R3 : Managed Insights to Actions Enabling a more effective

Report
Deloitte Analytics
Enabling a more effective, proactive
marketing organization
May 2014
Marketers today are looking to improve performance and reduce churn
through an enhanced customer experience
The competitive advantage:
Real-time actions, tailored to each customer
High value
customer
segments
Exhibiting
attrition
behavior
Customer
interaction
preferences
Customized
messaging to
connect with
customers
Real-time
offers to
encourage
conversion
Unique
interactions to
build loyalty
Automatic action allows you to proactively own the customer relationship
2
Case Study 1
At Client X, member retention was a strategic priority and a predictive model
helped to close the gaps between member renewal targets and attainment
Background
 Increasing membership retention is a strategic priority; Revenue from membership represents
approximately 53 percent of Client X’s net income
 In recent years, the membership base has stagnated and renewal rates for new members are low
 There is a gap between the forecasted member revenue plan and actual member revenue growth
 Understanding membership renewal patterns and reasons will help close the gap
Key Business Objectives
Project Objectives
 Short Term: Identify gaps by membership
type that can be filled to bring membership
income to plan
1. Identify members most likely to attrite
 Long Term: Inform development of a
membership renewal strategy
3. Of these members, identify most likely
responders to intervention strategies
2. Understand attrition drivers by member
 Additionally, provide valuable information
to the Winback team to optimize budgets
4
Insights and tools generated by the project enabled Client X to improve
membership retention
OUTCOME (RESPONSE)
DELIVERABLES
DRIVERS (EXPLANATORY VARIABLES)
BUSINESS QUESTIONS
Tools to enable
 Member renewal
indicator (Prediction
target variable)
 Member renewal
likelihood score
 Member renewal
reason Codes
 Intervention
Responsive-ness
scores
Member Characteristics
 Age
 Gender
 Income
 Tenure
 Distance To Club
 Cohort
 Education Level
 Investor Likelihood
 Dwelling Type
 Family Type
 Occupation
 Yrs. at Residence
 Donates Money
 Acq. Month
 Card Type
 Upgrade Ind
 Downgrade Ind
 Renewal Type
Member Behavior
 Promo Participation
 Promo Response
 Number of renewals
 Pct. on-time renewals
 Pct. late renewals
Member Interactions
 Purchase Amt 6/12 mos
 Upgrade in 6/12/24 mos
 RFM Decile
 RF Decile (50/50)
 WRFM Decile (39/60/1)
 R / F / M Decile
 # Unique cat Shopped in last
6/12/24 mos
 Cat Shopped in last 6/12/24
mos
 Activity by Channel
Store Characteristics
 Location Type
 Store Size, Tenure
 Number of Employees
 SIC Segment, Micro / Metro
 Comp in 10 miles
 Restaurant in 10 miles
 FIC – FY 11/12 Renewal
 Micro / Metro
 Portfolio–2010/11 Renewal
 Proto Size
 Remodeled in 2010/11
 Client X Combo




Base data exploration
Univariate analysis
Churn predictions exploration
Interventions targeting
Insights by member



Renewal scores
Renewal Reason codes
Intervention lists
Frameworks to enable actions from
insights



Periodically score members
Periodically recalibrate
Test plan
5
Definitions and assumptions within the modeling approach
Definition of
Attrition
 For the purposes of the models, attrition was defined as a failure to renew by 60
days following a member’s DTR date
 Member behavior was examined by quarter and normalized for each member’s
Normalization of
unique member year
Member Year
 Q1 for each member refers to the 1st through 3rd months of their membership (not
Q1 of the calendar year)
Model Scope
“Category
Groups”
 Five models were built:
o 1 model that predicts attrition for the entire member population
o 4 sub-models differentiated by member type (Advantage vs. Business) and
tenure (first-year members vs. tenured members) used to explain reasons
for attrition risk
 For the purposes of examining category purchase behavior, five “category groups”
were looked at:
o Consumables excluding o General Merchandise
Snacks/Candy
o Gas/Car Wash
o Snacks/Candy/Tobacco o Miscellaneous/Other
6
The suite of deliverables allowed Client X to both act upon the insights
generated from the analysis, and re-run the models in the future
1
Attrition
Models
2
Model Code
Model
Outputs
Member Scores and
Reason Codes
3
Use
Cases
Model Usage Examples
WMS
Model Formulas
Tableau Visuals of Scores
and Reason Codes
7
The attrition model was used 6 months before a member’s DTR date –
enabling Client X to identify potential attriters and intervene early
 The model predicts well in the highest risk groups (deciles 8, 9 and 10) when run at the DTR date
 When run at 6 months from DTR, the model retains a large portion of its predictive power
Run for members at DTR date
Decile Gains: 1.55
Membership (Thousands)
200
Run for members 6 mo. prior to DTR
2.24
3.13
200
180
180
160
160
140
140
120
120
100
100
80
80
60
2012 attrition
rate = 26%
40
20
Decile Gains:
1.55
2.13
2.65
8
9
10
60
40
20
0
1
2
3
4
5
6
7
8
9
10
1
2
3
Decile
Predicted to renew
4
5
6
7
Decile
Predicted to attrite
Renewed
Predicted to renew
Predicted to attrite
Attrited
8
The model also explained why high risk members are at risk. Client X used
this insight to develop intervention options
This graph depicts how often each factor appeared within the top 5 most significant predictors for each member in the
three ‘at risk’ deciles (i.e. top 30% most likely to attrite)
All Member Population
1•
Among the most at-risk members, lack of
shopping (from a $ perspective) or a
decrease in shopping across quarters are
the most common signals of attrition
2•
Lack of purchases in General
Merchandise also is a common factor
behind the at-risk population
3•
The next-most-common signal is a
tendency to buy very few items, but
relatively high-margin ones (i.e., spend
high $ on a big-ticket item, then attrite)
Top Reason Codes for High Risk Members in 2013 (Deciles 8, 9, 10)
Percent of Members in Deciles 8-10
0%
5%
10%
15%
20%
25%
Member shopped very little (in terms of $ of purchases) during year
1
Member shopped a lot in membership Q1 but very little in other quarters
Member does not buy much general merchandise
Member purchases low quantities overall, comprised primarily of high margin items
Member does not exhibit behavior of primarily purchasing high quantities of low margin…
Member did not shop a lot in Q4 of member year
Member purchases high quantities overall, comprised of mostly low margin items
2
3
Member is not assigned to a club that has one more more of BJs / Costco / Restaurant…
Member is assigned to a club that has one more more of BJs / Costco / Restaurant…
Member is likely not a minority small business owner
Member does not exhibit behavior of primarily purchasing high quantities of low margin…
Member is likely a minority small business owner
Member shopped a lot in membership Q4
Member does not make a lot of misc. item purchases
Member did not make a lot of visits before 11 or after 6
Member shopped a lot of general merchandise
Member had downgraded from PLUS in the past
Member makes a lot of visits before 11am and after 6pm
Member likely has a Sams credit card
What interventions can Client X take 6
months prior to the member’s DTR date to:
 Re-engage shopping frequency and
spend?
 Drive purchases / spend in the General
Merchandise category-group?
 Introduce ‘one and done’ members to the
rest of the club experience?
9
Case Study 2
Case Study 2 – Customer Churn Management
LARGE GAMING COMPANY
Topics
Segmentation analytics, Cost-to-serve analytics, Pricing Analytics
Key Issues
•
•
How is customer lifetime value measured under new a business model (retail  subscription)?
What would be optimal offers, prices, bundles, promotions, and payment model for customers?
Data
•
•
•
Analyzed 100K+ customer purchase and online gaming usage data points
Conducted 55+ customer interviews in US & EU
Enhanced input data points captured to enable analysis (e.g., added new fields in systems)
•
Identified groups of customers that displayed differentiated valuation for game benefits and their price sensitivity to establish different
customer segment types
Determined likelihood of drop-rates in subscriptions and tailored marketing messages by segment
Analysis
•
•
•
Utilized churn rates to determine inflection points in customer lifecycle that triggered different targeted marketing programs (e.g., optimizing
subscription models, driving micro-transaction content revenue)
Used insights to drive development of new offerings and pricing strategy
•
Reduced customer churn, improvement in SKU uptake at more optimal prices
Decisions
Impact
DISCUSSION TOPICS
Potential insights to be gained from analytics:
• How do you formulate consumer segments based on customer value?
• How do consumers’ purchase paths relate offline and online, and how can
marketing influence them at appropriate stages with right messages?
• What type of engagement is most relevant for consumers at certain
deflection points (e.g., price, offer, brand promotion)?
Potential operational considerations:
• How do companies need to restructure the way customer data points are
captured in order to be able to perform customer lifetime value and cost-toserve analytics?
Customer Lifetime Value Example
Case Study 2 (cont.) – Reduce Customer Churn with Lifetime Value Analysis
CHURN RATE ANALYSIS EXAMPLE
ILLUSTRATIVE
75
75
% of
Customers
Remaining
100
% of
Customers
Remaining
100
50
50
25
25
Starting with a 1 month non-online sub
*
Starting
with a 1 year non-online sub
0
0
2
4
6
8
10
12 14
16
18 20
22 24 26
Starting with a 1 month Online sub
Starting with a 1 year Online sub
0
28
30
0
2
4
6
8
10 12 14 16 18 20 22 24 26 28 30
Months after Initial Subscription Purchase
INSIGHTS
• Churn is higher for customers who purchase via retail
• Non-online buyers can be moved online by driving awareness of the benefits of the online options
• Moving non-online customers to online subscription increases Customer Lifetime Value by $100 on average

similar documents