Large Scale Multi-Country Segmentation and

Large Scale Multi-Country Segmentation and

• 2.
  Large Scale Multi-CountrySegmentation and Targeting for a Leading Mobile Phone Operator Using Latent Class Modeling on Both Behavioral and Attitudinal Data Marc O’Regan Ashok Kalidas Director Vice President Behavioral Insight MaPS Kantar Oleg Reshetin Olga Maksimova Project Director Consumer and Market MTS Insight Director MTS
• 3.
  Introducing MTS The LeadingTelecom Group in Russia, Eastern Europe and Central Asia Total Subscribers* (millions) • Operating mobile- and fixed-access networks in 105.20 six countries in Eastern Europe and Central Asia 103.40 – Russia, Ukraine, Belarus, Uzbekistan, Turkmenistan 101.38 102.37 102.38 and Armenia – Population coverage area of over 230 million • MTS is one of Russia’s top corporations Q3 2009 Q4 2009 Q1 2010 Q2 2010 Q3 2010 – Rated one of the most transparent companies in Q3 Q4 Q1 Q2 Q3 Russia by Standard & Poor’s Subscribers, Russia 2009 2009 2010 2010 2010 • MTS one of the BRANDZ™ Top 100 Most Mobile (millions) 68.70 69.34 69.08 69.42 69.67 Powerful Brands Households passed, 000s 7 485 7 502 7 756 7 942 7 799 – MTS included in the ranking in 2008, 2009 and 2010 Broadband Internet, 000s 1 227 1 298 1 359 1 437 1 466 – 72nd brand overall with a brand value of $9.7 bln Pay TV, 000s 2 085 2 124 2 122 2 176 2 028 • MTS’ Level III ADRs have been publicly traded * Including Mobile TeleSystems LLC, a mobile operator in Belarus, on the New York Stock Exchange since 2000 in which MTS owns a 49% stake • MTS is majority-owned by AFK Sistema (LSE:SSA), the largest diversified public financial corporation in Russia and the CIS
• 4.
  3i: MTS Strategy StrategicDirection Tactics Key Benefits Integration • Seamless user experience for all segments New pipelines • Rapid broadband infrastructure (fixed/3G/LTE) deployment and customer touch-points • Integrated sales channels Increasing customer lifetime Internet value • Enhanced connectivity Smarter pipelines • Compelling Internet user experience to capture additional value • Best-in-class content apps and services Generating shareholder Innovation returns • Delivery of exclusive devices Differentiation • Cutting-edge products and services for all customer segments through product and service mix • End-to-end user experience at home, at work and on the move
• 5.
  MTS MacroSegmentation MTS engagedKantar to conduct a macrosegmentation across its footprint, with the following guiding principles • Strategic, yet Actionable • Common regional platform, yet local solutions • Attributable to MTS CRM database, replicable in future market research • Business and consumer; customers and prospects • Foundation for strategic planning, performance management and operations
• 6.
  The Process: ACollaboration Between Many Kantar Operating Companies The vast scale and scope made the logistics enormously complicated Design Scoring Algorithm, Segment Synthesis, Feasibility & Field Other Development Reporting Setup Implementation • Over 10K surveys, 80M+ Exploratory analysis of customer Customer data database Analytic technique (domain, granularity, quality, volum • Kantar team spread across designed to works with e) Boston, London, Kyiv, Moscow, “systematic blocks of Chennai missing data” Reach agreement on what is/is not possible with database • 10 computers building models Variety of analyses to 24*7 for 30 days assess solution Sampling plan and survey content stability, replicability, and influenced by findings of • Over 100 potential solutions examined scoring accuracy exploratory analysis
• 7.
  Keys to Success 1.We balanced marketing richness (do the segments differ on values, needs, wants, attitudes) and behavioral/demographic targetability (can the segments be identified based on the customer data that are available) Hypothetical Case: Perfect Demographic/Behavioral High and Needs & Attitudes Alignment Approach A Balanced Demographic/ Balanced Behavioral Segmentation Targetability Segmentation Approach B Needs and Attitudes Segmentation Undesirable Low Low Marketing Relevance High
• 8.
  Keys to Success 2. We identified “real” differences • Segmentation unearthed a set of between countries, but maintained “common” segments across all countries a common framework and platform but with very different sizes across the entire footprint • Methodology was sensitive to survey biases and database inconsistencies across countries Needs by Country Segment Distribution 84% 78% 90% 88% 89% Overall 25% 20% 14% 14% 5% 22% 75% 73% 70% 67% 71% 65% 63% 70% Mainly in 72% 73% 60% 63% 57% 67% Russia 30% 20% 15% 10%1% 24% 62% 56% 43% 63% 63% Mature 47% 40% 56% 56% 55% Ukraine 25% 18% 13% 14% 2% 28% Markets 33% 30% 39% 30% Armenia 35% 34% Russia Belarus 20% 30% 13% 14% 11% 12% Ukraine Wide Belarus coverage Relevant plans and Easy to Uzbekistan Uzbekistan 9% 22% 13% 23% 31% 2% Latest area services use products products Cheap prices Clear and Excellent Mainly in Nascent and transpare and services services nt customer Armenia 26% 28% 17% 9% 19% charges service Markets Usage of rating scales clearly varies; analytic Common to method specifically designed to capture this All Markets Note: Data disguised to protect client confidentiality
• 9.
  Keys to Success 3. We provided the ability to score the customer database at a high accuracy (70%+) as well as the ability to identify segments using a short set of survey questions Attitudes, Perceptions, Self-reported Usage, Spend, Other Behaviors, Demographics Behaviors, Demographics WAREHOUSE Both Customers Customers SURVEY DATA and Prospects Only UNIFIED MARKET SEGMENTATION ? QUESTIONS ALGORITHM PREDICTION “GOLDEN” SEGMENT Even though the 2 classification algorithms use completely different sets of A short set of questions that data, we achieved greater immediately indicates the Attributes a segment to each customer in respondent’s segment, allowing than 70% accuracy in both the CRM database enabling post-hoc replication in other research studies analysis and proactive marketing actions
• 10.
  Overview of SegmentationMethodology Segmentation analysis process Hypothesize Conduct Strategic Evaluate Segmentation Segmentation Drivers Segmentation Solution Analysis • Mix of behavioral and • Marketing evaluation survey based – Clear and distinct picture characteristics. e.g. A B of segment members – Usage/billing data C – Actionable and well- differentiated on targetable – Functional and D dimensions emotional needs – Enables prioritization based – General category  Use Latent Class Analysis on segment value attitudes – Demographics • Statistical evaluation – Stability and replicability – Classification accuracy – Homogeneity within segments and heterogeneity across segments
• 11.
  Analytic Innovation Segment an incomplete database constructed from multiple data sources in one step – rather than developing a solution based on just one set of data and retrofitting it to another Nature of Information Available Survey Data Customer Database MTS Customers (~5K) Actual Behaviors, (surveyed customers) Needs, Attitudes, Demographics Self-Reported Behaviors, Comp. Customers (~5K) (surveyed consumers) Demographics X MTS Customers (~80M) (from customer database) X Actual Behaviors, Demographics We deployed an innovative solution based on a sophisticated latent class model explicitly designed to work with “missing” blocks of data
• 12.
  Implementation • Segmentation widelyused in the organization – New Tariff plans launched in Ukraine and Russia based on needs of one key segment “Thrifty Value Seekers” – KPIs tracked by segment (Size, ARPU, MoU) – Advertising – High level budgeting – Unified marketing planning within the Group • Following the recent acquisition of Comstar, a fixed line telecom company, MTS plans to enhance the segmentation beyond mobile
• 13.
  @ 5ADVERTISING 7. RESEARCH "".,,, , FOU N DA T I O N