Customer Analytics & Segmentation
3 likes694 views
This document provides an overview of customer segmentation techniques and applications for telecommunications. It defines customer segmentation as splitting a customer database into meaningful groups based on specific criteria. The goals are to gain customer insights, enable targeted marketing, and achieve competitive advantages. Various types of segmentation are described, including structural, categorical, and behavioral. Examples are given using dimensions like tenure, profitability, and risk. Effective customer metrics, technologies, infrastructure, and the segmentation lifecycle are also outlined.
Customer Analytics & Segmentation
- 1. http://www.datamine.gr Customer Analytics & Segmentation for telecommunications Techniques & Applications George Krasadakis September 2005
- 2. http://www.datamine.gr Customer Segmentation Overview & Definitions Types of Segmentation Segmentation Examples Input Description The Physical Customer Model Effective Customer Metrics Sample Segmentation schemes The time dimension Technologies & I.T. Infrastructure Segmentation Lifecycle Contents
- 3. http://www.datamine.gr Overview Customer Segmentation is the process of splitting a customer database into distinct, meaningful, homogenous groups based on a specific methodology Customer database Goals & objectives Analysis & Segmentation Statistical models, marketing expertise Profiling & interpretation The main objective of customer segmentation is to understand the customer base, and achieve sufficient customer insight that will enable the right treatment on the right set of customers at the right time…through the right channel Efficient use of customer segmentation infrastructure & techniques is expected to result in: Competitive advantages through flexible, targeted marketing actions & campaigns Customer Satisfaction & Loyalty (Churn management) Efficient Consumer Risk Management Process Automation & Optimization Effective Performance Monitoring, Executive information & Decision support
- 4. http://www.datamine.gr Definitions Customer Segmentation: the process of developing effective schemes for categorizing and organizing meaningful groups of customers Macro-Segmentation targets in schemes that are simple, easy to understand, in order to become a common corporate language regarding the customer base Micro-Segmentation defines rather complex schemes, with shorter lifecycle and large number of variables and filtering criteria, to be used by analysts or marketing experts. Supports decision making, marketing campaigns, monitoring & performance studies Customer Segmentation can be Market Driven in order to capture specific market attributes (consumer vs large accounts), or Data Driven in order to capture actual structures or patterns based on customer characteristics and behaviors Customer Profiling is the process of analyzing the elements (customers) of each segment in order to generalize, describe or name this set of customers based on common characteristics. It is the process of understanding and labeling a set of customers Business Intelligence is the set of technologies that enable companies to explore, analyze, and model large amounts of complex data. Consists of statistical modeling, data mining and multidimensional data exploration technologies - OLAP
- 5. http://www.datamine.gr Customer Segmentation: Types & Objectives The goals for segmenting the customer base can be either strategic, decision supportive in nature (executive information) or pure marketing-oriented for specific campaigns or promotional activities At a macro level, the main objective is to understand the customer base, be able to present its synthesis using meaningful groups of customers, monitor and understand change over time, to support critical strategies and functions such as CRM, Loyalty programs, product development At a micro level, to support specific campaigns, commercial policies, cross-selling & up-selling activities, analyze and manage churn & Loyalty Customer Segmentation can be further divided in the following groups: Structural: ‘natural’ segments that are very basic and result from the nature of the business. Geographical, product or commercial based segments (consumer or large accounts) Categorical: Based on ‘physical’ customer characteristics such as gender or age Behavioral: Based on indexes or scores that capture customer behavior in several dimensions
- 6. http://www.datamine.gr A simplified Example: 2 dimensions Tenure (CLS) Profitability or Revenue Highly profitable, ‘New comers’ Non profitable, ‘New comers’ Highly profitable, Loyal Customers Non or Low profitable, Loyal Customers 0 Attempt usage stimulation campaign, using further micro segmentation schemes Good Customers that must be retained: Add to Loyalty program The best set of customers. Must be treated differently through all available customer touch points (POS to CC) Poor performing customers. Must be analyzed for promising sub groups (age or demographic profile along with variances in usage) Limitations of the above oversimplified segmentation scheme No consideration of significant dimensions, such as Payment Behavior (Consumer Credit Risk) Demographic, socioeconomic or lifestyle and usage information is missing Use of scores or ranks can significantly improve the schema and its interpretation It is static, no time dimension or Transition Probabilities defined
- 7. http://www.datamine.gr A simplified Example: 2 dimensions Consumer Risk Profitability or Revenue Highly profitable, Low-Risk Customers Low profitable, Low-Risk Customers Highly profitable, High-Risk Customers Low profitable, High-Risk Customers 0 Attempt usage stimulation campaign, use further micro segmentation schemes Best Customers - must be retained: Add to Loyalty program High revenue generation but bad-payers. Must be treated accordingly e.g require credit card as payment method Poor performing, High Risk customers: analyze for understanding and modeling behaviors Limitations of the oversimplified segmentation scheme No consideration of significant dimensions, such as Tenure Demographic, socioeconomic and usage information is missing Use of scores or ranks can significantly improve the schema
- 8. http://www.datamine.gr Input Description The input should be sufficient in order to describe… Overall customer picture, based on summary figures (using weighting techniques): tenure, average revenue, aggregated AMOU, account analysis, activation requests – applications, total Revenue Ranking, Risk Assessment Utilization - how the subscriber uses each service (traffic data), indexes, correlations Spending & Payment behavior, including consumer risk assessment …enabling analysis at several levels: Physical Customer Level: demographics, socioeconomic data, aggregates & scores Account & Product Level: listing along with specific properties, Services & usage patterns, processed traffic data, Maintenance behavior & Contact History Seasonal Patterns, trends, time dimension
- 9. http://www.datamine.gr Dimensions & Filters Customer -Risk Class -Revenue Class -Socio -Economic data -Demographics -Location data (GI) -Tenure (CLS) -Traffic Patterns -Contact Patterns -Prior Classifications Product - Services -Accounts, status & types -Services & Tariffs -other properties Input Description Customer Segmentation is -by definition- multidimensional: must involve all the important aspects of each customer: risk, tenure, profitability, or Customer value must be combined in order to explain or optimize a set of metrics or specific behaviors Measures -total revenue -Balance by type (source) -frequencies -’recent’ statistics -’lifetime’ statistics -AMOU -ARPU -Specific Traffic metrics (services usage – destination analysis, incoming vs outgoing etc) -Churn Behavior -Campaign Responses -Customer Satisfaction metrics Segmentation schemes Macro segmentation for management & decision support and performance evaluation purposes Micro segmentation schemes, campaign specific, for product development, up selling or cross- selling program design, for loyalty – churn management, marketing actions
- 10. http://www.datamine.gr Composite input for advanced Segmentation Powerful Segmentation schemes can be designed based on combination of market knowledge, concepts, and extensive statistical or data mining modeling. Dimensions and measures such as: Voice usage (Frequency, duration - variance of duration) Systematic, Normal, Occasional Service Sensitive, Price Sensitive, Balanced Traffic Destination Local, long distance, international, competitors Incoming/Outgoing Traffic Balance Passive, Active, Normal VAS usage Entry Level, Experienced, Power users Traffic Density Analysis (scores of distinct IN/OUT MSISDNS) MSISDN dependency levels SMS versus Voice Balance (Incoming/Outgoing) Heavy SMS, Heavy Voice Activation history New, Returning, Recycling, Multi-Contract Contact Statistics Systematic, Normal, Occasional
- 11. http://www.datamine.gr The Physical Customer Model Physical Customer Account #1 Line #1 Line #n Invoicing, Payment Traffic patterns Traffic patterns Demographics, customer history, ratings, memberships Account #2 Line #1 Line #n Invoicing, Payment Traffic Patterns Traffic Patterns Scoring Engine Score, statistics, weight factors Score, statistics, weight factors Weight Models Overall scores PartialScore PartialScore Physical Customer Identification is a critical point in customer segmentation & insight: A physical customer may have several accounts with contradictive behavior regarding usage or payment. The physical customer (a) must be correctly identified and (b) must be scored in the top level in an efficient way
- 12. http://www.datamine.gr The need for an Objective Customer Assessment Physical Customer Account #1 Line #1 Account #2 Line #1 A Physical customer can have several accounts of contradictive behavior. The concept of Primary Account and suitable weighting mechanisms can efficiently address this complexity through an objective scoring at the top level A very good account: Tenure rank: top 10% Revenue rank: top 20% Credit Risk: bellow 10% Could participate in a Loyalty program A bad account: No traffic Bad payment behavior with frequent payment delays (suspensions, reactivations) Could be in a collection state Confusing, negative outcome for the Customer
- 13. http://www.datamine.gr Efficient Customer Metrics Billing & Payment Statistics Total amount Billed, Open Balance Analysis Billing Statistics (Averages,Variability) Payment-related Statistics (Delays, Suspensions, Fraud History), Credit Score (payment behavior) Profitability or Revenue Rank Score Account analysis (by status), Product & Services Traffic analysis Outgoing Calls / Duration versus SMS Incoming Calls / Duration versus SMS Most Frequent Destination Number (MFN) Operator Significance Indexes Distinct Number of IN/OUT MSISDNS Call Duration distribution Time of Day distribution Day of the week distribution Variability & Trend of average Call duration Operator (Destination) distribution (IN & OUT traffic) Cell distribution (GIS) Distinct Number of Cells used (Mobility) Data Calls frequency - duration Special Services - frequency – duration Customer Care Calls, Frequencies & Summaries by Service, reason Metadata Statistically derived Scores, clusters and existing segmentation schemes Marketing Research data, customer satisfaction surveys, on-line customer surveys, customer interaction data (CRM campaigns, Loyalty program memberships & usage, special offers) Micro-Macro segmentation, clustering memberships, control-placebo group memberships
- 14. http://www.datamine.gr Customer metrics versus time Modeling Customer Metrics on time scale is a challenging task due to: Complex, Seasonality patterns, segment depended, cycling behaviors, different life cycles for each segment Market Trends, competition & significant changes (e.g. number portability) Complex environment (services, tariffs, multiple accounts for each customer, contract versus prepaid markets) datamine’s approach in modeling change is based on capturing the complete picture of each customer at certain (predefined) moments of its lifecycle along with detailed history per customer: Picture of the customer on 6th and 10th month of it’s life (key metrics on traffic, averages on billing and payment, risk scores, rank) in order to capture key metrics in a mature state and also prior the critical first contract expiration. Running averages, comparable with the above, yearly averages along with variance and variation coefficients Trend measures, and seasonal components on frequent time intervals datamine’s approach is based on a flexible infrastructure that maintains sufficient historical information using intelligent techniques (scores, aggregates and/or random sampling on the actual transactional data) thus providing the capability of reproducing the state of the customer base and each single customer for any given time point, resulting in powerful reporting capabilities and customer base monitoring / comparative functionality.
- 15. http://www.datamine.gr Customer metrics versus time 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 CLS (months) (%)ofpopulation REMAINING(%) VOL(%) NOVOL(%) Measure Key Customer metrics on 5th to 6th month Measure Key Customer metrics on 10th month, apply segmentation and contact valuable customers for upgrade Study the synthesis of the remaining customers and compare with the initial population
- 16. http://www.datamine.gr Customer metrics versus time Allows dynamic report generation of the style: Select the top 70% of the customers (in terms of revenue) that have exactly one active account, running the 8th to 10th month, having credit risk below 20%, and outgoing traffic more than 80% to competitors …….. and built a campaign targeting in both customer satisfaction and word-of-mouth effects Or Select top 30% of the customers with more than one active account, with less than 40% credit risk, that have reduced their traffic or revenue more than 40% in the last x months….. and try specific usage stimulation campaigns or perform random sampling to identify the satisfaction levels Similarly Select top 30% of the customers with more than 30% of their outgoing traffic to prepaid, with less than 40% credit risk, that have used MMS service more than xx times in the last x months….. and try the effect of offering free web access or other hi-tech services in competitive pricing
- 17. http://www.datamine.gr Data Modeling technologies Descriptive Statistics (exploratory data analysis): cross tabulation analysis, using combination of filtering criteria, OLAP tools, advanced visualization – graphics techniques Statistical Modeling: univariate & multivariate statistical techniques, cluster analysis, scoring models, combination of statistical techniques Data Mining techniques: specialized algorithms such as Decision trees or Neural Networks I.T. infrastructure A ‘mature’ Data Warehouse, providing reliable, ‘clean’ customer information, from the top level (the physical customer) to Call Detail (CDR) and Contact History level Statistical and/or Data Mining Systems, any of commercial product such as SPSS Clementine, SAS Enterprise Miner or Microsoft SQL Server 2005 Business Intelligence Studio Specialized OLAP - like systems with sufficient list management functionality and segmentation deployment procedures Technologies & I.T. Infrastructure
- 18. http://www.datamine.gr I.T. Infrastructure Flattened customer data structures Reliable customer data with time dimension Physical Customer, Account & Contact, Customer Scores Billing data Payment behavior Segmentation data Utilization profile & Aggregate Traffic patterns Statistical Modeling Billing & Provisioning Systems Customer Profiling Account data Services & tariffs Billing & payment history Customer Care, Operational CRM Contact History, Complaints, Activation Requests REPORTING datamart CRM datamart Reporting Tools OLAP Customer Base KPIs monitoring Customer Segmentation System Customer Viewer Traffic Data CDR raw data, QoS data TRAFFIC processes Operational CRM Platform Marketing Data Products & services properties, Campaigns, Micro& Macro segmentation schemes ETL processes Data cleansing, Transformation to ‘flat’ data structures Descriptive statistics, traffic patterns Statistical models, churn prediction, credit scoring, fraud cases, segment-cluster-campaign memberships MARKETING DATABASE Sales Automations DATA PROVIDERS DATA WAREHOUSE - ANALYTICS DSS AREA - DATA CONSUMERS
- 19. http://www.datamine.gr Segmentation lifecycle Goals & Objectives Customer Segmentation Profiling & Interpretation Business Applications Define Business Objectives: segmentation can be sales driven, product driven, profitability or service positioning driven • Set the basis of the analysis (time frame, subset of customers) • Built the working team Review data requirements & examine availability • collect, analyze data & assess data quality • perform preliminary data analysis cleanse data • Select segmentation techniques (predefined or statistical) • Begin Segmentation • Analyze data • build statistical models • (re) design customer metrics • perform segmentation Interpret segments • understand the typical customer within each segment • analyze performance indicators for each segment examine segment behavior versus time (customer base synthesis) Apply the derived segmentation schemes to support specific business needs • Monitor the customer base evolution in terms of segments • measure segment transition probabilities • monitor the homogeneity of each segment Close the Loop: collect response and performance information • assess segmentation synthesis - profiling Performance Assessment
- 20. http://www.datamine.gr 22 Ethnikis Antistasis Avenue, 15232 Chalandri, Athens, Greece Tel (+30) 210.68.99.960 Fax (+30) 210.68.99.968 g.krasadakis@datamine.gr http://www.datamine.gr George Krasadakis Customer Analytics Manager