Customer segmentation
54 likes13,837 views
The document discusses customer segmentation methods. It notes that advanced segmentation methods like value-based segmentation using customer lifetime value (CLV) can help companies better understand customers and ensure long-term profitability compared to basic methods. A multi-layer model combining advanced methods provides the most effective segments to connect with customers through acquisition, loyalty programs, and other targeted actions. The key is to define clear objectives and continuously improve the segmentation model.
Customer segmentation
- 1. Customer Segmentation May 2011 White Paper «Which segmentation to connect with your customers? » MANAGEMENT CONSULTING Your CONTACTS Alexandre GANGJI, Partner Benelux [email protected] +32 (0)477 597 398 Hub weave : www.weave.eu 1
- 2. Executive summary As market dynamics are changing, companies are looking for new segmentation to improve the way they attract, manage and retain their customers. This re-segmentation exercises can be used to support strategic objectives, such as identifying new market potential, or tactically, to improve retention, ROMI and cross- selling. However, moving from traditional segmentation methods to more advanced methods require marketing organizations to invest beyond their basic capabilities Within the advanced methods, the value-based segmentation allows for prioritization of the customers portfolio according to their economic value (past or future). One of the most effective value-based approaches is Customer Lifetime Value (CLV). While CLV can be seen as complex, its pragmatic application yields most benefits and the use of customer lifetime value as a marketing metric tends to place greater emphasis on customer service and long-term customer satisfaction, rather than on maximizing short-term sales A multi-layer segmentation model leverages the particular benefits of each segmentation method by crossing basic segments and value-based approaches. This results in operational segments allowing more efficient targeting actions (acquisition, costs reductions, loyalty, etc.) and connecting with your customers Segmentation projects are typically approached in seven steps, from defining objectives of the segmentation exercise to assessing its effectiveness. Key success factors are to implement a standard data collection process, to respect segment definition rules and to ensure continuous improvement of the segmentation model 2
- 3. Effective segmentations allow companies to allocate scarce resources where they'll deliver the highest payoff Strategic Tactical • Opportunity identification • Prospecting / lead generation • Prioritization • Sales force allocation • Market investments / divestments • Channel strategy Domains of application • Positioning • Communications programming • Product / portfolio development • Pricing • Market driving (vs. customer focused) • Compensation • New identified market potential • ROMI Potential impacts • Share of new products in total portfolio • Conversion rate • Margin by segment • Cross-selling and up-selling 3
- 4. Advanced segmentation methods bring customer strategy and Return On marketing Investments to the next level Level of adequacy with customer Long-term needs and benefit Methods Description behaviors potential Volumetric Quantitative analysis of historical prescription or purchased Low Low volumes Geographic Separates customers into different geographical units, such Low Low as countries, states, regions, cities Socio-demographic Separates customers on the basis of age, gender, social Low Low class, and other factors Needs-based Divides customers according to needs which are being Medium Medium fulfilled by the products or services Behavioral Based on identifying customer behavior characteristics that Medium Medium help to understand why customers behave they way they do Value-based Based on the present and future profitability of a customer High Medium (for instance CLV) Multilayer approach Approach crossing different segmentation methods and High High dimensions (in particular a value-based segmentation with other effective segmentation methods) 4
- 5. Advanced segmentation methods allow companies to connect with their customers and ensure long term profitability Basic Advanced Long term customer Segmentation Segmentation profitability 1 2 Level of marketing intelligence 1 Although basic segmentation methods can be useful for companies with basic marketing capabilities, they have proven limited sales and marketing efficiency Behavioral and value-based methods focus on understanding customer decisions, behaviors, and financial 2 value. They have proven by many companies as very prevalent and effective segmentation methods 5
- 6. Executive summary As market dynamics are changing, companies are looking for new segmentation to improve the way they attract, manage and retain their customers. This re-segmentation exercises can be used to support strategic objectives, such as identifying new market potential, or tactically, to improve retention, ROMI and cross- selling. However, moving from traditional segmentation methods to more advanced methods require marketing organizations to invest beyond their basic capabilities Within the advanced methods, the value-based segmentation allows for prioritization of the customers portfolio according to their economic value (past or future). One of the most effective value-based approaches is Customer Lifetime Value (CLV). While CLV can be seen as complex, its pragmatic application yields most benefits and the use of customer lifetime value as a marketing metric tends to place greater emphasis on customer service and long-term customer satisfaction, rather than on maximizing short-term sales A multi-layer segmentation model leverages the particular benefits of each segmentation method by crossing basic segments and value-based approaches. This results in operational segments allowing more efficient targeting actions (acquisition, costs reductions, loyalty, etc.) and connecting with your customers Segmentation projects are typically approached in seven steps, from defining objectives of the segmentation exercise to assessing its effectiveness. Key success factors are to implement a standard data collection process, to respect segment definition rules and to ensure continuous improvement of the segmentation model 6
- 7. Deep dive on value-based segmentation Characteristics Benefits • Most readily available metric Low Revenue-based • But often poorly correlated to real profitability • Includes both actual and potential customer revenue Market • Can be estimated using overall spend in the category and / or potential opportunity / growth in revenue over time revenue potential • To be used when share growth is a leading objective • Accounts for the cost to serve the customer (allocates costs such as S&M acquisition costs, service / support, R&D, etc.) Current / past customer • Very useful in industries where the cost to serve varies significantly by customer profitability • Mostly feasible for companies with a small number of customers or with very advanced customer management systems Customer • Recognizes the customer as a corporate asset Lifetime Value • Encompasses the holistic value a customer provides (influence on others) Customer Influence • Useful in industries where a few customers have a disproportionate share of influence on others’ buying decisions (ex: Pharmaceutical companies) High Source: « Marketers Must Understand Customer Value to Make Segmentation Pay », Scott Wilkerson, Sept 2009, 7 www.managesmarter.com
- 8. Introducing Customer Lifetime Value, a powerful value- based approach Definition Areas of impact • The Customer Lifetime Value (CLV) is the discounted value of the future profits that will be generated by an • There is significant added value of a CLV-based individual customer segmentation, at various levels • As these future profits are uncertain, predictive – At campaign level: CLV can help you models have to be developed. These models are capture cross-selling and cannibalisation based on data and use analytics techniques effects • Traditionally, companies / marketing managers focus – At client level: CLV can rank our on analysing historical data (past sales, campaign segments in terms of profitability returns, etc), although very significant gains could be – More generally, a CLV approach can help produced through future customer behaviour you optimise marketing campaigns forecasting • To do so, predictive models have to be • CLV can also increase predictability in terms of value developed and ranking, and be used aside traditional metrics to • E.g., a customer that will generate €120, €80, €30, enhance customer insight €50 and €10 of profits in the next five years, will have a CLV of €238,11 if the discount factor is 10%. This value is unfortunately unobservable for now, and a predictive model needs to be developed 8
- 9. Case Study Areas impacted by a CLV analysis Example Impacts 1 • From a traditional product-centric point-of- view: The return of the campaign is computed taking only into account the return generated by the Car Insurance CLV captures cross selling and Added value at product cannibalization effects campaign level • From a CLV point of view: The return of the campaign takes into account the effect on the current accounts and the savings accounts Segment A Segment B Segment C Car 600€ for 2 600€ for 2 300€ for 1 years Insurance years years 10.000€ for 2 Current years 10.000€ for 2 Account years (cross-selling) Savings -10.000€ Account (cannibalism) 9
- 10. Case Study Areas impacted by a CLV analysis Example Impacts 2 • From a product-centric point of view: Segment A = Segment B = Segment C • From a CLV point of view: Segment A < CLV can rank segments in terms of Added value a Segment B < Segment C profitability client level Segment A Segment B Segment C Car 600€ for 2 600€ for 2 300€ for 1 years Insurance years years 10.000€ for 2 Current years 10.000€ for 2 Account years (cross-selling) Savings -10.000€ Account (cannibalism) 10
- 11. Case Study Areas impacted by a CLV analysis Example Impacts 3 • The CLV change allow to optimize the company’s profits by targeting the segment A for campaign 2 and the segments B & C Added value for for campaign 1 in priority CLV allows to optimize marketing marketing campaigns campaigns • Note that with our current tools, we wouldn’t be able to determine which campaign is most appropriate for the segments B and C Segment A Segment B Segment C Consider that campaign 1 and 2 occur more or less at the same Campaign 1 0€ 250 € 200 € time. We don’t want to contact the same customers twice. Which segment should be targeted for which campaign? Campaign 2 25 € 50 € 55 € 11
- 12. The CLV allows to make more profitable decisions than Case Study the propensity-to-buy ILLUSTRATION CLV vs. Propensity to Buy In this real case example, we compare, for 500 customers, the scoring of the propensity to buy model (who will buy the product?), …with the ranking of the CLV change (who will be profitable if I contact him/her?) 12
- 13. Case Study Comparing historic profitability against CLV ILLUSTRATION 1 Past information available 2 ‘Future’ information available for 3 Analyze correlation for the project the project assessment between predicted (1st quarter 2009) (2nd quarter 2009) and actual data Past Past Customer Past Customer Customer profits Profits Quarterly CPM Profits • Value: 47% 200901 • Value in Euro • Ranking: 19% • Ranking Actual (based on the agreed scope) Better Targeting • Customer value in Euro • Ranking of the customer CLV CLV Past activity, • Value in Euro • Value: 95% CLV customer • Ranking • Ranking: 81% characteristics, etc. CLV can reliably rank customer segments in terms of profitability 13
- 14. Case Study Pragmatic application of CLV yields most benefits • Focus on profitable products only and strategic areas Product scope • Focus on multiple iteration campaigns only (in order to identify and isolate the effect) Campaign scope CLV analysis • Define a realistic horizon from 3 to 5 years with focus on Product scope relative/marginal profitability and ranking • Exclude fix costs Profitability scope • Focus on profitability per segment (vs. deep dive per client) • Keep model simple (per month, per quarter vs. per day) Modeling 14
- 15. Case Study Customer Lifetime Value modelling Pi , j ,t CLVi t 0, j 1 h, J (1 d )t Where h is the horizon of the prediction: how far we want to go in the future J is the number of products/business lines considered Pi,j,t is the profit generated by the customer i at time t because of the usage of the product j d is the discount rate Usually h is taken via a business rule J is a tradeoff between implementability and realism Pi,j,t is predicted using statistical models d is selected in agreement between the management, finance and the accounting department 15
- 16. How to model the future profitability and activity of the Case Study customers? Approach Pro’s and con’s • Create “cells” or groups of customers • Pro: very simple and flexible. Good for long based on the recency, the frequency and term predictions RFM models the monetary value of their prior • Con: many segments needed if used for purchases, the model is then estimated individual customer valuation using decision trees or Markov chains • Assume an underlying stochastic model • Pro: statistically elegant, extensively discussed Probability (e.g. the Pareto/NBD model) in the academic literature models • Con: PhD needed… • Hazard functions • Pro: very flexible and extensively used in the industry (but not for CLV modeling) Econometric • Survival analysis models • Con: work only for contractual setting (when the end of the contract is observed) • Vector Autoregressive (VAR) model • Pro: very flexible and easy, powerful for short Persistence term predictions, can take into account many models types of drivers • Con: computationally expensive Source: topology described in Gupta and colleagues 2006 in the special issue on CLV of the Journal of Service 16 Research
- 17. Markov Chains approach: an example in the retail Case Study banking industry ILLUSTRATION Year 2009 Potential Sleeping High Active Mature Potential client client potential client client churner Lost Potential 90% 0% 5% 3% 2% 0% 0% client Sleeping 0% 90% 4% 1% 0% 0% 5% client High- 0% 5% 60% 15% 5% 5% 10% potential Year 2008 Active 0% 10% 3% 70% 12% 3% 2% client Mature 0% 8% 1% 5% 70% 11% 5% client Potential 0% 15% 0% 7% 8% 30% 40% churner Lost 20% 0% 0% 0% 0% 0% 80% 17
- 18. Vector autoregressive models: an example in the retail Case Study banking industry The model for the customer activity is Yi ,t f (Yi ,t 1 , X i ,U i ,t ), with: Yi ,1,t • Yi,t the vector of the activity of customer i in the product categories at future time t, – is a function (regression) of Yi ,t ... , Y • Yi,t-1 the matrix of the activity of customer i in the product categories at time before t, i , J ,t – The activity at time t is a function of the activity at time t-1, t-2, …., t-T. – Used for modeling: loyalty, attrition. Yi ,1,t T ... Yi ,1,t 1 – The activity at time t in the product category j is a function of the activity at time t- Yi ,t 1 ... ... ... . 1, t-2, …., t-T in the OTHER product category 1,…,J. Y – Used for modeling: cross-selling, halo effect, cannibalism. i , J ,t T ... Yi , J ,t 1 • Xi a vector of characteristic of customer i, – The activity at time t is a function of the age,…,etc. – Used for modeling: customer heterogeneity, customer segmentation. • Ui,t a vector of actionable drivers (marketing actions), – The activity at time t is a function of the marketing campaign m implemented at time t-1,…t-T. – Used for: marketing campaign optimization, target identification, etc. 18
- 19. Case Study Application to a retail banker settings • The Customer Lifetime Value is the discounted value of the future profits that will be generated by an individual customer • The CLV of a customer i is a function of the profit, Pi,j,t, he/she will generate in the future t via the product j J T Pi , j ,t CLVi . j 1 t 1 (1 r ) t • As Pi,j,t is unknown, a prediction model needs to be build: – The future profits will be derived from the predicted future activity Yi,j,t, as Pi,j,t = sj x Yi,j,t, where sj is the spread of product j and Yi,j,t is the outstanding amount on customer’s i account at time t. – The future activity of the customers Yi,j,t will be predicted using an adaptation of a Vector Auto- Regressive (VAR) model. 19
- 20. From Data Sources to CLV-based Strategy: Example Case Study from Belgian Universal Bank Input Model Output ILLUSTRATION Transactional information Customer Lifetime Value Customer past transactions, Customer Activity Customer One measure in Euro per purchases, etc. Model Profitability customer Model • Identification of the CLV-based score Customer characteristics activity drivers Profit as a function Identification of the customers for Socio-demographics information: • Customer activity of the customer a marketing action using the CLV age, address, etc. forecast activity CLV-based tool Identification of the optimal Price/Cost structure Customer Lifetime Value marketing actions using the CLV Information on the relationship Model between the customers’ activity and the profits CLV Estimation based on the CLV documentation discounted future profits Summary statistics Expert knowledge Information from the experts Management presentations Findings summary and Recommendations Data Cleaning Selection Data mining & modeling Databases Data Warehouse Task Relevant Data Pattern Evaluation 20
- 21. Case Study What you need for starting a value-based project • Data and information needed – You need to know who your customers are – You need to know what is your CURRENT customer profitability – You need to know what your customers did in the past • Maturity level needed – Typically, you already implemented customer analytics type of project: o Propensity to buy o Attrition modeling – The profitability of your customers is a key question • Type of business – Typically, with many customers, and a lot of past transactional data available – Example of industries: Retail Banking, Telecommunication, Pharmaceuticals, Retailers 21
- 22. Case Study Lessons learned from past projects • The risks – Politics, conservatism, etc. – Heterogeneity of your customer base: adapt your segmentation accordingly – Endless arguments on the price and cost structure: use marginal revenues! – Mature products vs. newly developed products: discard new products! – Profitability approach might be conflicting with existing sales incentives (e.g. volume- based) • The key success factors: – It’s a business project, not an IT one! The project has to be led by the marketing department Have someone from marketing leading the project Knowing that the effort will be 80% IT – Be pragmatic: use the 80/20 rule – Be realistic: it is impossible to predict what will happen in 20 years with a 90% accuracy – The Key Question is “HOW WILL THE VALUE BASED SEGMENTATION WILL BE ACTIONED?” 22
- 23. Case Study Conclusions Value-based segmentation allows to allocate resources where they'll deliver the highest payoff – Marketing actions can be implemented in a more optimal way – Customers can be targeted more profitably Value-based segmentation models and CLV Models can be estimated using standard procedures – We model the future activity of the customers using an adapted Vector Auto-Regressive Model – Markov-chains are an efficient alternative when an aggregated level (segment) is needed By taking into account the relevant drivers of customer activity, accurate and reliable predictions are made: – Owing to the model simplicity, the estimates can easily be interpreted – For prediction, we achieved a correlation of 95% between predicted and actual over the first three months 23
- 24. Executive summary As market dynamics are changing, companies are looking for new segmentation to improve the way they attract, manage and retain their customers. This re-segmentation exercises can be used to support strategic objectives, such as identifying new market potential, or tactically, to improve retention, ROMI and cross- selling. However, moving from traditional segmentation methods to more advanced methods require marketing organizations to invest beyond their basic capabilities Within the advanced methods, the value-based segmentation allows for prioritization of the customers portfolio according to their economic value (past or future). One of the most effective value-based approaches is Customer Lifetime Value (CLV). While CLV can be seen as complex, its pragmatic application yields most benefits and the use of customer lifetime value as a marketing metric tends to place greater emphasis on customer service and long-term customer satisfaction, rather than on maximizing short-term sales A multi-layer segmentation model leverages the particular benefits of each segmentation method by crossing basic segments and value-based approaches. This results in operational segments allowing more efficient targeting actions (acquisition, costs reductions, loyalty, etc.) and connecting with your customers Segmentation projects are typically approached in seven steps, from defining objectives of the segmentation exercise to assessing its effectiveness. Key success factors are to implement a standard data collection process, to respect segment definition rules and to ensure continuous improvement of the segmentation model 24
- 25. We recommend a multi-layer approach crossing value-based segmentations with other segmentations methods Prioritization of your customers Leveraging categorization portfolio according to their data from traditional economic value segmentation methods A Value-based approach B • Annual turnover Basic Segments Segments de marché • Sourcing / production segmentation costs Value Valeur XL L M S • Behavior-based • Network costs • Consumption level • Segment management ++ • Expectations level cost • Usage • Lifetime of a customer + • … • … - A + B Operational segments • Same economic value • Similar profiles •… Obtaining granular and actionable segments 25
- 26. Crossing basic segments with value-based approaches allows to define operational and actionable segments Multi-layer components Strategic segments Sub-segments Impact Behavior-based Value-based • A global and Geographic A.1 A.2 Group A consistent go-to- A.3 market strategy Socio-demographic within and across product lines B.1 Behavioral Group B B.2 B.3 • Integration of client’s long-term potential Value-based C.1 C.2 C.3 … Group C • Aggregation of C.4 C.5 customer value • Valuation • Market and product planning • A ‘ROMI’ approach Purpose • Prioritization • Campaign planning and execution • Resource allocation • Marketing communications planning • Channel assignments • Granular view of target markets and motivators • Easy to understand and to Benefits use • Actionable: basis for offer development, campaign • Sustainable segments to targeting and market/brand achieve a differentiated positioning position 26
- 27. Executive summary As market dynamics are changing, companies are looking for new segmentation to improve the way they attract, manage and retain their customers. This re-segmentation exercises can be used to support strategic objectives, such as identifying new market potential, or tactically, to improve retention, ROMI and cross- selling. However, moving from traditional segmentation methods to more advanced methods require marketing organizations to invest beyond their basic capabilities Within the advanced methods, the value-based segmentation allows for prioritization of the customers portfolio according to their economic value (past or future). One of the most effective value-based approaches is Customer Lifetime Value (CLV). While CLV can be seen as complex, its pragmatic application yields most benefits and the use of customer lifetime value as a marketing metric tends to place greater emphasis on customer service and long-term customer satisfaction, rather than on maximizing short-term sales A multi-layer segmentation model leverages the particular benefits of each segmentation method by crossing basic segments and value-based approaches. This results in operational segments allowing more efficient targeting actions (acquisition, costs reductions, loyalty, etc.) and connecting with your customers Segmentation projects are typically approached in seven steps, from defining objectives of the segmentation exercise to assessing its effectiveness. Key success factors are to implement a standard data collection process, to respect segment definition rules and to ensure continuous improvement of the segmentation model 27
- 28. The typical process to an effective segmentation : a 7- steps approach • Which business line(s) and customers are concerned, what is the time frame of the study ? • What is the depth and the final objective of the • Define your strategic objectives and the 1 study (strategic or tactical applications) ? associated key performance indicators 2 Define business • Determine the relevant segmentation Determine objectives assessment methods based on those objectives segmentation • Identify the data source(s) to be used scope (internal data collection for mature segments, customer survey for new 7 customers, etc.) Assess 3 • Identify audience and determine survey • Review and follow your key segmentation and / or data collection tools, content (for performance indicators (market Collect qualitative or quantitative data), and share evolution, sales growth, new effectiveness data administration plan (through reps, etc) customer base, etc) • Ensure / Check data quality and reliability • Assess segmentation update needs 6 • Analyse data 4 • Summarise customer responses into Go-to-market Analyse the predictive models of segment and customer • Develop strategies to serve individual segmentation behaviour segments 5 Identify • Review current market and product • Develop product segmentation based on offerings value to customer and value to business segment profiles • Define current customer base by segment • Develop recommendations for pricing dimensions (products consumed, actions based on segment specific geography, etc) and define “clusters” of purchase behaviour and buying process • Characterise each segment by determining customers with similar needs for products customers key differences and similarities (cluster consumption analysis) • Determine each segment size and purchasing power / profile 28
- 29. Key success factors Organize Respect Customize Develop Data collection & 5 golden rules of Segmentation strategy Continuous improvement segmentation process segment definition approach Standardize common To be useful, the • “De-average” the market Segmentation is a processes to achieve segments you continuous, rather than • Assess the potential of synergies and build an identified should be: linear, process: each segment (size, organizational growth, uniformity, structure to manage it • Markets and competition, etc) segments are • Determine data • Homogeneous within, • Select the best dynamic and unstable requirements heterogeneous segments to serve : over time across according to their Conduct • Collect data from profitability or your Segmentation adequate sources • Measurable competitive advantage • Manage and analyze • Identifiable (align segment Implement data characteristics with your Strategies • Actionable capabilities and Measure Effectiveness • Substantial competencies) • Define adequate business models to serve them profitably 29
- 30. Standard approach to segmentation design and roll-out Our approach ensures understanding and buy-in from all stakeholders impacted by a new segmentation Actionable segmentatio 3. Roll-out & n approach Scoping 1. Pilot 2. Go / No Go follow-up to retain and acquire new customers Double track testing Pilot results Communication of results • One sample with traditional • Communicate and discuss • Ensure visibility of results at marketing campaigns the comparison of value- group level (marketing • One sample with value- based approach to previous boards...) based segmentation traditional approaches • Ensure understanding and A posteriori analysis • Demonstrate value-based buy-in of all stakeholders • Previous campaigns to be benefits through business • Refine and extend first analyzed case approach segmentation approach to Initial analysis of results Go overall project scope • Compare value-based • Results and benefits are in • Assess impact on approach to previous line with expectations – organization traditional approaches proceed further • Share regular feedback/follow Collaboration and No Go up and assessment of value- communication with • Results and/or benefits are based segmentation clients teams not in line with expectations deployment with entities – stop the project • Collaboration and communication with client teams 30
- 31. Sample of references • CLV model definition and approach Retail banking Launch of a pilot to implement • Pilot phase reference value based segmentation • Results analysis and recommendations • Qualitative assessment of internal and external data Definition of operational • Analysis of customer’s expectations segmentation to enhance • Definition of operational segmentation sales performance • Definition of action plan • Analysis of agent’s expectations Definition of operational • Analysis of customers' expectations segmentation of GDF Suez • Identification of segmentation axis network • Analysis of partners' segmentation • Definition of operational segmentation Definition of operational • Definition of a customer relationship policy for each segment segmentation to enhance • Definition of needs for the new CRM software customer value • Definition of product offer processes Definition of multi-channel • Definition of multi-channel customer relationship policy for after sales operations for customer relationship policy each segment (Grand voyageur, Seniors, 12-25, Escapade…) • Definition of macro-segmentation (4 segments) Definition of macro- • Definition of a customer relationship policy for each segment for acquisition and segmentation and customer relationship policy retention • Deployment of new customer relationship policy Definition of segmentation • Definition of operational segmentation and customer relationship • Development of new services to improve relationships with insured, health policy professionals and employers 31