avoiding the “break-up”: a data-driven approach to increasing engagement and reducing churn
TRANSCRIPT
Avoiding the “Break-Up”:A Data-driven Approach to Increasing Engagement and
Reducing ChurnJim ForemanStaples, Inc.
The Dreaded “Leaky Bucket”
Why the Leaky Bucket Isn’t Really Truein a Data-Driven Company
Customer Behavior• Unlike water, customer behavior is unique and not random• It is based on individualized attitudes, wants, needs, and relationships• Most companies that group customers into generic segments or attempt to engage customers with a vanilla, one-size-fits-all approach will pay a price in terms of churn
Understanding A Break-Up• Whether in business or our personal lives, most Break-Ups have two root causes:
or 2. Unmet Needs 1. Misaligned Expectations
Misaligned Expectations• We have implicit or explicit expectations of every relationship in both our business and personal lives
EXPECTATION Personal Business
Honesty/Trust √ √
Physical √ √
Emotional √ ?
Experiential √ √
Financial ? √
Loyalty √ ?
• Few relationships are successful in the long term if the majority of each party’s expectations are not aligned
Misaligned Expectations: Who’s To Blame?• “You can’t please all of the people all of the time…” – though people and businesses often go to great lengths to artificially increase their attractiveness
Misaligned Expectations: Who’s To Blame?• “You can’t please all of the people all of the time…” – though people and businesses often go to great lengths to artificially increase their attractiveness
Unmet Needs• Even when initial expectations are sufficiently aligned, the deepening of any relationship may reveal wants/needs that are not (or no longer) being met• All relationships evolve over time – in order to succeed, both parties must evolve and adapt to the changing needs of the other party• In business, nirvana is a customer who:• Has an emotional connection to your company or brand• Feels like “…they really ‘get’ me!”• Becomes a brand evangelist and tells others about it
The Challenge• How can we sort through all of the “noise” to:
• Better-align with customer expectations• More deeply engage customers by evolving
our relationships with them• Demonstrate an ongoing understanding of
customer needs and how to best fulfill them• Reduce the likelihood of a break-up• Proactively identify customers who may be
on the path to a break-up
Our Approach
to hear from customers themselves
to learn from past customer behaviors
to anticipate future behavior
Ante: Both the ability and willingness to effectively communicate with your customers
Qualitative Analysis“The only stupid questionis the one not asked…”
• Customers are surprisingly willing to share their feelings (good or bad) and reasons for taking certain actions• Customer sentiment can be evaluated through:• Surveys / Focus Groups• Social Media• Leverage “Active” vs. “Passive” break-ups
Descriptive Analytics• Most companies have access to tools and large quantities of rich but under-leveraged data that can be mined for insights on customer engagement and attrition risk:
• Demographics/Firmographics• RFM / Transaction History Data• Coupon and/or Discount Usage (Type / Frequency)• Interaction / Promotion History and Response• “Big Data” – Online Browsing, Click-stream, etc.
• Techniques: Visual Analytics, Deciling, Cluster Analysis, etc.
Descriptive Analytics: Example• Differential Customer Profiling• Select key metric(s) – sales, profit, response, etc.• Decile (or cross-decile) customers by these metric(s)• Separately profile customers in top and bottom deciles• Identify dimensional differences between top and
bottom performers• Test marketing actions to incent desired behaviors• Re-decile and profile periodically to validate approach
• Potential Trap: Correlation vs. Causation
Predictive Analytics
• Though there are no guarantees, a solid understanding of the “what” and “why” of the past significantly enhances our ability to predict the future• Development of predictive models results in optimal actionability based on your analytical findings• Model-building tools and talent are readily available in the marketplace at very reasonable costs• Walk before you run: while large and complex models can be extremely powerful, even relatively straight-forward regression models can have a big impact
Predictive Models: Example• Tenured Attrition Model: Scores tenured but recently inactive customers by their likelihood of attrition (1=Most likely to be Retained, 10=Most likely to attrite)
• Key variables:• RFM (Recency, ∆ Frequency, ∆ Sales)• Categories purchased• Coupon Usage• Web site browsing behavior• Tenure
Predictive Models: ExampleModel Score
1
2
3
4
5
6
7
8
9
10
Attrition Risk
High
LowEmail: Lower
value offer
Lowest Risk: more likely to redeem offer, but less likely to drive incremental retention as they may buy again on their own
Highest Risk: less likely to redeem offer, “one foot out the door” already
Moderate Risk: best chance for incremental retention benefit
Email: Medium-High value offer
DM: Low-Medium
value offer
Attrition ModelScore
1
2
3
4
5
6
7
8
9
10
AttritionRisk
High
Low
1 2 3 4 5 6 7 8 9 10
High Sales Decile Low
IGNORE
EMAIL:Medium Value
Offer
EMAIL:Low Value
OfferDM:
High ValueOffer
TELESALESw/Offer
DM:Medium Value
Offer
EMAIL:Low Value Offer
EMAIL:High Value Offer
Predictive Models: Synergy• Often the creation and use of one model opens the door to other models that can work synergistically with each other to further drive insights and results
Attrition Model
RevenueModel
Lifetime ValueModel
Next Best ActionModel