Data-Driven Social Game Design:
A Deep-Dive into What The
Industry Is Doing
Dr. Volkan Ediz
Director of Product
GREE confidential
But first, who is GREE….?
GREE	
  Headquarters	
  
Tokyo,	
  Japan	
  
	
  
GREE	
  Interna7onal,	
  Inc.	
  	
  
San	
  Francisco,	
  CA	
  
	
  
GREE	
  Canada	
  	
  
Vancouver,	
  BC	
  
QUICK	
  FACTS	
  
6	
  Con7nents	
  playing	
  GREE	
  games	
  
1,882	
  Employees	
  Worldwide	
  
13	
  Games	
  made	
  in	
  North	
  America	
  
2004	
  
2011	
  
2013	
  
MILESTONES	
   GAME	
  STATS	
  -­‐	
  4	
  9tles	
  in	
  top	
  100	
  grossing	
  as	
  of	
  Nov.	
  2014*	
  	
  
Crime	
  City	
  (Studios)	
  
Reached	
  Top	
  10	
  Grossing	
  in	
  140	
  countries	
  
Top	
  100	
  Grossing	
  in	
  19	
  countries,	
  over	
  3	
  years	
  since	
  launch	
  	
  
	
  
	
  
*App	
  Annie	
  
GREE – A Global Gaming Powerhouse
Knights	
  &	
  Dragons	
  (Publishing)	
  
Reached	
  Top	
  10	
  Grossing	
  in	
  41	
  countries	
  
Top	
  100	
  Grossing	
  in	
  22	
  countries	
  
	
  
	
  
Where is Mobile Gaming Going?
•  Over 2.1b devices in
circulation WW today
•  The 2nd largest sector of the
gaming business
•  Represents 40% of App Store
downloads and 80%+ of
revenue	
  
•  Surpassing	
  $22	
  billion	
  by	
  ‘14,
over $35 billion business in
‘17	
  
•  Asia	
  Pacific	
  to	
  become	
  the	
  largest	
  
market	
  with	
  $12.2B	
  in	
  revenue	
  in	
  
‘14	
  
•  North	
  America	
  and	
  Europe	
  
represent	
  together	
  $8.1b	
  in	
  ’14	
  
WHERE WE ARE AND WHERE WE WANT TO BE!
Agenda
1)  Building Social Features
2)  Elements of Social Data
3)  Unique Challenges of Social Data
Why do We Need Social Features?
•  Growth Engine
•  Retention/Monetization
•  User Acquisition Cost
Savings
No other features can touch on all three at once.
How to Build Successful Social Features?
Deep Social Features
Campaigns
Reward
Structure
Integration
Social Integration
Social Connections
Invite / Referral
Codes
Social Footprint on
UI
Integrating Social
into Tutorial
Social Rewards and Campaigns
Social Rewards Campaigns
Basic Incentives
•  Rare Items
•  Stats Boots
•  Free Energy
•  Free Currency
Milestone Rewards
•  Rewards are given for hitting certain
milestones for Invites
•  Progressive milestones for each
referral
Daily Gifting
Limited-time
rewards
Rewarding multiple
postings/likes/
shares
Deep Integration Pays Off
Massive increase in
referred installs helps
to decrease cost of
user acquisition
Low Medium High Low Medium HighLevel of FB
Integration
Referred players are 2x-4x better retained and monetized
Agenda
1)  Building Social Features
2)  Elements of Social Data
3)  Unique Challenges of Social Data
Sources of Social Gaming Data
SOCIAL
DATA
In-Game Chat
Messaging Apps Guild Activity
Forums
Wiki-How
Ratings & ReviewsSocial Networks
Social Data Impacts Every Aspect of the Business
SOCIAL
DATA
Product
Development
Live Ops
Optimization
Targeted Sales /
Promos
Marketing
•  What do our customers want?
•  What features should we build
next?
•  How do we adjust the event
difficulty?
•  How do we increase user
engagement?
•  How do we boost customer
spending?
•  What is optimum volume,
frequency and duration of
sales/promos?
•  How do we target customers?
•  Which channels provide the
most engaged users?
•  How can we decrease cost of
user acquisition?
Social Data Requires Shift in Traditional Thinking:
Individual vs Collective Data
KPIs
SOCIAL
DATA
Game Strategy
Behaviors
Social
Activity
Marketing Evolution of
Groups
Solo Player
Guild of Players
Influencers
http://www.clashheads.com/
Social Data Requires Shift in Traditional Thinking:
Individual vs Collective Data
SOCIAL
DATA
Panda – Clash of Clans
$7K/month
Panda’s Guild
50 people - $100K/month
http://www.clashheads.com/
“I actually save
money playing these
games instead of
going out and
drinking,”
What would you pay for this influencer to try out your game?
1)  Building Social Features
2)  Elements of Social Data
3)  Unique Challenges of Social Data
Agenda
Split Testing In An Ideal World
What really happens in Social Games
SPLIT DATA
TEST
GROUP A
GROUP B
“Pseudo A/B groups” may provide directional
guidance.
1. Choose pseudo “control” and “test” groups where:
–  The install date of cohorts are as close as possible to the introduction of the A/
B test.
–  Both groups get exposed to same events.
–  Assume that the quality of installs are similar for both groups.
2. Compare metrics of “control” and “test” groups based on customer segmentation
GREE Confidential
“If there’s no A/B test conducted in the woods, can an analyst still ‘analyze’ the data?”
Methodology to Analyze Pseudo A/B Test
Day 2 Day 5 Day 6 Day7Day 1
Test
Introduced
Day 3
“Control” Period “Test” Period
Live Event Xn
Live Event Yn
Live Event Xn+1
Live Event Yn+1
Segmenting Customers Allow Us to See the Effect of
A/B Tests on User Behavior
Why segment customers?
•  A user who’s willing to spend early will be different from a user who buys later.
•  A user who buys large packages will be different from one who buys small packages.
•  We use the user’s first purchase time and first purchase amount to segment.
GREE Confidential
Hasty Minnow = I’m willing to try out IAPs and I
want instant gratification.
Hesitant Minnow = I will try out IAPs only after
careful thought.
Hasty Dolphin = I want some value for my IAPs
and I will spend quickly to progress.
Hesitant Dolphin = I consider value/cost trade offs
of IAPs and buy after getting a feel for the game.
Hasty Whale = I will invest quickly and heavily in a
game in order to speed up.
Hesitant Whale = I am prone to largesse, but I
have to feel that I’m getting the best.
Social Network Analysis Has Unique Challenges
Identify Key
Actors
Player Attribute
and Behavior
Social Sharing
and Filtering
Interaction
Analysis
Thank	
  You!	
  
	
  
Dr.	
  Volkan	
  Ediz	
  
Director	
  of	
  Product,	
  GREE	
  Interna7onal	
  
volkan.ediz@gree.net	
  

ediz_volkan_social_data

  • 1.
    Data-Driven Social GameDesign: A Deep-Dive into What The Industry Is Doing Dr. Volkan Ediz Director of Product GREE confidential
  • 2.
    But first, whois GREE….?
  • 3.
    GREE  Headquarters   Tokyo,  Japan     GREE  Interna7onal,  Inc.     San  Francisco,  CA     GREE  Canada     Vancouver,  BC   QUICK  FACTS   6  Con7nents  playing  GREE  games   1,882  Employees  Worldwide   13  Games  made  in  North  America   2004   2011   2013   MILESTONES   GAME  STATS  -­‐  4  9tles  in  top  100  grossing  as  of  Nov.  2014*     Crime  City  (Studios)   Reached  Top  10  Grossing  in  140  countries   Top  100  Grossing  in  19  countries,  over  3  years  since  launch         *App  Annie   GREE – A Global Gaming Powerhouse Knights  &  Dragons  (Publishing)   Reached  Top  10  Grossing  in  41  countries   Top  100  Grossing  in  22  countries      
  • 4.
    Where is MobileGaming Going? •  Over 2.1b devices in circulation WW today •  The 2nd largest sector of the gaming business •  Represents 40% of App Store downloads and 80%+ of revenue   •  Surpassing  $22  billion  by  ‘14, over $35 billion business in ‘17   •  Asia  Pacific  to  become  the  largest   market  with  $12.2B  in  revenue  in   ‘14   •  North  America  and  Europe   represent  together  $8.1b  in  ’14   WHERE WE ARE AND WHERE WE WANT TO BE!
  • 5.
    Agenda 1)  Building SocialFeatures 2)  Elements of Social Data 3)  Unique Challenges of Social Data
  • 6.
    Why do WeNeed Social Features? •  Growth Engine •  Retention/Monetization •  User Acquisition Cost Savings No other features can touch on all three at once.
  • 7.
    How to BuildSuccessful Social Features? Deep Social Features Campaigns Reward Structure Integration
  • 8.
    Social Integration Social Connections Invite/ Referral Codes Social Footprint on UI Integrating Social into Tutorial
  • 9.
    Social Rewards andCampaigns Social Rewards Campaigns Basic Incentives •  Rare Items •  Stats Boots •  Free Energy •  Free Currency Milestone Rewards •  Rewards are given for hitting certain milestones for Invites •  Progressive milestones for each referral Daily Gifting Limited-time rewards Rewarding multiple postings/likes/ shares
  • 10.
    Deep Integration PaysOff Massive increase in referred installs helps to decrease cost of user acquisition Low Medium High Low Medium HighLevel of FB Integration Referred players are 2x-4x better retained and monetized
  • 11.
    Agenda 1)  Building SocialFeatures 2)  Elements of Social Data 3)  Unique Challenges of Social Data
  • 12.
    Sources of SocialGaming Data SOCIAL DATA In-Game Chat Messaging Apps Guild Activity Forums Wiki-How Ratings & ReviewsSocial Networks
  • 13.
    Social Data ImpactsEvery Aspect of the Business SOCIAL DATA Product Development Live Ops Optimization Targeted Sales / Promos Marketing •  What do our customers want? •  What features should we build next? •  How do we adjust the event difficulty? •  How do we increase user engagement? •  How do we boost customer spending? •  What is optimum volume, frequency and duration of sales/promos? •  How do we target customers? •  Which channels provide the most engaged users? •  How can we decrease cost of user acquisition?
  • 14.
    Social Data RequiresShift in Traditional Thinking: Individual vs Collective Data KPIs SOCIAL DATA Game Strategy Behaviors Social Activity Marketing Evolution of Groups Solo Player Guild of Players Influencers http://www.clashheads.com/
  • 15.
    Social Data RequiresShift in Traditional Thinking: Individual vs Collective Data SOCIAL DATA Panda – Clash of Clans $7K/month Panda’s Guild 50 people - $100K/month http://www.clashheads.com/ “I actually save money playing these games instead of going out and drinking,” What would you pay for this influencer to try out your game?
  • 16.
    1)  Building SocialFeatures 2)  Elements of Social Data 3)  Unique Challenges of Social Data Agenda
  • 17.
    Split Testing InAn Ideal World
  • 18.
    What really happensin Social Games SPLIT DATA TEST GROUP A GROUP B
  • 19.
    “Pseudo A/B groups”may provide directional guidance. 1. Choose pseudo “control” and “test” groups where: –  The install date of cohorts are as close as possible to the introduction of the A/ B test. –  Both groups get exposed to same events. –  Assume that the quality of installs are similar for both groups. 2. Compare metrics of “control” and “test” groups based on customer segmentation GREE Confidential “If there’s no A/B test conducted in the woods, can an analyst still ‘analyze’ the data?” Methodology to Analyze Pseudo A/B Test Day 2 Day 5 Day 6 Day7Day 1 Test Introduced Day 3 “Control” Period “Test” Period Live Event Xn Live Event Yn Live Event Xn+1 Live Event Yn+1
  • 20.
    Segmenting Customers AllowUs to See the Effect of A/B Tests on User Behavior Why segment customers? •  A user who’s willing to spend early will be different from a user who buys later. •  A user who buys large packages will be different from one who buys small packages. •  We use the user’s first purchase time and first purchase amount to segment. GREE Confidential Hasty Minnow = I’m willing to try out IAPs and I want instant gratification. Hesitant Minnow = I will try out IAPs only after careful thought. Hasty Dolphin = I want some value for my IAPs and I will spend quickly to progress. Hesitant Dolphin = I consider value/cost trade offs of IAPs and buy after getting a feel for the game. Hasty Whale = I will invest quickly and heavily in a game in order to speed up. Hesitant Whale = I am prone to largesse, but I have to feel that I’m getting the best.
  • 21.
    Social Network AnalysisHas Unique Challenges Identify Key Actors Player Attribute and Behavior Social Sharing and Filtering Interaction Analysis
  • 22.
    Thank  You!     Dr.  Volkan  Ediz   Director  of  Product,  GREE  Interna7onal   volkan.ediz@gree.net