LTV measurement is one of most important aspects of growth in mobile gaming. How to start working with it? How to start with basic scripts and then move to more advanced strategies, including advanced models
User Acquisition focused on LTV on steroidsGameCamp
How to use statistics and ML for a better user acquisition (UA) with Feature Selection (FS) methodology. Is FS for me? How does it work, how to prepare and how can I benefit from it? Let's take look on examples of gaming companies that used Feature Selection (FS) for improve and scale their business.
LTV from Ads. Challenge and solution: Huuuge Games and Tap Tap Games exampleGameCamp
How to work with LTV measurement from ads? How to measure effectiveness of campaigns for games monetised via ads?Presentation based on Huuuge Games example and the system they built.
How to create good BI for UA activities? How we did it at Pixel Federation? How we structure our systems and our teams? What we can say based on our experience? Presentation run by Michal Grno at 8th edition of GameCamp (www.gamecamp.io).
This document discusses effective liveops strategies for games. It defines liveops as changes made to games after launch, generally without code changes, such as new items, events, or offers. Key components of liveops discussed include business intelligence, events, and offers/promotions. Good business intelligence is tied to player behavior and generates insights over time. Events can boost engagement or revenues and come in many forms. Offers should utilize player data and tools to message players effectively through various channels. The document stresses the importance of tools that allow modifying the game without needing engineers, and having capabilities like analytics, localization, targeting, and modifying game variables for events.
This document discusses the growth of social gaming and its potential social impact. It notes that daily active users of social gaming grew from 50 million in Q4 2007 to over 220 million friend visits in Q3 2009. It also provides examples of large amounts of money raised for social goods through gaming, such as $220,000 raised on YoVille and $830,000 in 2 weeks for social goods on FarmVille, which could provide meals for 500 children for a year. The document suggests that social gaming may have the potential to change the world through facilitating social applications and fundraising for important causes.
The seed round deck of New Legends Studios, a gaming startup.
This includes slides from the reading deck as well as the extra slides part for the presentation deck we used during Q3/Q4 2014. Hopefully, this will help someone out there. If you want the PowerPoint version just ping me on twitter (@civaxo).
You can read the related post over here: http://civax.net/2015/06/new-legends-seed-deck/
Eric Seufert - GDC San Francisco 2016 Presentation - Taking Flight Again: The...Eric Seufert
The document discusses the planning and launch strategy for Angry Birds 2. The key points are:
1) The strategy was to run a massive awareness campaign to generate free installs from earned media and reduce the cost of paid user acquisition by increasing click-through rates.
2) A surge of new Angry Birds 2 players was expected to flow into Rovio's portfolio of games through optimized cross-promotion targeting and improved cross-promotion ads.
3) The goal was to increase retention across Rovio's portfolio and leverage the launch of Angry Birds 2 to boost engagement and monetization for their other games.
Driving profitability of Google App Campaigns in scale. What is easy, what is...GameCamp
Strategic and operational approach. Learnings taken from Huuuge Games campaigns. Which best practices I saw especially important, which ones can be skipped? Presentation delivered by Itay Milstein, Head of Growth from Huuuge Games at 8th edition of GameCamp (www.GameCamp.io).
GDC Talk: Lifetime Value: The long tail of Mid-Core gamesTamara (Tammy) Levy
1) The document discusses lifetime value (LTV) of mobile games and how it is impacted by factors like retention, monetization, and live operations over time.
2) It provides data on typical key performance indicators (KPIs) like retention, ARPDAU, and LTV for different genres and shows how LTV can grow beyond 30 days of play.
3) The author advocates using player retention and monetization curves to more accurately project long-term LTV and how live operations like new content can increase monetization and retention over the lifetime of the game.
Dmitriy Molchanov has worked in game design at Playrix since 2011. He has contributed to several popular mobile games including Fishdom, Township, and Gardenscapes. Gardenscapes in particular has over 10 million daily active users and a team of over 200 people. Molchanov describes some of the key elements of Gardenscapes' match-3 game design including handcrafted levels with clear objectives, seasonal elements to keep the game fresh, and a positive storyline and characters. He also outlines Playrix's product structure and development process, emphasizing the importance of quality control and iterative improvement of ideas and features.
Space Ape's Live Ops Stack: Engineering Mobile Games for Live Ops from Day 1Simon Hade
To view the accompanying video see http://links.spaceapegames.com/liveops
Around half of the $80m revenue generated by Space Ape’s three mid-core build and battle games is attributable to in game events. By adopting a flexible forward looking approach to tools development Space Ape efficiently operates their games with very small non-technical teams maintaining major weekly content update cycles.
In this talk, Space Ape’s senior Live Ops specialists give a demo of their tools and workflows and share the content strategies that have allowed them to grow revenues whilst enabling the studio to focus the majority of its development capacity on creating new games and IP.
DESIGNING SUCCESSFUL LIVE OPS SYSTEMS IN FREE TO PLAY GACHA ECONOMIES
Space Ape shipped Transformers:Earth Wars in the Summer pre-baked with the community events tools that had worked so well in their previous game, Rival Kingdoms. However, they soon realised that many of the old tricks did not apply to the game’s gacha collection economy which had more in common with Kabam’s Contest of Champions than the linear economies of most Build and Battle games. In this talk Space Ape’s Live Ops Lead Andrew Munden (formerly Live Ops Lead at Kabam) will share the content strategies that work in gacha collection games as well as how to build a manageable content furnace and balance player fatigue in a sustainable way.
A BRIEF HISTORY OF IN-GAME TARGETING.
Analytics lead Fred Easy (ex Betfair, Playfish/EA) will share the evolution of his offer targeting technology from it’s belt and braces beginnings to sophisticated value based targeting and the transition to a dynamic in-session machine learning approach.
UNDER THE HOOD: RIVAL KINGDOM'S CMS TOOLS
Game changing content is introduced to Rival Kingdoms every month, with in game events at least every week. Product Manager Mitchell Smallman (formerly Rovio, Next Games) and Steven Hsiao (competitive StarCraft player turned community manager turned Live Ops lead) will demonstrate the content management tools that allow them to keep the game fresh for players without developer support. This will include the tools for configuring competitive events, inserting new content into the game as well as how they measure performance of the changes and optimise on the fly. Learn how these tools enabled them to grow revenue for 6 consecutive months with no marketing spend.
To find out more about the developer go to www.spaceapegames.com
How ASO Has Changed in 2019 and What’s Next. Our experience in creatives and ...GameCamp
Session about experience of Creative Mobile in A/B testing game creatives on Google Play with their mobile title ZooCraft: Animal Family. Optimization Tips and Tricks for Early 2020 and ASO loop as they see in their company. Presentation delivered by Denis Larionov and Serg Kharchenko at 8th edition of GameCamp (www.gamecamp.io).
Quick Introduction to F2P Mobile Game AnalyticsKyle Campbell
- The document introduces key metrics for analyzing free-to-play mobile games such as new users, daily active users, retention rates, revenue metrics like ARPDAU, ARPU, and LTV. It also discusses using data from games, ad networks, and user acquisition to build models and reports.
This document discusses EA Sports, including its company information, market structure, value chain, and management strategies. It provides details on EA Sports' locations, games, competitors, and barriers to entry in the market. Regarding its value chain, it describes EA Sports' role in investment, design, production, distribution, hardware, and reaching end users. Key management strategies include deals with major sports leagues, a long-term ESPN contract, and offering mobile games for free initially to attract gamers.
Improving LTV with Personalized Live Ops Offers: Hill Climb Racing 2 Case Stu...Jessica Tams
This document discusses improving monetization in games through personalized live operations offers using the case study of Hill Climb Racing 2. It describes segmenting players based on both their purchasing behavior and gameplay interactions to create targeted offers. Machine learning models were used to predict player preferences and maximize revenue. This approach resulted in a 52% increase in lifetime value for Hill Climb Racing 2 players through more relevant offers.
Hyper-casual 2.0. Is hyper-casual dead? What we can tell based on the data?GameCamp
What changes do we see in hyper-casual space? What changes in user's behaviour, mechanics of games? In which countries it is dead already, in which countries is it booming?
PlayFab runs a LiveOps backend services platform that handles more than 35 million monthly active players, on more than 450 live games, from studios and publishers that include Miniclip, Rovio, Hyper Hippo, Capcom, Bandai-Namco, and Atari. Getting to that level of scalability hasn’t been easy, and this talk describes the times when PlayFab nearly went down – and what architecture changes we needed to make each time to reach the next level of growth. This talk also shares some of the unique challenges of operating a shared platform, where problems are often not PlayFab’s fault, but always PlayFab’s responsibility, including game bugs that look like DDoS attacks, platform partners who break their APIs, and the joys of cascading server failures.
Idle Games: The Mechanics and Monetization of Self-Playing GamesKongregate
- Idle games are a new genre of self-playing games that have grown popular for their high player retention stats and revenue generation. They allow progress without interaction, rewarding players for returning after periods of idleness.
- Key mechanics include rapid cost/reward growth curves that create a satisfying sense of progress, goals/achievements, and "prestiging" systems that allow resetting games for power boosts. Regular updates and "bumpy" growth curves keep players engaged.
- Popular idle games like AdVenture Capitalist and Clicker Heroes employ monetization strategies like cash infusions, boost multipliers, and protective purchases. Case studies show these can be very profitable for high spend
Advanced approach to Google Universal App campaigns.GameCamp
How to use Google Universal App campaigns for gaming users acquisition? What are tips and tricks that help you manage campaigns effectively? How to structure activities for different game genres?
This document discusses big data analysis and data science. It introduces common data analysis techniques like predictive modeling, machine learning, and recommendation systems. It also discusses tools for working with big data, including Hadoop, HDFS, Pig, HBase, Mahout and languages like R and Python. The document provides an example of using these techniques and tools to build a recommendation system using streaming data from Flume stored in HDFS and analyzed with Pig and HBase.
17th Athens Big Data Meetup - 2nd Talk - Data Flow Building and Calculation P...Athens Big Data
Title: Data Flow Building and Calculation Pipelines via PySpark and ML Modeling via Python
Speaker: Theodoros Michalareas (https://linkedin.com/in/theodorosmichalareas/)
Date: Tuesday, September 24, 2019
Event: https://meetup.com/Athens-Big-Data/events/264702584/
In this study session, we raised a topic about new trends of AI technologies following a combination of deep learning and big data.
It would call for new AI architecture and require new challenge we should do to keep up with front runners.
Google Play store listings tests and optimisation. Outfit7 approach.GameCamp
How to do Google Play store listings tests in the right way? What elements are most important in tests? Benchmarks for tests. Outfit7 approach and examples.
Game monetization and promotion in Asia (with focus on Indonesia, Japan, Sout...GameCamp
What you should know once you want to monetize and promote your game in Asian countries (especially in: Indonesia, Japan, South Korea, Taiwan, Malaysia, Vietnam, Singapore and Philippines). How to create content and creatives? Usage of video and video creatives.
Machine Learning - why the hype and how it does its magicAmirali Charania
This document discusses machine learning and provides an overview of the topic. It begins by explaining the hype around machine learning and how major companies and analysts are recognizing it as an important trend. It then covers machine learning fundamentals, explaining what machine learning is and providing examples of classification and regression problems. The document also outlines the typical machine learning framework and process, including extracting data, developing models, and evaluating performance. Finally, it briefly discusses real-world use cases for machine learning and shows a demo of selecting a machine learning algorithm.
Cross-device: one of biggest marketer's headaches. How to approach it?MariuszGasiewski
How to work with cross-device in XXI century? How to move from idea that "cross-device is important" to real actions and activities. Case studies and examples.
The Future of Enterprise AI Depends on Continuous Quality with Mike GualtieriEggplant
The document discusses advice for successfully implementing enterprise artificial intelligence (AI). It emphasizes the importance of continuous quality for the future of enterprise AI. It recommends setting pragmatic expectations for AI, leaving no use case unexamined, demystifying machine learning for business people, insisting on comprehensive access to enterprise data, knowing when to quit on unsuccessful use cases, using auto-ML to speed up the model building process, scaling with ModelOps, keeping humans involved in decision making, continuously testing AI systems, and using AI to test other AI systems.
This document outlines a project to develop a machine learning model to predict house prices in the United States. It describes the company developing the system, provides background on machine learning and the problem domain, and outlines the objectives, requirements, methodology, design, and expected results of the project. The proposed methodology involves collecting house data, preprocessing it, training a random forest model on 80% of the data and testing it on the remaining 20%, and using the trained model to predict house prices. The system is intended to help buyers search for homes within their budget and avoid being misled on prices.
The document discusses how predictive modelers should approach big data. It notes that while big data and AI are increasingly important, big data places stresses on infrastructure that must be addressed. It recommends leveraging cloud computing and parallel processing to more efficiently handle large datasets. Additionally, it states that simply having big data is not enough - companies must ensure organizational buy-in and address challenges related to people, not just technology.
The document discusses Amazon Web Services (AWS) machine learning capabilities. It provides an overview of the AWS ML stack, which offers the broadest and most complete set of machine learning capabilities across vision, speech, text, search, chatbots, personalization, forecasting, fraud detection, and more. It also discusses several specific AWS machine learning services, including Amazon Rekognition (image and video analysis), Amazon Fraud Detector (online fraud detection), Amazon Kendra (enterprise search), Amazon CodeGuru (automated code reviews and profiling), and Contact Lens for Amazon Connect (contact center analytics).
MastersGate Ecommerce Summit 2018 - Automatizácia a AI v online marketingu - ...Adriana Mučičková
The document discusses automation and AI in internet marketing. It provides definitions for key terms related to machine learning, including optimization, automation, artificial intelligence, machine learning, and deep learning. Charts are presented showing how machine learning can reduce the time spent on repetitive tasks and allow more time for strategic work. Trends toward personalization, accessibility, and improved customer experiences are discussed in the context of online shopping.
The document discusses big data and analytics in supply chains. It provides examples of how companies like IBM are using big data techniques to optimize supply chains and customer experiences. Analytics are categorized from descriptive to predictive to cognitive. Challenges of big data include different data types and volumes. The document advocates for collaboration using analytics to create business value. Speakers from IBM, Analytic Impact and Supply Chain Insights discuss their work applying big data and analytics.
Similar to How to start working with LTV measurement in mobile gaming? How to move to advanced strategies? (20)
Practical approach to creative testing and creative optimisation at Google UA...GameCamp
This document provides best practices for optimizing mobile game creatives on Google Ads. It discusses identifying the best performing placements like AdMob, display, YouTube, and search. It also discusses targeting the audience by focusing creatives on core gameplay or genre. Creative testing should be done in separate ad groups to see performance in different networks before moving top performers to evergreen ad groups. Frequency of creative uploads and removing low performers after 7-14 days is recommended. Insights from universal app campaigns can also be applied to optimize across different networks.
Creativity and science behind creative testing. creative testing framework an...GameCamp
Presentation of Claire Rozain (from Product Madness) delivered at one of GameCamp webinars dedicated to Creative testing and optimisation (www.gamecamp.io).
How to combine creative and data approach in creative testing? How to make it as the process and learning loop and not one-off? Presentation based on examples and learnings from different campaigns and traffic sources.
Blog of Claire Rozain: https://ClaireRozain.fr.
8 Types of mobile game and app creatives you should tryGameCamp
Over that past three years, we have worked with around 80 apps and tested approximately 8,000 different videos. From this, we’ve been able to recognize and define the characteristics of certain types of concepts that frequently perform well. We have also seen these concepts work for specific app categories.
In this presentation, we would like to share our typology and discuss the game categories where we’ve seen these specific concepts perform particularly well. We hope this may help in building or improving the testing plan for your game and, as a consequence, achieve better results in UA.
Presentation done by Adam Turowski, Co-Founder of Miri Growth at GameCamp webinar: https://youtu.be/yg7aPWmyz2o?t=396
Building the BI system and analytics capabilities at the company based on Rea...GameCamp
Building the BI system and analytics capabilities at the company. How we built it and grew its capabilities?
How we developed our analytics capabilities? What was easy, what was more difficult. Preparation, process step by step. Presentation from GameCamp webinar: http://www.gamecamp.io/events/webinar-growing-measurement-capabilities-of-gaming-and-apps-company/
Ad-hoc tasks in the Data Scientist team based on Outfit7 exampleGameCamp
This document discusses key performance indicators (KPIs) and ad-hoc analyses at the data and analytics department of a company. It outlines the main KPIs such as revenue, retention, and advertising metrics that are important for monitoring. It also describes how ad-hoc analyses are conducted for issues like product support, data anomalies, and app crashes. Regular monitoring of KPIs and conducting relevant ad-hoc analyses are important for optimizing products and addressing issues.
Scaling UA activity - the challenges of growthGameCamp
What to look out for when trying to scale up the user acquisition. What metrics to choose, how to approach this topic, what to measure and how to decide if we are successful or not. Presentation run by Lukasz Kwiecien, Head of Marketing of Gamesture at 8th edition of GameCamp (www.GameCamp.io).
Growth Strategies: Lower Your CPI by Implementing ASO into UA Workflow.GameCamp
To have a sustainable, repeatable and balanced Acquisition Strategy you’ll need to think of it as system of compounding loops. And although the Paid Acquisition loop (UA) is crucial part of it, ignoring the Organic Acquisition loop (ASO) might mean the difference between a game that generates 7-figures or 8-figures revenue per year. Presentation based on examples and case studies. Presentation delivered by Erik Hegely, Head of Growth of Pixel Federation at 8th edition of GameCamp (www.GameCamp.io).
Growth trends in mobile gaming based on the dataGameCamp
What Growth trends can we see in different gaming genres and in different countries? How Organic and Paid is behaving in Growth? Based on the Google and industry data. Presentation delivered at 8th edition of GameCamp (www.GameCamp.io).
Facebook's and Social Creative Best Practice that worked for HuuugeGameCamp
Practical learnings from Huuuge Games' experience at Facebook creatives. What worked for us what did not. How Facebook creatives should be different than Google creatives? Presentation delivered at 8th edition of GameCamp (www.GameCamp.io).
Using Data Science for Behavioural Game DesignGameCamp
In this talk, the speaker shows how data science and behavioral science can help to design better games, increase retention, and in-app purchases, and achieve viral growth. Presentation delivered at 8th edition of GameCamp (www.GameCamp.io).
Learnings and insights from Benchmarks+. Performance Insights for 1.2bn Month...GameCamp
GameAnalytics shares global gaming trends surfaced from the Benchmarks+ platform. Their analysis spans more than 88,000 games, 150 million transactions, 1.2 billion players and 16 billion sessions on average each month - with performance insights dating back to 2016. Expect to learn the latest industry trends for core engagement and monetization KPIs like retention, session length, ARPPU and conversion across all major genres and sub-genres, now including aggregated insights for the top 5%, 10%, and 20% of market performance. This is one of the most detailed pieces of gaming analysis currently available for developers, studios and publishers of all shapes and sizes.
In-app offers monetization from basics to advanced technics based on Crazy Pa...GameCamp
From simple user segmentation to a complex interconnected system, we will follow Crazy Panda`s progress in developing personalized monetization. The key message of the session is not only about when or who to show those offers to, but also about how to make them truly personal. Presentation delivered by Ivan Kozyev, Head of Analytics of Crazy Panda at 8th edition of GameCamp (www.GameCamp.io).
Building cost-effective mobile product & marketing app analytics based on GCP...GameCamp
Ever wondered whether it makes sense for your company to build your own app analytics stack? This session will feature a case study about us doing just that, using tools like Google Firebase, BigQuery, and Data Studio. Martin Jelinek guided the audience through the whole process, including the initial decision making (why we chose to build, what tools to use), execution (events and metrics definitions, implementation and testing, cost control), and a final evaluation of pros and cons of such an approach & our learnings. Presentation run by Martin Jelinek (www.appagent.co) at 8th edition of GameCamp (www.gamecamp.io)
Enabling the growth of the game through performance marketing in 2020GameCamp
Every year performance marketing is changing and with it, growth strategies. This session covers in this topic, at the beginning some of the general rules in today's market and how to grow a game profitably through performance marketing channels and what are the most important metrics to keep in mind. In the next part, we will connect those growth metrics with signals which can be given to the Product/Game teams to help make good decisions. Presentation is be based on activities that Nordeus is doing to achieve such goals. Presentation delivered by Andrej Kugonic at 8th edition of GameCamp (www.GameCamp.io)
Which markets you should focus on in mobile gaming? Where is the highest LTV versus CPI. Based on Google data. Abstract from presentation from 7th edition of GameCamp
LTV measurement and multi-touch attributionGameCamp
Can multi touch attribution models help you increase the LTV of your campaigns? How should you split the credit for revenue generated by users who saw multiple ads? What is easy what is more difficult?
Big Problem, BigQuery: User Feature Engineering in Event-driven AnalyticsGameCamp
How to find best predictor of good LTV in a game (feature/event in a game) ? What is needed for good feature selection? How to construct and aggregate data well? How it can be used in practical way? Strategic and operational approach.
Not so trivial lessons learned from launching mid-core games.GameCamp
Key points on understanding future marketing and analytics challenges when developing ideas for your future games. We analyzed 1000 games of different genres and extracted common types of how the key metrics of apps develop after worldwide launch. We analyzed only notable games with the 1st year sum of revenue + downloads more than 2M and compared their performance during their first year. Learnings from the analysis.
Ad LTV in mobile gaming: deep dive into industryGameCamp
What is the situation in Ad LTV right now, what is a good benchmark for mobile game today? Data based on analyzing unique and granular data collected over a period of months and based on billions of in-app ad impressions. What are good and bad best practices looking at games from many genres? What is must have and what is nice to have? What should be most important for smaller company, what should be the focus for bigger organisation?
What monetisation trends can we see? Based on the data and predictions Which gaming genres are getting better in monetisation, which ones are getting more challenging? Abstract of presentation shown at 7th edition of GameCamp
How to start working with LTV measurement in mobile gaming? How to move to advanced strategies?
1. Proprietary + Confidential
How to start working with
LTV measurement in
mobile gaming. How to go
to advanced strategies?
Mariusz Gąsiewski,
gasiewski@google.com
Warsaw, 17.9.2018
2. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Not all users are the same!
3. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Payer fraction growth in mobile F2P games
Source, DeltaDNA, September 2018
4. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
It is not enough
5. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Example approaches in the field
Historical Average or Cohort Simple Predictive Models Advanced Predictive Models
Example:
Probability Models
combined with machine learning
Pro:
Very high accuracy, can become significant
competitive advantage for a company in
monetisation once done right.
Con:
Requires data science team and significant
amount of data.
High computational power needs
Historical Average or Cohort Simple Predictive Models
Example:
Feature selection using Random Forest
Pro:
Strong outcome, possible to do with pretty
limited ressources and for middle size
games. High actionability.
Con:
Less accurate than advanced statistical
methods. Data scientist experience strongly
recommended.
Con:
Doesn’t account for differences between
customers. Combined with strong risk once
not done carefully.
Pro:
Easy to compute with average revenue per
user. Good as the start even for small
studio. Actionable once done at the country
level.
Example:
Simple models based on spreadsheets and
formulas
6. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
It really matters...
Version 1 Version 2
7. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
K-factor analysis example
8. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
9. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Example LTV calculator
Example LTV calculator
https://goo.gl/A4fmKc
10. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Other simple approaches
Source: http://www.gamedonia.com/blog/5-ways-to-calculate-lifetime-value-for-free-to-play-games
11. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Example approaches in the field
Historical Average or Cohort Simple Predictive Models Advanced Predictive Models
Example:
Probability Models
combined with machine learning
Pro:
Very high accuracy, can become significant
competitive advantage for a company in
monetisation once done right.
Con:
Requires data science team and significant
amount of data.
High computational power needs
Historical Average or Cohort Simple Predictive Models
Example:
Feature selection using Random Forest
Pro:
Strong outcome, possible to do with pretty
limited ressources and for middle size
games. High actionability.
Con:
Less accurate than advanced statistical
methods. Data scientist experience strongly
recommended.
Con:
Doesn’t account for differences between
customers. Combined with strong risk once
not done carefully.
Pro:
Easy to compute with average revenue per
user. Good as the start even for small
studio. Actionable once done at the country
level.
Example:
Simple models based on spreadsheets and
formulas
12. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Is first payment good prediction of LTV?
13. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Feature selection
14. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Can be used for effective UA
15. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
What makes feature selection?
Source, http://blog.citizennet.com/blog/2012/11/10/random-forests-ensembles-and-performance-metrics
16. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Red dots represent users that made in-app purchase
Blue dots represent users that didn’t make in-app purchase
Imaginary Traditional View
17. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Red dots represent users that made in-app purchase
Blue dots represent users that didn’t make in-app purchase
Imaginary Traditional View Actual Case
18. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
...to more advanced approaches.ExtraRandomForest
A supervised learning algorithm that builds multiple decision trees and merges them together to get a
more accurate and stable prediction
19. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Source: https://www.machinelearningplus.com/machine-learning/feature-selection/
20. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Source: https://hub.packtpub.com/4-ways-implement-feature-selection-python-machine-learning/
21. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Example
22. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
MoonTon increases high
quality user volume by
leveraging Feature
Selection Methodology
Background
MoonTon was founded in 2014 in
Shanghai, focusing on developing
high quality mobile games. It has
become well-known in SEA
markets for its hero game’s
achievements both in install aspect
and revenue aspect.
Goal
MoonTon values the power of
machine learning and tries to
explore a way to balance the
installers’ quantity and quality from
game ecosystem perspective.
Approach
• Utilize Feature Selection
algorithm model to find out the
common in-app actions that high
value players act.
• Start UAC Actions campaigns
which optimize towards feature
selection events.
“We are impressed by the effect of
machine learning technology and we will
surely invest more and dig deeper to
unleash its potential next.”
--Cai Xuwei
Marketing Head
MoonTon
+18%
retention rate
-52%
cost/retention
+423%
ROI
23. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Example approaches in the field
Historical Average or Cohort Simple Predictive Models Advanced Predictive Models
Example:
Probability Models
combined with machine learning
Pro:
Very high accuracy, can become significant
competitive advantage for a company in
monetisation once done right.
Con:
Requires data science team and significant
amount of data.
High computational power needs
Historical Average or Cohort Simple Predictive Models
Example:
Feature selection using Random Forest
Pro:
Strong outcome, possible to do with pretty
limited ressources and for middle size
games. High actionability.
Con:
Less accurate than advanced statistical
methods. Data scientist experience strongly
recommended.
Con:
Doesn’t account for differences between
customers. Combined with strong risk once
not done carefully.
Pro:
Easy to compute with average revenue per
user. Good as the start even for small
studio. Actionable once done at the country
level.
Example:
Simple models based on spreadsheets and
formulas
24. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Which customer is more valuable to your business?
CPI
Value generated in Month 1
A B
$9 $19
$10 $10
Value generated in Month 2 $0 $10
ROI end of Month 1 11% -47%
Value generated in Month 3 $0 $10
ROI end of Month 3 11% 58%
25. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
BTYD: predictive model based on 3 inputs (RFM)
Unique
identifier
Transaction
value
Date of
transaction
26. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
BTYD: example of data output (prediction)
Customer
ID
P(Alive)
Predicted
Transactions
Predicted
AOV
Predicted
Value
1 39.0% 0.2 $68.34 $16.36
2 99.7% 31.3 $45.53 $1,424.94
3 99.1% 7.8 $122.52 $958.64
4 97.9% 7.6 $62.29 $475.35
5 81.3% 2.2 $85.51 $183.55
6 99.9% 9.0 $126.15 $1,138.03
7 97.5% 13.9 $95.65 $1,328.45
8 67.0% 2.3 $167.61 $379.78
9 28.9% 4.8 $230.58 $1,017.02
10 52.8% 1.6 $126.83 $202.36
27. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
BTYD: Example of an RFM model at work
28. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
BTYD: Example of an RFM model at work
Looking ahead, academic research shows that at Week 78 Customer A is actually
more valuable than Customer B.
Predictive Models help us understand and value this future behavior
29. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Scalable Machine Learning System Successfully Onboard
Cloud & Big Query
Data Architecture / Processing
Data Analytics
TensorFlow Marchine Learning
Predictive Payer + pLTV
Business Intelligence
Reporting via Data Studio
30. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Clash of Kings Boosts
CVR up by 7X with
TensorFlow & UAC
Approach
• Build a TensorFlow Deep Learning Model on
Google Cloud
• Predict gamers’ likelihood to pay based on
hundreds of behavioral signals with an automated
Machine Learning System
• Leverage Google UAC to acquire “will pay gamers“
based on predictive signals
“TensorFlow & Machine Learning are awesome tools, It
helped us explore a smarter way of acquiring high value
customers.”
-- Bin Gao
Senior UA Lead
Elex Clash of Kings
2.6X
ROAS
Over 7X
Paid User/Installer
63% decrease
of cost per paid
user
31. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Example approaches in the field
Historical Average or Cohort Simple Predictive Models Advanced Predictive Models
Example:
Probability Models
combined with machine learning
Pro:
Very high accuracy, can become significant
competitive advantage for a company in
monetisation once done right.
Con:
Requires data science team and significant
amount of data.
High computational power needs
Historical Average or Cohort Simple Predictive Models
Example:
Feature selection using Random Forest
Pro:
Strong outcome, possible to do with pretty
limited ressources and for middle size
games. High actionability.
Con:
Less accurate than advanced statistical
methods. Data scientist experience strongly
recommended.
Con:
Doesn’t account for differences between
customers. Combined with strong risk once
not done carefully.
Pro:
Easy to compute with average revenue per
user. Good as the start even for small
studio. Actionable once done at the country
level.
Example:
Simple models based on spreadsheets and
formulas
32. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Worth to look at
Source: https://www.youtube.com/watch?v=-ZEXx6PNRIc
33. Source: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis non erat sem
Proprietary + Confidential
Worth reading
Source: https://medium.com/googleplaydev/predicting-your-games-monetization-future-ce176169b056