0 100%
100%
Rule 1: Target Hot Leads (Life Events, Enquirers)
Rule 2: Affinity Targets
Rule 3: High Value Mul2-‐Buyers
Rule 4: Exclude “Bad” Prospects
50% Coverage = 50% Total Sales
100% Coverage = 100% Total Sales
Baseline Gains
Rule Gains
Marketing Segments and Predictive Models Working Together – Gains Chart
Customers Contacted
Total Sales
0 100%
100%
Some improvement due to beVer op2miza2on of exis2ng rules
Most improvement ader core rules are exhausted
Some improvement through beVer exclusion of weak prospects
40%
70%
Rule Gains
Baseline Gains
Marketing Segments and Predictive Models Working Together – Gains Chart
Predic2ve Model
1. Customer Intelligence & Insight
6. Marke?ng Offer Selec?ons
Creating an analytically-powered marketing platform: six key use cases
13
5. Real Time Customer Analysis
2. Campaign Targe?ng 3. Campaign Automa?on (in-‐line scoring)
4. Marke?ng Op?miza?on
1. Customer Intelligence & Insight
14
Generate a more complete 360-‐degree view by amalgama2ng mul2ple, disparate data sources and appending predic2ve insights. Advanced analy2cs finds hidden pa]erns and predictors in large amounts of structured and unstructured data that are most relevant to customer profiles.
Use Case #1: Know Your Customer!
2. Campaign Targe?ng
Advanced analy2cs models help improve accuracy of targe?ng. This allows markers to send fewer offers with higher predicted conversion rates, lowering marke?ng costs and improving ROMI.
Use Case #2: Present Offers and Messages that Resonate
15
3. Campaign Automa?on (in-‐line scoring)
Predic2ve Customer Intelligence scores can be embedded in Campaign flows and scored at any 2me during campaign processing, making analy?c sophis?ca?on immediately available to the marke2ng lifecycle.
Use Case #3: Automate Campaigns
16
4. Marke?ng Op?miza?on
Combine predic?ve analy?cs scoring to reveal likelihood of certain events (e.g. propensity to accept an offer, risk of aVri2on, etc.). Evaluate predic2ve scores alongside business constraints and within business rules to op2mize decisions.
Use Case #4: Optimize Through Business Rules, Constraints, and Analytics
17
5. Real Time Customer Analysis
Predic2ve Customer Intelligence’s real 2me scoring engine allows the power of the deep algorithms to be introduced at the moment of impact, including the inclusion of contextual data -‐ informa2on collected as the interac2on is happening. This again adds depth and accuracy to the understanding of the customer profile, which supports campaign execu2on.
Use Case #5: Interact in Real-Time and Considering Context
18
6. Marke?ng Offer Selec?ons
Predic2ve Customer Intelligence scores provide an alternate recommenda2on for marketers to consider alongside standard naive bayes/self learning algorithms for offer selec2on, grounded in mul?ple algorithmic techniques that examine many dimensions of data. This empowers the marketers with op2ons that may improve accuracy of offer selec?on.
Use Case #6: Add Predictive Layers to Offer Selection
19
STEP V Measure & Refine
Business Intelligence Engine
STEP II Generate Insights
Customer Intelligence Segmentation
Offer Propensity Churn risk
purchase predictors Customer profile
Etc…
STEP I Gather Data
Data Integration
Customer Analytics Platform
STEP IV Act
Delivery
STEP III Decide
Campaign Execution
Campaign Targets
Customer analytics produces data for targeted campaigns Predictive INSIGHTS PROFITABLE ACTIONS
Real-‐Time Push
Batch Real-‐Time Interac?ve
Real-‐Time Campaign Cross Channel Offers
Event
Offer
Channel
20
Acquisition models Campaign response models Churn models Customer lifetime value Price sensitivity Product affinity models Segmentation models Sentiment models Up-sell / Cross-sell models Etc.
Campaigns Offers/Messaging Customer experience design Omni-channel campaign management Contact optimization Real time marketing Lead nurturing Marketing event detection Digital marketing
Customer insights drive optimized, integrated decision making
Big Data Predictive Customer
Insight Real time or historical Enterprise Marketing
Solutions
Chat
Voice Email/SMS
Social media
IVR & Call Center
Web and Mobile apps
Outbound, Mail, etc.
Omni-channel Customer Interactions
Integrated Decisioning
Shared Contextual View of the Customer
HOW? Interaction data • Email & chat transcriptions • Call center notes • Web clickstreams • In-person dialogues
WHY? Attitudinal data • Opinions • Preferences • Needs and desires • Sentiments
WHO? Descriptive data • Attributes • Characteristics • Self-declared information • Geographic demographics
WHAT? Behavioral data • Orders • Transactions • Payment history • Struggles • Interests
POS, Kiosk ATM
21
Communications provider C Spire Wireless uses predictive analytics and decision models to optimize cross-selling and prevent churn
Business Challenge ⏐ Outcompete the resource-rich wireless giants, C Spire Wireless needed to beat them at the small things that matter most: getting closer to customers and keeping them satisfied. Its challenge was to convert what it knows about customers into actionable insights that help account reps craft the optimal offers that meet their needs and head off customer dissatisfaction.
Smarter Solution ⏐ C Spire Wireless is using predictive models to examine the complexity of its customers’ behavior and determine which service mix is optimal for each customer’s need, as well as the indicators of imminent churn. By embedding these insights into its customer-facing processes, C Spire Wireless has empowered its reps to optimize their interactions with customers.
270% increase in cross-sales of
accessory products
Increased satisfaction by creating a more
personalized customer experience
50% increase in effectiveness of customer
retention campaigns
Excellent buy-in from front-line crew
Connecting more closely to customers
What should we offer this customer? • Use models to predict churn risk, propensity to respond to different offers • Use rules to enforce eligibility, policy, and regulatory compliance
“We’re not only getting a more complete picture of our customers’ needs, we’re translating those insights into a higher-value customer experience.”
- Justin Croft, Manager of Brand Platforms and Analytics
Systems of record PULSE database is constantly
updated with every customer interaction – including purchases,
demographics, and prior offers / responses
Systems of engagement Personalize interactions across all touch points Connect CRM, Web and mobile into one seamless experience
Point of Sale
Web
IVR
SMS
© 2015 IBM - Internal Use 24
Top Related