34
BIG DATA Opportunities and Challenges Use Cases in PT.Telekomunikasi Indonesia Komang B Aryasa Big Data Project Director [email protected] Jul 2nd, 2015

komang@ - Big Ideas with Big Data in Indonesia · PDF fileBIG DATA Opportunities and Challenges Use Cases in PT.Telekomunikasi Indonesia Komang B Aryasa Big Data Project Director komang@

Embed Size (px)

Citation preview

BIG DATA Opportunities and Challenges

Use Cases in PT.Telekomunikasi Indonesia

Komang B Aryasa Big Data Project Director

[email protected]

Jul 2nd, 2015

Contents

1   Introduction & Background

2   Big Data in PT Telekomunikasi Indonesia

3   Use Cases in PT Telelkomunikasi Indonesia

Introduction & Background

Huge Number of events happen in 30 seconds .... People always connected - anywhere, anytime

Social Media Landscape

Source: eMarketer

 

Examining  more  than  400  large  companies,  Bain  &  Co  found  that  those  with  the  most  advanced  analy>cs  capabili>es  are  outperforming  compe>tors  

Source: Bain Company, business recorder, insurance journal

•  Gartner: Average service provider could potentially generate additional margin of $300M/year

•  McKinsey: Data-driven Companies are 5% more productive and 6% more profitable

The effect of Big Data (Analytics) to Company Performance

Source: Bain & Company 2013, IDC 2014

Indonesia Big Data Market size predicted to be USD 41,8 Mn in 2017 (CAGR 39,5)

Global

Indonesia  Big  Data  Market  Size  ($  Mn)  

2012   2013   2014   2015   2016   2017   CAGR  

7,9   11,0   15,3   21,4   29,9   41,8   39,5%  

Global  Big  Data  Spending  projected  to  be  USD  47,1  Bn  in  2020  growing  from  USD  14  Bn  in  2015    

Global & Indonesia Big Data Market Size Companies are Spending BIG on Big Data …

Big Data Market in Indonesia

Source: IDC

Big Data in PT Telekomunikasi Indonesia

Business Model & Challenges

Internal Use Cases

•  Internal Business Problem •  Decision Support System •  Cross-channel •  Contextual

Customer Experience Use Cases

•  Voice of the Customer •  Customer Experience

External Use Cases

•  Where is the money? •  Which Verticals ? •  What skills/tools required

Data Monetizing

2 31

Focus Areas for CSPs When Exploiting Data/Analytics Opportunities Source  :  Gartner  

Data

Enterprise Information

Management

Data Governance

Master Data Management

People Process Technology

Data Organization

Chief Data Officer

Process Maturity

KPI Development

Logical Data Warehouse

Analytics

Distributed Processing

Programming & Databases

Communication & visualization

Domain Knowledge & Soft Skills

Math  &  Sta>s>cs  

•  Machine Learning •  Statistical Modelling •  Experiment Design •  Supervised learning,

decision trees, Bayesian inference, logistic regression

•  Able to engage with senior management •  Story telling skills •  Translate data driven insights into

decision and actions •  Visual art design •  R packages •  Knowledge of visualization

•  Passionate about business •  Curious about data •  Influence without authority •  Hacker mindset, Problem solver •  Strategic, proactive, creative, innovative,

and collaborative

People

Data Scientist

•  Computer science fundamentals •  Scripting language e.g python •  Statistical computing package •  Database SQL & NoSQL •  Relational algebra •  Parallel databases and parallel query •  MapReduce concepts

Technology

Unsructured & Semi Structured Data Structured Data

Data Mining

Interactive Analytics

Data Quality �

Data Integration, Storage & Computing �

RTD

CEP/PME

Customer Knowledge Mgmt.

Internet Behavior Analytics �

Location Analytics

Real Time Operations �

Data Governance �

Text �Network Data � CRM �Billing � Web �Video �

Campaign Business Intelligent CRM Reporting &

Visualization

Data Security �

User Mgmt

Deploy

Operations Management�

Configure

Maintenance

Fault

Performance

A&A

Enterprise Data Warehouse

Acqu

ired  

Accessed

 An

aly>c  

Applica>

on  

Audio �

4A

Analytics �

Data Sources �

To  handle  VOLUME  of  Data   To  handle  VARIETY  of  Data   To  handle  VELOCITY  of  Data  

Product Analysis �

Paralelize Data Mart Query

Big Data Integration Platform   Streams �

Sqoop Flume Hive

Original Data Transformed Data

Use Cases in PT. Telekomunikasi Indonesia

Use Cases

01 Broadband Churn Prevention

02 Monetizing 1 Million SME

03 Brand Tracking & Monitoring

04 Government Social Media Monitoring

Use Case Broadband Churn Prevention

01  

Use Case Broadband Churn Prevention

1.  Decreasing churn rate from 5% to 2%

2.  Building prediction model to anticipate customer that will churn

3.  Maintaining existing revenue growth & Manage retention program & loyalty

Objective :

Revenue, Payment, Payment method, Debt, Product/Service performance beliefs

Billing, Payment

Inbound call, outbound call, visit

Customer Complaint

Paket, Pricing, promotion, place, bundling, dll

Portfolio Product

Customer location mapped to regional offices

Customer Flag

FTTH, Copper,MSAN,SNR, R2BB

Network

Age, income, Gender

Demography

1

2

3

5

6

7

Top Domain access, Hits, Cookies, Interest, Hobbies

Usage Pattern

Call Usage, Broadband Usage, Browsing behavior

Pricing

4 8

Data Sources

Window Definition

•  Window observation : Period during which customer behavior is analyzed

•  Window action : Period during which the customers are scored

•  Window prediction : Period during which the attrition behavior of the customer is predicted

Observa>on  Windows  6  Months  

Jun  –  Dec  2014  

Ac>ve  Windows    

January    2015  

Predic>on  Windows    

February  2015  

Jun  14   Dec  14   Jan  15   Feb  15  

Observa>on  Window   Ac>on  window   Predic>ve  window  

CT0  in  Jan  2015  =  33.403  

May  14   Nov  14   Dec  14   Jan  15  

Apr  14   Oct  14   Nov  14   Dec  14  

Sampling to Build The Model

CT0  in  Dec  2014      

CT0  in  Nov  2014    

Y Total  Customers %Percents

Actives 1,241,529                                             93.0%

Churn 92,117                                                         7.0%Total 1,333,646                                             100.0%

Total  data  sampling  for  building  Churn  Model    

CT0  :  Change  tariff  to  zero  

Model Accuracy

Apr-­‐15  

Predicted Actual Total 0 = Stay 1 = Churn

0 = Stay 1.303.843 17.122 1.320.965 1 = Churn 66.466 18.894 85.360 Total 1.370.309 36.016 1.406.325

1   Model  1  (LOS  <  6  month)  

Mar-­‐15  

Predicted Actual Total 0 = Stay 1 = Churn

0 = Stay 1.279.841 20.470 1.300.311 1 = Churn 92.957 19.456 112.413 Total 1.372.798 39.926 1.412.724

Feb-­‐15  

Predicted Actual Total 0 = Stay 1 = Churn

0 = Stay 1.318.427 17.336 1.335.763 1 = Churn 54.381 16.805 71.186 Total 1.372.808 34.141 1.406.949

accuracy  :  88,5%   Sensi>fity:  52,5%   Spesifity:  89,2%  

Accuracy  :  89,8%   Sensi>fity:  60,5%   Spesifity:  91,1%  

Accuracy  :  93,2%   Sensi>fity:  56,2%   Spesifity:  94,6%  

2   Model  1  (LOS  >  6  month)  

Apr-­‐15  

Predicted Actual Total 0 = Stay 1 = Churn

0 = Stay 1.148.901 14.986 1.163.887 1 = Churn 47.757 13.808 61.565 Total 1.196.658 28.794 1.225.452

Accuracy  :  94,9%   Sensi>fity:  48%   Spesifity:  96%  

Mar-­‐15  

Predicted Actual Total 0 = Stay 1 = Churn

0 = Stay 1.112.536 17.334 1.129.870 1 = Churn 76.615 14.654 91.269 Total 1.189.151 31.988 1.221.139

Accuracy  :  92,3%   Sensi>fity:  45,8%   Spesifity:  93,6%  

Feb-­‐15  

Predicted Actual Total 0 = Stay 1 = Churn

0 = Stay 1.137.569 14.157 1.151.726 1 = Churn 44.008 12.726 56.734 Total 1.181.577 26.883 1.208.460

Accuracy  :  95,2%   Sensi>fity:  47,3%   Spesifity:  96,3%  

Result

Network  &  Package      

Network  type   MSAN  Internet  Package   New  Speedy  -­‐  Reguler  

512K  Zone  2  

Customer  Performance  (Average  )  Internet  Usage  Total  (MB)   2362  Internet  Upload  (MB)   213  Internet  Download  (MB)   2148  Internet  Billing  (Rp)   99584  Internet  Billing  before  Tax  (Rp)   90531  Internet  Payment  (Rp)   102876  Internet  outstanding  (Rp)   5761  Number  >cket  created   0  

Network  stability      

Count  SNR  less  than  13   1  

Last  R2BB  Measurement   0  

Modeling  Result      

Probability  to  be  churn   4.70%  Probability  to  be  stay   95.30%  Predicted  value   Stay  Top  3  Important  Variable  (Root  cause)  

*  R2BB  last  month  *  Outstanding  last  month  *  SNR  last  month  

Suggesaon   Cross-­‐sell  

IBU  Y,Kota  Pekalongan  Jawa  Tengah  Customer  :  Fixed  lines  &  Internet  Length  of  stay    <  6  months  

Upgraded  to  INDIHOME    

High  Usage  

Next  Best  

Offered    

Result

Next  Best  Offered    

Customer  Profile      

Customer  number   128622  Name   P  HIMAWAN  Address   Jalan  Ba>k  Rengganis  19,  

Sukaluyu  

City   Kota  Bandung  Region   03  No  Speedy   131169107874  No  Telp   222504986  LOS  Speedy  (months)   36  LOS  Customer  (months)   141  

Network  &  Package      

Network  type   COPPER  R2BB  code   SP4  R2BB  (MB)   4096  Internet  Speed   INET512RH  Internet  Package   Speedy  SOCIALIA  512  Kb-­‐  

HSSP  

Customer  Performance    (Average  During  Last  6  Months)  Internet  Usage  Total  (MB)   0  Internet  Upload  (MB)   0  Internet  Download  (MB)   0  Internet  Billing  (Rp)   214500  Internet  Billing  before  Tax  (Rp)   195000  Internet  Payment  (Rp)   151125  Internet  outstanding  (Rp)   76375  Number  >cket  created   0  Number  of  call   0  Number  of  mobile  call   0  Number  of  int.  Call   0  POTS  billing  (Rp)   52360  POTS  outstanding  (Rp)   -­‐4407  

Network  stability      

Count  SNR  less  than  13   0  

Last  R2BB  Measurement   0  

Modeling  Result      

Probability  to  be  churn   83.07%  Probability  to  be  stay   16.93%  Predicted  value   Churn  Top  3  Important  Variable  (Root  cause)  *  R2BB  last  month  

*  All  outstanding  last  month  *  Freq.  Internet  download  decrease  

Suggesaon   Retenaon  program  

Bapak  X,  Kota  Bandung  Jawa  Barat  Customer  :  Fixed  lines  &  Internet  Length  of  stay    >  6  months  

RETENTION  PROGRAM    

Use Case Monetizing 1 Million SME

02  

Small  Medium  Enterprise  (SME)    

Profiling  &  Exploi>ng  Data  SME  

Acquisi>on  New  Customers  

Leveraging  Customer  Exis>ng  

Opportunity  areas  

q Data  collec>on  and  cleansing    q Standardize  data  format  q Data  normality    q Data  visualiza>on  q Data  segmenta>on  

Synergy  value  

1.000.000  profiled  SME  

v Cross-­‐sell,  upgrade,  &  add-­‐on  v Membership  &  point  reward  v Merchant  &  Redemp>on  v Customer  profile    v Scheduler  campaign  program  

Sales  acquisi>on  70.000  

subscriber    

Sales  leveraging  incremental  revenue  IDR  

70  Bio  

q Campaign  offering  q Lead  genera>on  based  on  profile  q Product  package  &  bundling  q Tight  monitor  RE  to  PS  process  q Repor>ng  and  dashboard  

Objectives

Server  Analy>cs  

Server  Campaign  Management  

Campaign  Management  Module  

1.  Sejng  prospect  2.  Event  Trigger  3.  Predic>ve  Analy>cs    

Channel  Management  Module  

1.  Email  blas>ng  2.  SMS  blas>ng  3.  WA  blas>ng  4.  Twiker  push  5.  Facebook  push  6.  Outbound  /  Inbound    7.  Visi>ng  /  CSR  Screen  8.  Web  In  pos>ng  9.  Push  ads  /  inser>ons  10.  Toko  online  

Database  Management  Module  

1.  Data  cleansing  2.  Data  Feeding  by  ETL  3.  Data  mart  integra>on    4.  Data  standardiza>on    5.  Data  visualiza>on    

Repor>ng  &  Dashboard  

1.  Incremental  revenue  

2.  Incremental  usage  3.  KPI  achievement  

1   2   3   4  

1.  Acquisi>on  2.  Cross-­‐sell,  Upgrade,  add-­‐on  3.  Filtering,  event  trigger  4.  Associa>on  analysis  

Data  sources  

Twiker    FB  

Data    customer  

Data    Transac>on  

Data  interac>on  channel  

Process Data Monetizing SME

Use Case Brand Tracking & Monitoring

03  

Use Case Brand Tracking & Monitoring Social Network Data about Indihome

Most Mentioned User

[PERCENTAGE]  

[PERCENTAGE]  

[PERCENTAGE]  

Sentiment

Posi>ve   Nega>ve   Netral  

Gender Top Topic

Tweets

June 2015

Brand Tracking Sentiment about Indihome

•  In general, the conversation regarding Internet Service Provider tends to be be either positive or neutral. ü  Among all providers, ISP B generates the highest positive sentiment. In contrast, Indihome generates the highest

negative sentiment proportionally, mostly regarding some complaints about Indihome services.

•  In Jabodetabek, Indihome also has the highest negative sentiment which mostly about Indihome’s service complaints.

600  58  40  

21  21  20  17  17  16  16  15  15  14  14  14  14  14  14  14  14  13  13  13  13  12  

telkomcare  telspeedyjogja  telkomspeedy  cempaka2000  

speedy_sti  dionmv  

fertida_olshop  paulawymetha  luthBi_rahmadi  

lapakolrosachan  kotabalikpapan  promoiklanindo  twitsurabaya  radiolazuarfm  

haqqi  halobali  infosmg  

uhurukalawa  halocities  chrisraph  

chusnulhida  sekitarsolo  

liferecordsmy  kim_eun_joon  halojakarta  

792  601  

361  56  50  49  34  32  31  31  24  21  20  19  17  15  15  14  14  13  13  13  12  10  8  

telkomcare  telkomspeedy  

_mBi_  telkomindonesia  

updateblog  speedyinstan  

chrisraph  wiBi_id  

spe  wizkidayo  

speedy_ydeeps  rp  

telkomsel  zuhairimisrawi  uhurukalawa  useetv_cable  onnowpurbo  useetvcom  traxmagz  infosmg  

speedy_bpn  alfamartku  yogawat  haqqi  

rooBiriyuzaki  

Top Users of Brand

•  Telkom and its subsidiaries accounts are still dominating the top mentioned and top active users list. ü This may be concluded that the conversations about Indihome are still mostly driven by Telkom’s

promotional tweets and costumer care tweets. ü Other than that, the top mentioned and top users list are dominated by individual accounts, each

with a quite low number of buzz and most of them are tweeting their complaints. ü _mfi_ is the third highest mentioned account since his complaint tweet received a high number of

retweets by bot.

Top  Acave  Users  Top  Menaoned  Users  

Telkom  and  Subsidiaries  

Ins>tu>onal  

Media  

Public  Figures  

Products  (Gadget,  etc)  

Community  

Other  Account  

Use Case Government Social Media Monitoring

04  

Government Social Media Monitoring Bandung

Tweet Map to monitor city current issue

Use Case Social Media Analytics for City of Bandung

33  

Happiness Index with Social Media

Ciazen  topic  

Hot  Topic  

Happiness  Index  

…Pelihara  Persatuan…Menangkan  Persaingan…  Jayalah  Indonesia…Jayalah  Telkom  Indonesia…  

Thank