Data Mining & Campaign Management The Maxis Experience

Size: px
Start display at page:

Download "Data Mining & Campaign Management The Maxis Experience"

Transcription

1 Data Mining & Campaign Management The Maxis Experience SAS Forum International May 2006 Geneva, Switzerland Evelyn Jimenez Head of Customer Lifecycle Management Maxis Communications Berhad, Malaysia

2 Presentation Agenda Maxis and the Malaysian mobile industry Customer Lifecycle Management Sample Campaigns Best Practices 1

3 Malaysia at a glance Located in Southeastern Asia, Malaysia is basically divided into 2 regions: Peninsular Malaysia and East Malaysia Land area: 328,550 sq km Population: 26.7 million (2005 est.) Median age: 24 years Multi-cultural society: Malays, Chinese and Indians GDP (purchasing power parity): $249 billion (2005 est.) Internet users: 10 million (2005) Mobile users: 19 million (2005) 2

4 Maxis is the leading mobile service provider in Malaysia with a cumulative base of 8M subscribers Penetration Rate (%) 100% 75% 50% 25% Malaysia Penetration Rates Prepaid launch spawned mobile growth Lowering of prepaid starter pack prices 44% 31% 37% 56% 73% Malaysian mobile penetration rate reached 73% at the end of 2005 with 19.5 million subscribers 85% are predominantly prepaid subscribers 0% Y2005 Market Share DiGi, 25% Maxis, 40% Y2005 Revenue Share DiGi, 21% Maxis, 46% Maxis is the market leader with 40% market share or 8 million subscriber base Maxis also leads in financial performance with share of Revenue and EBITDA at 46% and 49% respectively Maxis ARPU registered at 9% above industry average Celcom, 35% Celcom, 33% 3

5 Maxis on its way to becoming the preferred mobile service provider in the region Awards won in 2005: International ventures: ASIAN MOBILENEWS AWARDS 2005 Mobile Operator of the Year, Malaysia Award India Aircel (Dec 05) FROST AND SULLIVAN Malaysia Telecom Awards Service Provider of the Year - Mobile Service Provider of the Year MALAYSIA BRAND EQUITY AWARD st Brand Visibility Award BEST MOBILE BROADBAND SERVICE PROVIDER PC.com Best Product Award 2005 BEST POSTPAID TELCO SERVICE PROVIDER PC.com Best Product Award 2005 SUPER BRANDS 2005 Mobile Service Provider - Platinum Telecom Company Gold Aircel: Number 1 operator in Tamil Nadu and Chennai India s population of 1.1billion currently at a low penetration of 7.4% and at the inflexion point of growth Indonesia NTS (Apr 05) NTS: Scheduled nationwide launch for GSM1800 and 3G in 2006 Market with high potential for growth 4

6 While acquisition is still needed, customer retention becomes more critical in a maturing market Environment Challenges Market reaching maturity with 73% penetration rate Intense competition with much lower cost of entry: cheaper starter packs and further reduction in tariffs Mobile service now a commodity, no longer a luxury Subscriber stickiness is a critical business driver Growing value-added services and mobile data business Campaign Challenges Aggressive acquisitions expand the market, but it also creates dual-sim phenomenon and phantom churn Targeted CLM campaigns are more efficient and sustainable on a long-term basis Millions of daily CDRs processed to timely, meaningful and actionable information Speed to market Quantity and quality of campaign executions Campaign effectiveness monitoring 5

7 Presentation Agenda Maxis and the Malaysian mobile industry Customer Lifecycle Management Sample Campaigns Best Practices 6

8 Focusing on churn reduction and revenue enhancement within the customer s lifecycle Acquisition Customer development Harvest Win Back Customer Value Target/ acquire subscriber Welcome Program Revenue Enhancement Migration Pro-activity based on If events: - Lifetime - Usage/ purchase - Behaviour Churn Prevention/ Reduction Expiration Time / insights Analytical insights Behaviour Scoring Response rates Entry Scoring Contact Policy Fraud Detection Segmentation (Value/Needs) Tariff Plan Optimisation Cross Sell / Up Sell Credit / Collections Churn Propensity Churn Segmentation Satisfaction score 7

9 Data mining and below-the-line channels are our enablers toward an effective lifecycle management From. To. Standard Data Reports Data Mining Segmentation Predictive Modeling Customer Profitability Traditional Marketing Target Marketing Churn management Usage stimulation Plan migration Above-the-line Advertising Below-the-line channels SMS Outbound calls Direct mailers 8

10 The Maxis Customer Lifecycle Management Process Multiple segments + multiple offers Define segments based on customer usage and recharge behavior Define multiple offer types and variants Use control groups to measure impact Test offers + measure impact (ROI, saves) Test different offers against different segments Calculate net save rates (target group vs. control group) Measure ROI of each offer-target combination Scale successful campaigns Identify offers which resulted in the highest save rate and scale up to the whole segment (BTL) Segment Offer Target Expected saves* Deactive segment ARPU >RM45 Group A Time to churn <35 days Group B Time to churn Group C Time to churn K subs 30K subs days days 60K subs 60K subs + Control groups 60K subs Offer variants RM20 RM60 100% bonus for bonus for Bonus if RM10 top- RM30 top- next top-up up up in in 7 days 3 times 7 days + no offer Segment A B C Offers Uplift*: 25% Uplift: 67% Cost**: RM12 Cost: RM30 ROI: 700% ROI: 1,000% Uplift*: 6% Uplift: 90% Cost**: RM12 Cost: RM30 ROI: -66% ROI: 2400% Uplift*: 8% Uplift: 125% Cost**: RM11 Cost: RM31 ROI: -66% ROI: 3000% Uplift: 25% Cost**: RM7 ROI: 1,300% Uplift: 2% Cost: RM6 ROI: -130% Uplift: 0% Cost: RM6 ROI: -150% A No K/month 2,500 B C No.102 No K/month 210K/month 4,000 3,500 Total 600K/month 10,000 Today: >10 offer variants tested each week across different segments have built the ability to execute 3-4 campaigns/week 1 scaled campaign ( K) 2-3 tests / refinements ( K) Todate: 25 campaigns tested 9

11 Presentation Agenda Maxis and the Malaysian mobile industry Customer Lifecycle Management Sample Campaigns Best Practices 10

12 Subscribers get hit several times with CLM offers while in the churn pipeline Reactivate Broad Launch to Deactive Segment Segment Offer Control Take Net Take Net Save SMS Offer 1 SMS Offer 2 SMS Offer 3 SMS Offer 4 Presence of a sweet spot for saves within the lifecycle Certain value segments have higher save rates SMS Offer SMS Offer 6 11

13 Revenue enhancement from upfront revenues and resulting increase in usage TEST BROAD LAUNCH Offers Tested SMS Offer 1 SMS Offer 2 SMS Offer 3 Target: SMS usage-tiered segments Net Take rate Negligible take rate from very low SMS users Reprice offset by increase in usage of about 30-50% Offer SMS Offer 1 Expected incremental usage (based on CLM Test) Expected total revenue 30-50% uplift Upfront revenue + incremental usage SMS Offer 4 SMS Offer 5 SMS Offer 6 12

14 Predictive churn model looks at past 3 months behavior and predicts behavior for the next month Looking back at behavior before actual churn History History History History Churn Last 3 Month Last 2 Month Last 1 Month This Month Next Month Step 1 Step 2 Step 3 Step 4 Step 5 Build Model based on random sample behavior Determine significant variables from variables list Generate highpropensity churner segments Apply Model to score whole postpaid base on monthly basis Perform Save Offer on monthly basis 13

15 There are 5 significant variables that predict postpaid attrition with churn probability score as high as ~ 90% Churner 90.5% Non Churners 91% Churner 71.6% Churners Model lift value = 7.3 Churner 60.4% Churner 55.1% 1% Randomly Selected Non Churners Top 1% Model Predicted Churners Non Churners High Score 9% Churner 51.2% Churners Base= 9,629 Churners = 0.3% Churners Base= 9,629 Churners = 2.2% Churn Attributes Used: 358 Original and Transformed Variables 5 Significant Variables The Major Profile: Significant Reduction in bill size Significant reduction in voice calls amount SMS usage level Overall incoming call level Less outgoing calls vs incoming calls The Model produced 13 groups of subscribers with different churn probability and behavior It can be broadly classified into High Score (9%) and Low Score (91%) The high score segment consists of 5 distinct groups of behavior The Model identifies 7 times more churners! 14

16 The predictive churn campaign yielded not only a net save of 13% but also a net increased usage incidence of 27% Task Identifying potential churners Criteria High & medium value customers High propensity to churn Churn Model Segments Test Offers Gross Takes Outbound Calls to Target List, whilst control group is isolated from campaigns Offer 1 Offer 2 Offer 3 Top 7 % take up 60 % take up 20 % take up Control, no offer Post Campaign analysis Status 1 month after launch Active : 91% Suspend : 1% Active : 75% Suspend : 5% Deactive : 7% Deactive : 20% Saved potential churners Analysis of usage of customers in the contact list & control group showed improvements Usage Band % Increased Usage 36% Maintain 16% Reduced Usage 48% Net save 13% Usage Band % Increased Usage 9% Maintain 36% Reduced Usage 55% 15

17 Presentation Agenda Maxis and the Malaysian mobile industry Customer Lifecycle Management Sample Campaigns Best Practices 16

18 Based on our experience, it takes several components to make the CLM operations successful Focus first on critical business issues Dedicated CLM team Skills development, training & consultancy Top Management support Ownership by the business, not IT Analytical minds & lots of creative brainstorming Unified CLM datamart Continuous campaign monitoring and transfer of learnings Effective execution via tools, processes & quality controls 17

19 Full campaign automation and real-time, event-based marketing are our key next steps Standard analytics Data mining: Predictors, Customer Lifetime Value Churn Forecasting ROI tools & CLM processes Market testing of new products & services Churn management, usage stimulation, migration Full campaign management automation Profile indicators & personalized offers in all touch points Real-time, event-based marketing 18

20 Thank You Football. Gets the world talking. Log on to

21 20

22 Thank You Football. Gets the world talking. Log on to