Presentation from an IBM Business Analytics seminar, held the 22th of november 2012 at IBM Client Center Nordic.
Description:
IBM has studied the success factors needed to create optimal customer experiences. Analysis is a key factor to recognize the most profitable customers, optimize sales activity and pricing as well as improve the quality of the company's encounter with the customer. We discuss how to use your customer data actively to predict and influence future customer behavior and create loyal customers.
Colin Linsky, Predictive Analytics Worldwide Retail Sector Leader, IBM
2. Agenda
Business Analytics – The Competitive Advantage
Business Analytics in Action
– Customer Analytics
– Market Basket Analysis
– Next Best Action
The Analytics Centre of Excellence
Harvesting and Actioning Consumer Insight
2
4. Business Analytics
BI PA
What What to do
Why?
happened? next?
From Sense and Respond to Predict and Act
4
5. Predictive Analytics – What is it?
• A true analytics process is the one that transforms raw data into actionable insights, the true
transformation from "So What?" to "Now What?".
• Business Analytics is the process that transforms raw data into actionable strategic
knowledge to guide decisions aiming to increase market share, revenue and profit.
• Drive your business by making informed decisions based insights derived from analyzing
one of you most valuable company assets, data.
• Analytics takes data and translates it into meaningful, value-added options for leadership
decisions.
• Actionable, statistically supported insights from data that help drive competitive advantage.
• “By 2014, 30% of analytic applications will use proactive, predictive and
forecasting capabilities” Gartner Forecast, 2011
http://www.readwriteweb.com/enterprise/2011/01/business-analytics-predictions.php
6. Key Moments of Truth
Research and Browse
Browsing and cart use
Attract
Pre-purchase
Checkout and payment
Delivery
Multi-Channel use Grow
Sign-up to a Loyalty Program
Response to a campaign or promotion
Credit application Retain
Complaint
Claim
Customer Service Request
Fraud
Warranty registration
Blog/Twitter
Social Media
Product out-of-stock Risk
Destruction of perishables
Low velocity product sales
Demand forecast
8. Driving Smarter Business Outcomes
Capture Predict Act
Enabling a complete view of Understand customers micro-behavior Deploy predictive analytics
the customer combining across channels, predict their next within business processes,
enterprise and social media move and make the next best offer across access platforms,
based data maximizing operational
impact
Text Data Statistics …
Data Collection
Mining Mining
Deployment
Platform Technologies
Pre-built Content
Attract Up-sell Retain
…
10. Customer Life Cycle – Customer Experience Framework
Research
Product
Advocate Up/Cross Purchase
Product Sold Product
Get Customer Use
Service Product
11. Customer Life Cycle – Customer Experience Framework
Marketing
Social Intelligence Research
Product
Sales
Advocate Up/Cross Purchase
Product Sold Product
Get Customer Use
Service Product
Feedback Management
Support/Services
12. Customer Life Cycle – Case Studies
Marketing
Social Intelligence Research
Product
Sales
Advocate Up/Cross Purchase
Product Sold Product
Get Customer Use
Service Product
Feedback Management
Support/Services
12
13. Customer Life Cycle – Customer Experience Framework
Marketing
Social Intelligence Research Cost of e-mail marketing as a
Product
cost percentage of revenue
71,000 responses analysed and (CPR) was cut almost by half
online buzz increased by over Sales
400%
Advocate Up/Cross Purchase
Product Sold Product
Analyzes 30 to 40 data points
per customer to deliver
actionable insights, giving in a Delivers preventive health
Decreased churn
3.1% boost in response rate information to individuals
from 19% to just
Get Customer Use in a format that motivates
under 2% Product
Service them to take action
Feedback Management
More easily identify potentially Support/Services
fraudulent claims, increasing
customer profitability by 20%
14. Example: Predictive Analytics and merchandising
In-store promotion
decisions
Association
POS Transaction Data detection
Capture Predict Act
16. Example: Loyalty, targeting, promotions and incentives
Promotional Display
Buy X get Z for only Domain Expertise
$1.49!
Market basket
insights
• If A then B Transactions
• If C then D from all
• If E and F then G customers
• If H, then H then I
Special Offer – This Week Only
10% off on any of these
combinations: A + B…G + H….
Predictive Models
Offers
Transactions from this
% $ 1 Gillette razors customer
Statement • Cardholder since YYYYMM
insert % $ 2 L’Oreal shampoo • Average transaction value
• Monthly transaction value
3 13 % $ 3 House brand shampoo
• Categories purchased
% $ 4 House brand hair color • Brands purchased
6 12
456 % $ 5 Colgate toothpaste Descriptive
6636 • Age
% $ 6 Nivea skin care • Gender
% $ Men’s fragrance • Family situation
7
• Zip code
% $ 8 Woman’s fragrance
% $ 9 House brand sun care Interactions
Statement • Web registration
insert % $ 10 Optician • Web visits
• Customer service contacts
% $ 11 Feminine hygiene • Channel preference
12 15
% $ 12 Online photo service
773 11 3 Attitudes
9245 % $ Family planning
13 • Satisfaction scores
• Shopper type
% $ 14 Pampers diapers
• Eco score
% $ 15 House brand diapers
16
17. It’s not just about marketing - what should
we do for these customers?
18. Example: Next Best Action
Customer
Reporting,
KPIs and
Alerts
Association
Browsing Business
Rules
LTV
Transactions
Domain
Expertise
Propensity Predictive
Modeling
Products Customer
Predictive Engagement
Model Scoring
Inventory
3rd Party, CSR,
Classification
Social Media,
Survey …
Analytical
Decision
Management
Segmentation
Supply Chain
Capture Predict Act
19. The Largest Online Shopping Mall in Japan
Merchants: over 37,000
Customers: over 80 million
Top page PV: 8 million / day
# of orders: 500,000 / day
Gross Mercandise Sales (GMS): 3 billion yen
GMS growth: +18% YoY
21. The vital ingredients…
Predictive Expertise
– Models predict customer segment and category affinity
– Customer Segmentation (Funnel)
– Market Basket Analysis (Prior sales)
– Category Affinity (Products and activity – Browse/Purchase)
– Current Interaction history (What’s happening during the interaction)
– Price Sensitivity Calculations and Offers
– Inventory Based Suggestions
Decision Management
– Combine predictive intelligence with business know-how
– Prioritize offers based on profitability and propensity to respond.
– Deliver recommendations and personalizations to a website or point of sale
Business Intelligence
– Understand your current state and your potential state
– Monitor results and fine-tune your business
– Inform strategy with a view into the future
Synthesis of data sources and data types
– Overlay browsing history onto purchase history to profile customers
– Use profile to drive better recommendations, offers and actions
27. Customer analytics scenario
Data and Model Customer
Management Services
Campaigns
Data Driven Segmentation Multi-Channel ECommerce
and Profiling Deployment
Single View
Targeting Models of the Sales Tools
Customer
Customer Performance
Reporting
Customer LTV POS
Measurement Data
Quality
Ad hoc Queries
Feedback
3rd Party Data
Sources
Infrastructure
Modelling
Data
Sources Measurement
Deployment
Governance
28. Analytics Centre of Excellence:
Best practices, governance and production
Collaboration
– Analysts
– Best Practice
– Recycling
– Consumers
Model Management
– Strategic Asset
– Test & Production
– Governance
Automation and Scheduling
– Analytics as part of business process: event or time based
– Back-office actions
Scoring
– Batch
– Real (Right?) Time
Integration
– Seamless integration into existing systems and business processes
– Open, flexible and customizable
Sentiment evolution over time comparing Ariel and Persil. The positive peak in January relates to the Persil Gel launch by Henkel Arabia. The positive peak in July relates to the Persil Cleaner Planet Plan announcement in New Zealand.
From each chart one can drill down to the snippet view where concepts, hotwords and sentiment are highlighted and additional metadata is shown, including the original URL of the post.
The evolving topics are automatically detected without preconfiguration and are visualized in so-called „topic rivers“ that show the temporal evolution of topics along with the keywords that are most frequently mentioned within the snippets that make up the topics.