2. about
Relationship programs
A market leading Foods & Beverages Business [India]: Customer
Loyalty Program across 350 outlets – 320,000 members and 12,000,000
transactions mined; Retailer Loyalty Program across a network of
3,000 outlets
A market leading Beauty / Fitness Chain [India]: Customer Loyalty
Program across 110 centers in India
Indian arm of a market leading global brand of Beauty products: High
Value Customer panel set up and mass customization initiatives
across the brand’s shop-in-shops in India
Indian arm of a global Communications Service Provider: Business
analytics for Revenue management and Customer Loyalty Program for
the CSP’s India operations
A 1,000 store global jewelry chain [UAE]: Business analytics and
Customer Loyalty Program across 180 outlets in UAE, Europe and
India – a consulting assignment
3. about
Predictive modeling/analytics
A US based satellite co: Pricing and discount modeling solution based
on 5-year historical data and 10-year look-ahead (prospective) data
A US based consumer marketing co: Predictive analytics solution based
on historical marketing program data for the last 5 years
An Ecommerce [B2B2C] brand merchandising business [India]: Web and
Business Analytics – a full business solution for Online businesses
A market leading Packaged Foods business [India]:Models predicting
Market Share of leading Indian packaged foods brands based on Retail
Audit and Panel data
HR Analytics for one of the largest employers in India: People
performance metrics modeling, Attrition modeling and Salary
Intelligence
HR Analytics for India’s largest Assessments company: Talent Pool
Supply-Demand Modeling, Capability-Effectiveness models
4. about
Consumer research
A F&B Industry focused Private Equity fund: Consumer / Brand
Perception Studies to evaluate two leading Fine Dining chains for
potential investments
A Global Education Major: Sizing up the BPO markets in Pakistan and Sri
Lanka for the group to design entry strategies
One of India’s largest fine dining restaurant chains: Consumer
Perception, Feedback analysis and Mystery Customer Exercises
An Indian Fortune 500 Petro-major: Retail Chain Set-up – 56 feasibility
studies to date, Auto-LPG sales potential studies, Retail Outlet facility due
diligence
India arm of a global luxury brand: Mystery Shopper Exercise across 7
cities and 9 stores – first ever store evaluation in India
An Indian Fortune 500 Petro-major : ‘Oil Conservation Fortnight’
Effectiveness Studies, Retail Audits, ‘Non-Fuel Options’ Study at Retail
Outlets
6. Customer segmentation/profiling
The rare breed – rarely eat out (less times than once a month)
A negligible percentage of the sample (3.2%)
Equal no of respondents split between impulsive and planned decision making
Distance traveled anywhere between next door to > 10km
Most prefer Indian Cuisine - buffet/steak & grill and prefer fine dining to casual
Spend: Rs. 70 to Rs. 3,000 a meal on an outing
Most are under 25 to 30 years and are from the salaried class
The gregarious – eat out once a month – mostly with friends
Decision equally split between impulsive and planned, most travel between 2 to 10 km
Like Indian cuisine best, followed by both western and eastern cuisines
Dislike fast food but have an equally good inclination to have all other food types
Do not distinguish much between casual/fine dining styles
Young – predominantly under 25, most are salaried
Spend:Rs.1,000 to Rs.2,500
The moderates – eat out once a fortnight
Most have a plan to dine out but all might not have decided on the restaurant to visit
Travel between 2 to 10 km
Prefer Indian but have an equally good inclination to go for the other cusines, like the
buffet and the BBQ
Prefer the casual to the fine dining style
Spend: Rs. 1000 to Rs. 2500
Are from the business/salaried class
7. Customer segmentation/profiling
Email
SMS
Mobile
Home Number
Office Number
Number of Respondents
Others
Weekday Weekend
Time of week
9. Attrition Modeling
Interval between visits 11 days 22 days 36 days 52 days 66 days 82 days > 100 days
October 630 61
November 1252 423 29
December 1874 1000 189 29
January 1493 1608 462 146 26
February 862 1639 658 248 99 19 4
March 958 1878 842 383 168 73 25
April 914 1870 779 376 202 88 76
Base: 21,383 High frequency visitors made more visits before
attrition
Member visits numbered between 3 and 18
Assumptions:
1. Enrollments up to April considered
2. Members not transacting for the second time after being enrolled, until July ’04
10. Attrition Modeling
•Worrying Non-usage activity
linear 5000 180.00
trend, a 4229
4500 4302 160.00
steady churn 163.78
4000 3735
of 3,500 to 3525 140.00
4,500 per 3500 3223 138.77
120.00
month could 3000 119.87
Members
100.00
be
Days
2500
anticipated 90.36 80.00
2000 1751
•Rising ave 60.00
1500 65.08
lifetime 40.00
1000 691
suggests 44.36
early 500 26.64
20.00
members 0 0.00
losing October November December January February March April
interest
No of churns Ave Lifetime (Days)
11. Redemption Ready Members – 300 to 700
reward points
60000
End of gestation for a majority of members
50000
40000
30000 Equation predicts a rapid decline
20000 Surge in redemptions at this level
10000
0
Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul
'03 '03 '03 '04 '04 '04 '04 '04 '04 '04 '04 '04 '04 '04 '04 '05 '05 '05 '05 '05 '05 '05
-10000
12. The next 2 months across milestones
Member Eligibility: Predictions for the next two months
160,000 Flatter, steady growth
140,000 134,687 140,313
129,479
124,682
Number of eligible members
120,000 Rapid decline at low end
100,000 Faster growth enables higher level sweet spot
80,000 81,494
70,436
62,533
60,000
57,144
50,934 46,662
40,000 39,355
28,222
26,547
21,540 25,992
20,000 22,447
11,809 17,876
10,783 3,104 14,889
628 830
0 340
Nov '03 Dec '03 Jun '05 Jul '05 Aug '05 Sep '05
All eligible 300 to 700 700 to 2,000 2,000 to 5,000
13. Advanced Analytics for Decision Support
Marketing Effectiveness modeling for a 8 billion USD company
14. Marketing Program Performance
Incremental Sales, Volume and Margins
Incremental Sales from the National Sales program (2002 through 2006) Each of the programs
comprising the NSP has
1,300,000,000 contributed 1% or less to
annual sales
1,200,000,000
All but two programs – Fall
1,100,000,000 Rebate 2002 and Spring
Rebate 2006 – have
1,000,000,000
contributed significantly to
sales but individual
900,000,000
programs are yet to yield
800,000,000 consistent y-o-y increments
2002 (1.42%*) 2003 (2.03%) 2004 (1.59%) 2005 (1.6%) 2006 (0%)
Fall Warranty $3,795,387 Annually, off-season
Fall Rebate $0 $9,313,176 $6,001,539 $11,947,742
programs add between $12
Spring Warranty $9,991,971
Spring Rebate $8,494,085 $10,735,324 $8,473,328 $0
million and $20 million to
Annual $866,918,809 $950,862,630 $1,051,523,573 $1,274,812,840 $1,414,647,693 the top line
15. Marketing Program Performance
Incremental Sales, Volume and Margins
Incremental Sales from the National sales program (2002 through 2006)
$160,000,000
$140,000,000
$120,000,000
$100,000,000
Sales ($)
$80,000,000
$60,000,000
$40,000,000
$20,000,000
$0
Incr. Sales Program Sales Incr. Sales Program Sales Incr. Sales Program Sales Incr. Sales Program Sales
Spring Rebate Spring Warranty Fall Rebate Fall Warranty
2006 $0 30,713,129
2005 $8,473,328 $80,833,729 $11,947,742 $24,596,407
2004 $10,735,324 $24,483,535 $6,001,539 $42,806,992
2003 $9,991,971 $30,576,490 $9,313,176 $37,921,499
2002 $8,494,085 $39,547,183 $0 $4,853,516 $3,795,387 19,499,789
Incremental sales due to the programs have grown over the years but for a dip during Fall 2004
As a proportion of sales tracked through respective programs, incremental sales have varied between 10% and 49%
Except for the first Fall program (2002) and the 2006 Spring Warranty program, all others have generated significant incremental sales
16. Marketing Program Performance
Incremental Sales, Volume and Margins
What volumes do programs drive?
2006 0 [Spring Warranty] Spring Warranty in
2003 generated the
most volumes over the
2005 8,829 10,450
Spring Rebate
4 years
Spring Warranty
Fall Rebate Again, Fall 2004 sees a
2004 9,890 5,329 Fall Warranty
dip in volumes but the
Fall season next year
2003 12,461 9,320
does very well
0 [Fall Rebate] Except in 2005, Spring
2002 9,685 4,993 programs have
performed better than
0 5,000 10,000 15,000 20,000 25,000
Volume (number of units) the ones during Fall
17. Marketing Program Performance
Programs that generate maximum Rate of Returns
The 2 warranty programs
Margins and Returns from the Programs
have yielded good returns
50.00% 5,000,000 though, the second among
40.00% 35.52% 36.95%
4,000,000 the two works well on both
Rate of Return (Net Margin / Incremental
30.00% 26.22% 31.91%
3,000,000
margins and returns
18.58%
20.00% 14.70% The 2005 Spring Rebate
12.31% 2,000,000
10.00% program yields
Net Margin ($)
1,000,000
0.00% considerably high negative
Sales)
Spring Fall Spring Fall Rebate Spring Fall Rebate Spring Fall Rebate 0
-10.00% Rebate Warranty Warranty 2003 Rebate 2004 Rebate 2005
net margins and returns
-20.00%
2002 2002 2003 2004 2005 -1,000,000 despite generating decent
-2,000,000 incremental sales and
-30.00%
-3,000,000 margins
-40.00% -53.15%
-50.00% -4,000,000
The 2003 Spring Warranty
program has, so
-60.00% -5,000,000
far, fetched the best
Rate of Returns NetMargin
returns
18. Impact if program was not offered
Change in mix of products sold
Change in the mix of products sold if incentives were withdrawn
6.00%
% drop in the number of units sold
5.00%
4.00%
There is a likelihood of
3.00%
about 4% to 5% lesser
number of Premium
2.00%
products being sold if the
1.00% programs were
unavailable to
0.00% consumers. The drop is
2002 2003 2004 2005
Drop in Premium Products sold 2.78% 4.69% 4.99% 3.76%
not as much for non-
Drop in Non-Premium Products sold 1.42% 1.91% 0.88% 1.54% Premium products
20. PPU business concerns
Usage Variance – Corporate
Corporate Customers by Weekly
Customer Code
5000
3782
6654
• Fluctuations are wild 8041
8084
across the 4000
9231
board, irrespective of
quantum of usage
Usage in minutes
• Dips are prolonged in 3000
many cases; few large
peaks
2000
• Trend observed in
some cases while in
others, growth is flat 1000
over the year
0
1 3 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52
weeks
22. Scenarios
Forecasting Usage for an Enterprise Customer
350,000
300,000
250,000
200,000
Usage in
Most Likely
minutes 150,000
Optimistic
Pessimistic
100,000
50,000
0
1st Month 2nd Month
Next months
23. Statistics
Correlations
Correlations
Meeting Minutes Port
Enterprise customers Meeting Pearson Correlation 1 .969** .990**
Sig. (2-tailed) .000 .000
• No of attendees (Ports in use) N 352 352 352
have a very high correlation with Minutes Pearson Correlation .969** 1 .957**
Meetings (in the case illustrated Sig. (2-tailed) .000 .000
alongside) for Enterprise N 352 352 352
Port Pearson Correlation .990** .957** 1
customers. And in many cases,
Sig. (2-tailed) .000 .000
they very strongly drive minutes N 352 352 352
as well **. Correlation is significant at the 0.01 level (2-tailed).
• Meetings drive Minutes strongly Correlations
in case of Corporate customers Meeting Minutes Port
Meeting Pearson Correlation 1 .808** .783**
• In essence, teams may drive Sig. (2-tailed) .000 .000
meetings in the enterprise N 196 196 196
segment while initiators drive Minutes Pearson Correlation .808** 1 .753**
meetings – and hence minutes – Sig. (2-tailed) .000 .000
in case of corporate customers N 196 196 196
Port Pearson Correlation .783** .753** 1
Sig. (2-tailed) .000 .000
Corporate customers
N 196 196 196
**. Correlation is significant at the 0.01 level (2-tailed).
24. XYZ Loyalty Program
Enterprise Incentive Strategy
Group Bonus: 5
additional XYZ
marginal scope for improvement
Loyalty Minutes for a
High
group meeting
4262
incentivise group meetings
Activity Reward: 10% extra
Port Utilization
in Loyalty Minutes if > x ports
are active in a month (could
be relaxed in specific cases)
2226
9140 incentivise by number of ports used in a month
3372
4323
Low
Low Number of Ports in Use High
27. Market Basket Analysis
Suggestive Selling Tools for an Online Business
Business Challenge
arrive at purchase likelihood
estimates across all products at unit
and category levels
visualization of the model along with
prioritization of product pairs and
triplets that sold well together
28. Market baskets – Category
Confidence Stats for associations
Supposed direction
of cause-effect
Stronger lines show higher
degree of association
Computers and Accessories Stationery
0.11
0.69
Watches & Clocks 0.37
0.06
0.12 0.58
Apparel and accessories Bags
Electronics 0.29
0.48
0.19 0.22
Utilities
Travel Bag, T-Shirt & Travel organizer
29. Market baskets – Products
Confidence Stats for associations
Photo Pen Holder Photo Frame
0.27 0.73
I Don't Sleep
Round Neck T-Shirt - (Men) I Didn't Get Smarter
0.324 0.622 Round neck T-Shirt - (M)
Arrow Shirt White
0.35 0.567
Black Polo Neck
Van Heusen - Blue
T-Shirt - (Men)
0. 35
Black I Didn't Get Smarter 0.54
Round neck T-Shirt - (M) Leather wallet
with white stitch
0.35 0.486
Leather Pen Stand 0.432
White Polo Neck T-Shirt - (Men)
Magic Calculator with Pen
31. Client-wise Profiling
Distribution of clients and attrites across the attrition range
450 70.00%
400
60.00%
350
# of clients
50.00%
attrites as a % of all associates
% contribution to total attrition
300
250 40.00%
% contribution to
# of clients
total attrition
200 30.00%
150
58 clients with modest attrition rates… 20.00%
100
10.00%
50
0 0.00%
0%
0.1% to 4.99%
5% to 14.99%
15% to 24.99%
25% to 34.99%
35% to 44.99%
45% to 54.99%
55% to 64.99%
65% to 74.99%
75% to 84.99%
85% to 94.99%
95% to 100%
% Attrition (Range)
32. Client-wise Profiling
Distribution of associates & attrites across the attrition range
1,800 …contribute the highest to overall attrition 70.00%
by sourcing in and losing massive numbers…
1,600 Mean # of 60.00%
associates
1,400
50.00%
attrites as a % of all associates
% contribution to total attrition
1,200 % contribution
Mean # of associates
to total attrition 40.00%
1,000
800 …while 19 clients sourcing the second highest 30.00%
average number of associates have relatively
600 higher attrition rates but contribute less than
20.00%
10% to overall attrition numbers
400
10.00%
200
0 0.00%
0%
0.1% to 4.99%
5% to 14.99%
15% to 24.99%
25% to 34.99%
35% to 44.99%
45% to 54.99%
55% to 64.99%
65% to 74.99%
75% to 84.99%
85% to 94.99%
95% to 100%
% Attrition (Range)
33. Client-wise Profiling
Clients with the worst attrition rates: 100% to 50%
100% attrition on a sizeable
base of associates
3,000 110.00%
100.00%
2,500
90.00%
2,000
Attrition rates climb down with larger offtakes
80.00%
1,500
70.00%
1,000 R2 = 0.9603
60.00%
500
50.00%
0 40.00%
Total # of associates Percentage_Attrited Poly. (Percentage_Attrited)
34. Designation-wise Profiling
Designations with the worst attrition rates
6000 55.00%
Attrition drops progressively, with relatively more popular designations…
50.00%
5000
45.00%
4000
40.00%
3000 R2 = 0.9906 35.00%
30.00%
2000
25.00%
1000
20.00%
0 15.00%
Total # of associates Percentage_Attrited Poly. (Percentage_Attrited)