Are you acquiring customers that lose you money? eMetrics presentation on using customer segmentation, machine learning, random forests to predict returns and identify profitable & loss making customers.
11. Example Segments
Average Customer Sales
(Annual)
Average Customer Actual
Sales Less Returns Costs
(Annual)
Average Purchase Frequency
(Annual)
Keepers
Highest Value £908.3 £908.3 2
Super Loyal £389.9 £389.9 7
Bundler £205.1 £205.1 2
Potential Sensitive £147.4 £147.4 3
Thrifty £ 68.6 £ 68.6 2
Returners
Accidentals £345.0 £297.5 6
Low Risk Explorers £287.0 £200.7 5
High Risk Explorers £312.1 £171.5 5
Typical Overbuyers £320.9 £129.4 4
Serious Overbuyers £368.1 £ 98.2 4
Delinquents £405.1 £ 56.9 4
Negative Value £189.7 -£ 6.8 3
12. Accessories
Purchased 96 items
Gross Sales = £1,144
Womenswear
Purchased 104 items
Gross Sales = £1,845
Menswear
Purchased 28 items
Gross Sales = £349
Health & Beauty
Purchased 8 items
Gross Sales = £352
After refunds & ops costs
£23
Total Sales
£3,690
How Can It Be Used?
After cost to acquire & retain
- £7
-£45 after refunds & costs -£223 after refunds & costs
-£61 after refunds & costs £352 Net - it get’s kept
@brockvicky
13. Decision trees operate by splitting
data on conditions
Random Forests are collections of
Decision Trees utilising different
subsets of features
Comparing the outputs from each tree in the Forest, exposes
relative importance of each feature
Returns Drivers @brockvicky
14. ● As customers become loyal they
tend to return products more often
● Customers returning their orders
are less likely to churn
● Exception is return-sensitive
customers those who churn when
they return an item
Churn Rate Analysis
“We LOVE returns - they make us money.
Returns are a positive thing” ?
@brockvicky
15. ● Marginal net revenue
declines as customers
place more orders
● Costs suddenly become
very very important
@brockvicky
Churn Rate Analysis
17. Target for keeps
Suppress or nudge loss making behaviours
Service to reduce churn, not reduce margin
Policy and service optimisation
SANITY NOT VANITY
@brockvicky
18. THANK YOU
VICKY BROCK |DATA FOUNDER & GEEK
HELLO@VICKYBROCK.COM
VISIT MY BLOG: VICKYBROCK.COM/BLOG
@brockvicky