Data Driven Decisions @RummyCircle presented at Google Business Group (GBG) Mumbai on 21 March 2015 at Mithibai College

1. Data Driven
Business Decisions
By Sachin Uppal
Marketing Director,
Play Games24x7 (RummyCircle.com)

2. “You can’t manage what you
don’t measure.”
W. Edwards Deming and Peter Drucker

3. Attendees Today
47%
29%
14%
3%
7%
Occupation Break-Up of Attendees
Entrepreneur
Business Professional
Student
Working Professional
Other

4. How they heard about this event?
16%
15%
69%
How did you hear about this event?
Friend Referrals
Search + Social
Received an email from GBG Mumbai

5. Most common Questions
1) Where to find data?
2) How to use data for decision making?
3) How to make forecasts based on past data?
4) How to make predictions based on past and present
elements?

6. 1. Data driven businesses
i. Online business definition
ii. Business Model
iii. Data driven approach
iv. Key Principles and online business components
2. Where to find data and how to use it?
i. Why the data driven approach?
ii. Maturity in Data Analytics
iii. Business intelligence structure
iv. Optimization, Behavioral Learning, Big data & Predictive Analytics
3. Data driven decision making – Case Studies
i. Hourglass View of Business – Top, Bottom and Full Funnel
ii. Acquisition – Consumer Behavior and Preferences
iii. Product – Design, Pricing and Revenue optimization
iv. Retention – Big data driven predictive analytics
Agenda

7. A for profit business or non for profit organization that
sells its products and services on any digital delivery
platform (like: Website, Mobile, Tablet, Digital TV etc.)
where product / service content is delivered through
internet.
Case in point, Play Games24x7’s Online Rummy
game business, RummyCircle.com, where the service is
delivered online.
P.S.: I have limited the definition to a fairly tactical one to keep things in perspective. I
am not covering, corporate social responsibility, impact on environment, impact on
lives of people and world peace at this moment.
1.1 Online Business Definition

8. 1.1 Case in point – RummyCircle.com
RummyCircle: India’s
best and largest online
rummy site
3 offices, 160
employees
Funded by Tiger GlobalCash positive and
growing!
We are
hiring!

9. 1.1 Growth depends on How you work with Data

10. Maximize the Lifetime Value of a Customer.
1.2 Business Model
Minimize CAC
Maximize LTVBusiness Need
Minimize
Churn
/ Engage

11. Maximize the Lifetime Value of a Customer.
1.2 Business Model – An Example
 Calculate Customer Life Time
 Calculate Lifetime Value
Monthly Acquisitions Monthly Retention
Month Customers M0 M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11
M0 100 100% 60% 40% 25% 18% 10% 0%
M1 100 100% 60% 40% 25% 18% 10% 0%
M2 100 100% 60% 40% 25% 18% 10% 0%
M3 100 100% 60% 40% 25% 18% 10% 0%
M4 100 100% 60% 40% 25% 18% 10% 0%
M5 100 100% 60% 40% 25% 18% 10% 0%
Monthly Acquisitions Monthly Retention
Month Customers M0 M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11
M0 100 R0 R1 R2 R3 R4 R5 R6
M1 100 R0 R1 R2 R3 R4 R5 R6
M2 100 R0 R1 R2 R3 R4 R5 R6
M3 100 R0 R1 R2 R3 R4 R5 R6
M4 100 R0 R1 R2 R3 R4 R5 R6
M5 100 R0 R1 R2 R3 R4 R5 R6

12. 1.2 Business Model
 Calculating Customer Life Time
Monthly Acquisitions Monthly Retention
Month Customers M0 M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11
M0 100 R0 R1 R2 R3 R4 R5 R6
M1 100 R0 R1 R2 R3 R4 R5 R6
M2 100 R0 R1 R2 R3 R4 R5 R6
M3 100 R0 R1 R2 R3 R4 R5 R6
M4 100 R0 R1 R2 R3 R4 R5 R6
M5 100 R0 R1 R2 R3 R4 R5 R6
• Suppose for a given month M: R0, R1, R2,…, R6 are the 0-, 1-, 2-,…, 6-month Retention Rates.
• So 100 customers of M0 generate revenues as follows:
 100*R0 contributed in M0
 100*R0*R1 contributed revenue in month M+1
 100R0*R1*R2 contributed revenue in month M+2 and so on.
• So we can say:
 100*(1-R0) customers never generated revenue
 100*R0*(1-R1) generated revenue for 1 month
 100* R0 * R1(1-R2) generated revenue for 2 months and so on.
Player Life Time (PLT) = R0(1-R1) x 1 + R0R1(1-R2) x 2 + R0R1R2(1-R3) x 3 + …

13. 1.2 Business Model
 Player Life Time Value
• Now we look at the revenue generating customers and calculate Average Revenue Per Paying
Customer (ARPPC)
𝐴𝑅𝑃𝑃𝐶 =
𝑅𝑒𝑣𝑒𝑛𝑢𝑒 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑 𝑏𝑦 𝑝𝑙𝑎𝑦𝑒𝑟𝑠 𝑎𝑐𝑢𝑖𝑟𝑒𝑑 𝑖𝑛 𝑚𝑜𝑛𝑡ℎ 𝑀
𝑆𝑢𝑚 𝑜𝑓 𝑚𝑜𝑛𝑡ℎ𝑙𝑦 𝑐𝑜𝑢𝑛𝑡 𝑜𝑓 𝑝𝑙𝑎𝑦𝑒𝑟𝑠 𝑤ℎ𝑜 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑 𝑡ℎ𝑒 𝑎𝑏𝑜𝑣𝑒 𝑟𝑒𝑣𝑒𝑛𝑢𝑒
Life Time Value (LTV) = PLT * ARPPC
Monthly
Acquisitions Avergae Revenue Per Paying Customer
Month Customers M0 M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11
M0 100 Arp0 Arp1 Arp2 Arp3 Arp4 Arp5 Arp6
M1 100 Arp0 Arp1 Arp2 Arp3 Arp4 Arp5 Arp6
M2 100 Arp0 Arp1 Arp2 Arp3 Arp4 Arp5 Arp6
M3 100 Arp0 Arp1 Arp2 Arp3 Arp4 Arp5 Arp6
M4 100 Arp0 Arp1 Arp2 Arp3 Arp4 Arp5 Arp6
M5 100 Arp0 Arp1 Arp2 Arp3 Arp4 Arp5 Arp6

14. 1.3 Data Driven Approach

15. 1.3 “How” you execute - matters!
Source: Dilbert.com

16. 1.3 Data Driven Approach
Business Objective: Based on our strategy, we set a business
objective. This is also based on past data of what works and
what doesn’t.
Analysis: We break down the business problem by looking at
the relevant data.
Hypothesis: Basis our analysis, we come up with Hypothesis
that could explain the behavior.
Experiment: We setup experiments to test our hypothesis.
Define KPIs. Avoid overload of data.
Outcome: We measure the results and report our outcomes.
Insight: We learn from the outcomes.
This process is repeated and we continuously improve our
business decisions and outcomes.

17. 1. Measure / Track Everything
2. Experiment Boldly
3. Test Religiously
4. Learn Iteratively
5. Use data to your advantage!
1.4 Key Principles

18. Key business components from growth perspective:
New User Acquisition - New players through attractive
campaigns
Product Design - Provide a wholesome Rummy experience
Payments – Seamless multi-option payment mechanism
Retention - Innovative offers and promotions
Focus on Customer service components:
Customer Delight – Low Turn-Around-Times
Smooth Operations – Time bound delivery / on -schedule
Risk Management – Safe, fair, low risk - enhance consumer
trust
1.4 Key Online Business Components

19. 1) Identify sources of traffic accurately.
2) Model Word of Mouth traffic based on paid traffic
sources or referral traffic.
3) Build detailed funnels for each source and each
path.
4) New Acquisitions = (WOM + Referrals + Paid) traffic.
So, your CPA is not based on Paid Conv from Paid
Traffic. But (total cost) / (total paid conversions).
5) If you do not factor in future retention, you are
building a half model.
6) Future retention has a cost too, much lower cost
than first acquisition.
7) Nothing can compensate for a poor product
experience.
1.4 Key Principles & Business Components

20. 1. Data driven online businesses
i. Online business definition
ii. Business Model
iii. Data driven approach
iv. Key Principles and online business components
2. Where to find data and how to use it?
i. Why the data driven approach?
ii. Maturity in Data Analytics
iii. Business intelligence structure
iv. Optimization, Behavioral Learning, Big data & Predictive Analytics
3. Data driven – Case Studies
i. Hourglass View of Business – Top, Bottom and Full Funnel
ii. Acquisition – Consumer Behavior and Preferences
iii. Product – Design, Pricing and Revenue optimization
iv. Retention – Big data driven predictive analytics
v. Risk Management - Process Automation
Agenda

21. Business
Intelligence
Engineering
Acquisition
Risk
Management
Operations
Customer
Support
Retention
Product
2.1 Where to find data?
Data
exchange
Facebook
Ad 1 or Ad
2? Green or
Red colour?
Limit Rs.
1000 or Rs.
5000?
Tournament
at 9 PM or
10 PM?
I responded
within 2
hours!
How much
bonus to
whom? Rs.
1000 or
Rs.200?
2 Steps or 3
Steps? 4
options or
5?
Add 2 or 3
servers?
Tech. 1 or
2?
Data is available in every single department of your company.

22. Business
Intelligence
Engineering
Acquisition
Risk
Management
Operations
Customer
Support
Retention
Product
2. Culture of Analytics – Across Business
Data
exchange
• Data helps in
making Fact /
Evidence based
decisions. Faster
and more
accurately.
• It’s no more an
activity or
department, it is
the Culture.

23. 2.1 Why Data driven approach?

24. 2.1 Maturity in Data Analytics
Source: SAS Institute

25. 2.2 Tools and Techniques
Source: Dilbert.com

26. 2.2 Tools and Techniques
Below are some standard tools and techniques we employ on an
ongoing basis:
• Behavior or Preference measurement - A/B Testing
• Look for Statistical Significance – for complete funnel
• Monte-Carlo Simulation (smaller data samples)
• E.g.: Acquisition and Product tests.
• Regression Analysis
• Predicting Y with variables X1, X2, X3… Xn
• E.g.: Display advertising has an impact on brand search
volume.
• Big Data
• Predictive models based on real-time events and / or past data for real
time actions or predictive outcomes. E.g.: Retention case study.
Image Source

27. 2.3 Business Intelligence Structure
BI enables the business to make intelligent, fact-based decisions

28. 2.3 BI Structure Contd.

29. 1. Data driven online businesses
i. Online business definition
ii. Business Model
iii. Data driven approach
iv. Key Principles and online business components
2. Where to find data and how to use it?
i. Why the data driven approach?
ii. Maturity in Data Analytics
iii. Business intelligence structure
iv. Optimization, Behavioral Learning, Big data & Predictive Analytics
3. Data driven – Case Studies
i. Hourglass View of Business – Top, Bottom and Full Funnel
ii. Acquisition – Consumer Behavior and Preferences
iii. Product – Design, Pricing and Revenue optimization
iv. Retention – Big data driven predictive analytics
v. Risk Management - Process Automation
Agenda

30. 3. Hourglass View

31. 3 Top Funnel View
We apply the lens of Data and
Analytics to expand the conversion
Funnel to uncover the entire path to
conversion and understand each
metric and optimize it to grow the
funnel!
Paid Conversions
Impressions
• CTR
• Segment
Clicks
• CTV
• LPO
Visits
• Trials
• Logins
Leads
• Game
Play
• Offer
Paid
• Loyalty
Points
• Bonuses
Re-
Purchase

32. 3.1 Full Funnel View - Optimization
• eCPM
• Reach
Impressions
• CTR
• Segment
Clicks
• CTV
• LPO
Visits
• Trials
• Logins
Leads
• Game
Play
• Offer
Paid
• Loyalty
Points
• Bonuses
Re-Purchase
Impressions
• Reduce
eCPM
• Expand
Reach
Clicks
• Increase
Click Thru
Rate
• Reduce Cost
/ Click
Visits
• Increase
Click to Visit
Rate
• Landing
Page
Optimization
Leads
• Game Trial
Rate
• Lead to Paid
Rate
Paid
• CPA
• Offers /
Bonus
• Referrals
Loyalty
• Bonuses
• Tournaments
• Loyalty
Points

33. 3.1 Full Funnel View - Optimization
Questions to Answer
1. Where should we buy impressions?
2. How much should we bid for it?
3. Should we pay for impressions, clicks or conversions?
4. Which audience shall we target?
5. What message should we show?
6. How frequently should we show our ads?
7. When should we retire our ads?
• eCPM
• Reach
Impressions
• CTR
• Segment
Clicks
• CTV
• LPO
Visits
• Trials
• Logins
Leads
• Game
Play
• Offer
Paid
• Loyalty
Points
• Bonuses
Re-Purchase

34. 3.1 Full Funnel View - Optimization
Questions to Answer
1. How to increase our game trials?
2. How to use bonuses to drive action?
3. How to improve visit rate?
4. What communication should we do at different stages of player cycle?
5. What interaction points do we presently have on website?
6. How to increase efficiency of each point to increase conversions?
7. When to prompt users to send referrals?
• eCPM
• Reach
Impressions
• CTR
• Segment
Clicks
• CTV
• LPO
Visits
• Trials
• Logins
Leads
• Game
Play
• Offer
Paid
• Loyalty
Points
• Bonuses
Re-Purchase

35. 3.1 Bottom Funnel View - Optimization
New Paying Customers
WOM +
Referrals
Existing Paying
Customers
Optimizing the Post Conversion Journey is
equally important. If your customers are not
driving new sales for you, your virality
factor is Zero. And your cost of new user
acquisition will grow so much that it would
be practically impossible to break-even or
make profits.
Questions to Answer:
1. How to get players to send more
referrals?
2. When to ask players to send referrals?
3. Should we incentivize?
4. How much should we incentivize?
5. Two Step Vs Three Step process
6. Twitter and FB Sharing?

36. 1. Acquisition – Consumer Behavior and Preferences
2. Product – Design, Pricing and Revenue optimization
3. Retention – Big data driven predictive analytics
3.2 Case Studies

37. 1. Acquisition – Consumer Behavior and Preferences
2. Product – Design, Pricing and Revenue optimization
3. Retention – Big data driven predictive analytics
3.2 Case Studies

38. Challenge: Cost per acquisition (CPA) grows
significantly when we scale-up the new user acquisitions.
Objective: Scale-up new users acquisitions while
keeping CPAs low.
Approach: Full funnel optimization through multiple
experiments.
3.2.1 Acquisitions – A scale-up challenge

39. Approach: Full funnel optimization (FFO) through
multiple experiments.
Optimize Media buys – Reduce Cost
Optimize Creatives – Improve Conversions – Reduce
Cost / Lead
Optimize Landing Pages – Improve Conversions –
Reduce CPA
3.2.1 Acquisitions – A scale-up challenge

40. Business Objective: Bring down CPA
Hypothesis: Aggregated demographic
targeting increases CPA.
Experiment: Micro-segmentation of demographic to
identify the high converting audience
• Data helped us identify the highly converting
demographic
• We made micro-segments of that audience
• Highly focused media buys to scale-up campaign
Outcome: CPA down by 10%
Insight: Micro-segmentation leads to more structured
spends and lowers CPA.
3.2.1 Full Funnel Optimization – Media Buys

41. Simple Moderately Complex
3.2.1 FFO – Experiment
Which one won?

42. Simple Moderately Complex
3.2.1 FFO – Experiment
Moderately complex creative showed an 18% Improvement in Visit to
Paid rate.

43. Business Objective: Reduce CPA by improving
conversions.
Hypothesis: Absolute simplicity is better in interactive
banners.
Experiment: We made two versions of an ad to test if
Absolutely Simplicity in Interactive Banner works better
than Moderate Complexity.
Outcome: While Absolute Simplicity drives higher clicks,
it doesn’t drive higher conversions as audience need a
problem that they think is challenging but quickly
solvable.
Insight: Players need a challenge that is solvable,
quickly.
3.2.1 FFO – Experiment

44. 3.2.1 FFO - Landing Pages

45. Business Objective: Increase Click to Visit Rate and
Visit to Lead Rates
Hypothesis: Registration form has multiple elements
that need to be optimized.
Experiments: Alter the registration form components
like:
Number of fields
Left or Right
Register for Free
Bonus message
Outcome: CTV up by 25%. LTP up by 20%.
3.2.1 FFO - Landing Pages

46. i. Acquisition – Consumer Behavior and
Preferences
ii. Product – Design, Pricing and Revenue
optimization
iii. Retention – Big data driven predictive
analytics
3.2 Case Studies

47. 3.2 Add Cash Process

48. 3.2 Choose Amount Screen

49. 3.2 Product – Shopping cart challenge
Challenge: Flat-lining conversion funnel percentages
Approach: Optimize Shopping cart (or Add Cash funnel)
funnel for conversions through multiple experiments:
• Build plug-n-play framework for running A/B tests
• First experiment - Reduce the number of choices

50. 3.2 Funnel Optimization - Reduced Choices
Objective: Increase conversion rate.
Hypotheses: More choice in selecting amounts results
in lower conversion.
Experiment: Split users entering the funnel into two
groups randomly
• Control group - 5 amount tiles
• Test group - 4 amount tiles

51. 3.2 The Experiment – Control Group

52. 3.2 The Experiment – Test Group

53. Outcome:
Increase in conversions by 12% in the reduced options
path
Increase in average deposit amount by 15%
Increase in ARPU by 17%
Insights:
Less is more – helps players make a quick decision
Behaviour – Propensity to deposit higher amounts
reduces when lower amounts are available
3.2 Experiment Results

54. i. Acquisition – Consumer Behavior and
Preferences
ii. Product – Design, Pricing and Revenue
optimization
iii. Retention – Big data driven predictive
analytics
3.3 Three Real Life Situations

55. 3.3 Retention - High Early Churn Rates
Challenge: A significant proportion of players do not last
beyond first day on the site.
Objective: To improve early retention of new cash
players
Approach: Offer a strong incentive to players sitting on
the fence to return to play.

56. 3.3 Context – Churn Trends
Problem: How do you substantially improve the Day1
retention while ensuring a positive net return?
Points to consider:
• The incentive does not affect players who are likely to
return anyway.
• The churn patterns beyond the first day are not
expected to change.
How do you make an attractive yet profitable offer?

57. Identify which players are likely to return and which
players are potentially one-timers.
Using this model, we performed an a priori segmentation
of players into Fence-Sitters and Enthusiasts.
3.3 The Proposed Solution – Segment Identification

58. 3.3 The Experiment - Control Group (Fence-Sitters)

59. 3.3 The Experiment - Test Group (Fence-Sitters)

60. Insight: Customization of incentives and journeys for
different segments can significantly increase the impact
of any promotion.
3.3 Results
Category Day1 Retention
Fence-Sitters: Test 59%
Fence-Sitters: Control 54%
Enthusiasts 67%
Over a 2-month period, this campaign yielded an ROI
of 62%.

61. 1. Measure Everything
2. Experiment Boldly
3. Test Religiously
4. Learn Iteratively
5. Use data to your advantage!
Summary

62. Connect on Twitter: @sachinuppal
LinkedIn: in.linkedin.com/in/sachinu/
Email me: Sachinuppal AT Gmail.com
Questions