why successful games need analytics

73
Data Science Club #3 Why Successful Games Need Analytics Ivan Trancik, CEO @ Cellense 1

Upload: data-science-club

Post on 22-Jan-2018

147 views

Category:

Engineering


4 download

TRANSCRIPT

Data Science Club #3Why Successful Games Need Analytics

Ivan Trancik, CEO @ Cellense

1

Data Science Club #3

Agenda

1. Introduction

2. Who is behind successful games?

3. How to apply analytics to games?

a. Market research (Nex Machina)

b. Player Experience (Hill Climb Racing 2)

c. Monetization (Unkilled)

4. Q&A

Data Science Club #3

Hello, my name is Ivan ([email protected])Serial Big-Data entrepreneur

Co-founder @ Infinario (now Exponea)

● Real-time Marketing Cloud

● Maybe you’ve already heard about it :-)

Founder & CEO @ Cellense, BuffPanel

● Full-stack Game Business Analytics

● Bootstrapped, 10+2 employees

● 3 Top #100 Grossing Mobile Games (US)

● 3 Top #1 Best-selling PC Games

Data Science Club #3

About Cellense

1. Introduction - Gaming Market

Data Science Club #3

Introduction - Global Gaming Market

2. Who is (not) behind successful games?

Data Science Club #3

Indie developer?

Data Science Club #3

Marcus “Notch” Persson (Mojang) - Minecraft (sold for $2.5 billion)

Data Science Club #3

Ilka Paananen (Supercell) - Clash Royale ($10 billion valuation)

Data Science Club #3

Marek Rosa (KeenSWH) - Space Engineers (2.5 millions copies)

Data Science Club #3

Šimon Šicko (Pixel Federation) - Diggy’s Adventure (30+ million EUR revenue 2017)

3. Analytics & Games

Data Science Club #3

Games & Data

Games:

1. Global B2C type of product

2. Work only in scale

3. Digital (& online) by design

Data Science Club #3

Games & Data

Games:

1. Global B2C type of product

2. Work only in scale

3. Digital (& online) by design

=> vast amounts of valuable data!

Data Science Club #3

Games & Data - Applications

Games:

1. Global B2C type of product

2. Work only in scale

3. Digital (& online) by design

=> vast amounts of valuable data!

=> interesting optimizations on all fronts to be made :-)

3.a Market Research

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)

Hard questions:

1. Why so underwhelming sales?

a. PR too small?

b. Niche too small?

2. What did more successful games do differently?

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)

Data sources

1. SteamSpy

2. SteamAPI

3. Google Trends

4. Nex’s data

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)

Nex clearly outperformed median game at much higher price

point.

Niche overview:

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)

Out of over 1000 games in its niche, only 3 (+Ruiner) relevant

games had better sales than Nex!

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)● General info

● #owners

● Price

● Playtime total

● Userscore

● Scorerank

● Developer

● Pre-release

○ News articles timeline

○ Pre-release search trends

● Post-release search trends

● Release info

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)News articles mentioning Helldivers pre-launch

The peak in news count was almost 10 months before release. Then there was steady

influx of new articles, with bumps every 3-4 months.

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)News articles mentioning Nex Machina pre-launch

The peak in news count was almost 6 months before release. Coverage was comparable

to other games.

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)Organic interest for Helldivers pre-launch

PR push resulted in steady interest in the upcoming months.

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)Organic interest for Nex Machina pre-launch

PR haven’t resulted into any significant organic interest.

Data Science Club #3

Market Research - Nex Machina (Steam, PS4)

3.b Player Experience

Data Science Club #3

Hill Climb Racing 2 - about

Fingersoft

Released Q4 2016

iOS & Android

multiplayer racing game

120 million players

(HCR1 over 700 millions)

Data Science Club #3

Hill Climb Racing 2 - gameplay

Fingersoft

Released Q4 2016

iOS & Android

multiplayer racing game

120 million players

Data Science Club #3

Very short soft-launch (8 weeks)

“Grow KPIs as much as possible”

=> Roadmap prioritization based on the data!

=> FTUE / Early Retention Only

Retention

Data Science Club #3

Level progression funnel - wins only

Retention - Level balancing

Data Science Club #3

Level progression funnel - w/ drop-off after win

Retention - Level balancing

Data Science Club #3

Churn rate - for how many players is the level last (after win & loss)

Retention - Level balancing

Data Science Club #3

Fail rate & failed-over-passed rate - how many times players fail for one completion?

Retention - Level balancing

Data Science Club #3

FUU factor - is losing frustrating or hopeful?

Retention - Level balancing

Data Science Club #3

First time fail - at which level player loses the first time / loses all lives for the first time

Retention - Level balancing

Data Science Club #3

Play time - how long does it take to win or lose?

Retention - Level balancing

Data Science Club #3

Crucial business bottleneck: early matches after player loses

HCR2 is a PvP game => let’s rebalance matchmaking!

Retention

Data Science Club #3

Tweaking win-rate to optimize retention.

What’s the best win-rate for a game like HCR2? 50%? 60%? 80%?

Retention

Data Science Club #3

Tweaking win-rate to optimize retention.

What’s the best win-rate for a game like HCR2? 50%? 60%? 80%?

=> Let’s try them all :-)

Retention

Data Science Club #3

Retention

Data Science Club #3

Retention

=>

Data Science Club #3

Early-retention

● Do a battery of player drop-off tests across multiple variables to be sure you’re

always prioritizing the most important issues

● Always confirm hypotheses by A/B tests

Retention - Summary

Data Science Club #3

LEVEL BALANCING CHALLENGES

- find issues in level design

- optimize difficulty curve

- balance revenue against churn

- improve design of blocker levels

Retention - Level balancing

Data Science Club #3

LEVEL BALANCING CHALLENGES

- find issues in level design

- optimize difficulty curve

- balance revenue against churn

- improve design of blocker levels

SOLUTION

- insights based on level balancing analyses

Retention - Level balancing

Data Science Club #3

LEVEL BALANCING CHALLENGES

- find issues in level design

- optimize difficulty curve

- balance revenue against churn

- improve design of blocker levels

SOLUTION

- insights based on level balancing analyses

Retention - Level balancing

ANALYSES

- level progression funnel

- win & fail rate

- failed-over-passed rate

- churn rate (after win & loss)

- time spent per fail & win

- game specific

- boosters used

- stars achieve

- first time fail

- …

3.c Monetisation

Data Science Club #3

Huge topic

● Retention

● Core Game design

● Currency spend onboarding

● Economy Balancing

...

● Live-ops Offers

Monetization

Game Executive 2017

Monetization - Live-ops Offers

Game Executive 2017

Monetization - Live-ops OffersSTP Model

Well describes the process in

these questions:

● WHO?

● HOW MUCH?

● HOW OFTEN?

● WHAT?

Game Executive 2017

Monetization - Naive approach (Segmentation)Simple segmentation based on spend

Advantages:

● Easy to understand

● The most widely used

● Enables basic targeting

Game Executive 2017

Monetization - Naive approachSimple segmentation based on spend

Disadvantages?

Game Executive 2017

Monetization - Naive approach (Segmentation)Simple segmentation based on spend

Disadvantages:

● Not taking into account different

spending patterns at the same

cumulative spend

● Not answering “WHAT?”

at all

● Not taking player’s activity into

an account

Game Executive 2017

Monetization - Business Analytics approach (Segmentation)

RFM SegmentationRecency, Frequency and Monetary value

● Directly models purchasing behavior

● Great for predicting purchase

effectiveness

● Easy to understand and apply in content

creation process

● Basis for targeted offers

● Still no answer to WHAT?

Game Executive 2017

Monetization - Business Analytics approach (Segmentation)

Player Progression SegmentationPlayers in the game are not the same -

● They unlock different content

● Progress to different points in the story mode

● Different split between game modes

● Different approach for spending currencies.

Game Executive 2017

Monetization - Business Analytics approach (Segmentation)

RFM Segmentation Player Progression Segmentation

Game Executive 2017

Monetization - Business Analytics approach (Targeting)

Number of all segments can be pretty high - we can focus only on a few

● Number of players

● Their payment potential

● (Un)-Availability of content

● Spending habits (which ones are spending premium currencies the most)

Game Executive 2017

Unkilled - about

Madfinger Games

Released Q3 2015

iOS & Android

zombie FPS with PvP

6 million players

Game Executive 2017

Unkilled - gameplay

Madfinger Games

Released Q3 2015

iOS & Android

zombie FPS with PvP

6 million players

Game Executive 2017

Monetization - Business Analytics approach (Example Targeting)

Identified Targetings

● Competitive Hard-core Non-payers

● First-time Payers without Premium Content

● Elite Spenders

● Newbies & At Risk Players

● Hooked Non-payers

Game Executive 2017

Monetization - Business Analytics approach (Example Targeting)

Identified Targetings

● Competitive Hard-core Non-payers

● First-time Payers without Premium Content

● Elite Spenders

● Newbies & At Risk Players

● Hooked Non-payers

Game Executive 2017

Monetization - Business Analytics approach (Example Targeting)

Competitive Hard-core Non-payers

● Last two story tiers.

● Ranked in PvP

● Already completed at least 1 rare or epic weapon

● Interaction with Premium Currency spending

Game Executive 2017

Monetization - Business Analytics approach (Example Offer)

Competitive Hard-core Non-payers

● $5 Price tag (From RFM segmentations)

● Communicate Extra Value for Competitive Players

● Offer stuff they already like to use

Game Executive 2017

Monetization - Business Analytics approach (Example Offer #1)

4 VIP chests + 2 bonus VIP chests (hook)

Regular value is 4 VIP chests for $5, if offer included only

chests it should be interesting

Random hero skin (incentive)

Every player has at least 1-2 heroes at this stage.

12x combination of VIP gadgets (incentive)

VIP gadgets are powerful for PvP, 12 gadgets should suffice

for 12 matches which is plenty

Game Executive 2017

One-time crafted offers

● Created based on an actual player’s payment potential and their current game

progression

● Confirmed by A/B tests

Monetization - Results

Data Science Club #3

● Player segmentation can be a very effective tool to drive the revenue

● 2x-10x revenue during live-ops events

=> Could be as much as 80% of the revenue!

Monetization - Results

=> WE’RE HIRING! <=

[email protected]