Advanced Customer Analytics for B 2 B A

Advanced Customer Analytics for B 2 B: A pragmatic approach Paulo Costa Advanced Analytics Director, Cisco Systems San Diego, 09/21/2015

Our Charter: Deliver Insights that will impact Sales and Marketing Initiatives and that are derived through Advanced Analytics (Statistics, Data Mining, Optimization) • A team of Statisticians, Econometricians, Mathematicians and Data Miners working collaboratively - Using a tested Analytical Framework, Modeling Methodology and Business Impact Evaluation Protocols - More than 100 years of combined experience with more than 40 academic papers published and presented. - Specialized in Data Mining, Forecasting and Optimization - FY 09 CMO Global Marketing Award - 2008 Gold Medal for Analytical Modeling from the NCDM

RAD Analytical Framework Summary Where to Compete How to Compete Identify sub markets of interest: countries, industries, subindustries, empl-size, etc, on which CISCO needs to focus Determine the optimum marketing mix for each sub-market for a given set of marketing objectives (Wallet Share Improvement, Acquisition and Retention). Map RAD Objectives Define/Optimize Mktg Mix Identify in each priority sub market what is the core marketing objective associated with each company Determine the appropriate marketing mix (4 Ps) for each sub market and for each core marketing objective Analytical Models that support it: Bi-Dimensional or Multi-dimensional Game Boards Analytical Models that support it: Portfolio analysis with dimensions such as market attractiveness, market position and competitive strengths Analytical Models that support it: Purchase Behavior Segmentation Propensity to Buy Market Basket Price Optimization Partner Optimization R A D …. … … Promotion . … . . Price . . . Place …… … . . Product Ex. : each company in the sub market S 1 will be assign to one WAR Game Board Cell Customer Valuation Models Segmentation Models

RAD Models Forecasting, Economic Analysis & Models Development, Acquisition Retention Product P 2 B, Refresh Cycle, Market Basket, Adoption Curve Promotion “Marketing Mix Optimization”, Brand Optimization, Funnel Stage Identification Price Optimization, Price & Adoption Curve Place Partner Led, Partner Optimization, Partner Wallet Share Customer & Prospects Valuation Models (Wallet & LTV) Customers and Prospects Segmentation Model

Modeling Methodology Task I Business Discovery The purpose of this task is to understand the business question, marketing objectives and business process to allow the specification of the analytical solution. Task II Data Extraction The purpose of this task is to identify possible independent and dependent variables, map the necessary data sources, perform the extraction and quality assure the extracted data. Task III Sampling The purpose of this task is to decide about sampling and if necessary, to design and perform the it. Task IV Data Task VIII Data Transformation Exploration The purpose of this task is to investigate possible associations within the extracted data. This task also includes missing data and outlier treatment. Task VI The purpose of this task is to perform all the necessary transformatio ns to comply with techniques data assumptions and/or to maximize model performance. Modeling The purpose of this task is to develop the analytical model using a set of adequate techniques. Assessment The purpose of this task is to select the best analytical model and perform the selected model business validation. Expansion The purpose of this task is to perform the necessary expansion to the population and make the model available in the appropriate platform.

Product Refresh Cycles Major Business Questions Answered • What is the average refresh cycle for a particular Cisco Technology across the retail banks in China? • What is the refresh cycle for a particular Cisco Technology in the company XYZ in Australia?

Funnel Position Location Major Business Questions Answered • How many and what are the major states in the marketing funnel? What is the relationship within them? • Can we locate a particular company in the funnel and track it evolution?

P 2 B Models Major Business Questions Answered • To which Cisco customers can we cross sell (or up sell) a particular Cisco Technology in the next 2 quarters? How much will they probably spend? • With which Cisco prospects do we have the best opportunity to convert to first time buyers of Cisco Technology (wireless, UC, security, routing and switching) in the next 2 quarters? How much will they probably spend? .

P 2 B Models Deployment Model Testing Protocol (hidden participant control) Companies from the market in study -++ +++ - - - + + -- - - +++ -- - +++ - + + ++ ---++ +++ + How can we verify if a predictive model produces the expected results? 1. Start with all companies from the market in study; 2. Use the model to score companies on likelihood of responding to the action; 3. Randomly select companies to be part of the control group; 2 4 3 - -- 4. To create the target group select companies with the highest scores (and which are not already in the control group); 5. Merge both groups to create a single list for action, but hide the scores and the group associated with each company. (This allows the analysis of the model independently of the effect of the marketing actions that will follow); 1 --- +++ + ++ Control group (random) 5 +++ +++ Target group (only highest scores) ! ! +++ ! ! 6 List of companies for actions 6. Use the list in marketing actions (may include touches: !); 7. When ready to analyze results separate each company back to its original group; 8. Compare success metrics from each group. 7 - -- 8 +++ --- + Control group ++ +++ +++ Target group +++

Customers & Prospects Wallet • Major Business Questions Answered • How much a customer or prospect could spend, in a particular fiscal year, on a specific Cisco technology? • What was the Cisco Wallet Share for Routing per company WW in FY 14, FY 13 FY 12, FY 11 and FY 10? Is it growing? • What will be the Wallet for Routing per company WW in FY 15 and FY 16? • Which customers should we Retain? Which ones should we grow with? Which should we acquire?

Segmentation • Major Business Questions Answered • Which customers and prospects respond similarly to the marketing mix? • How have Cisco customers been buying Cisco technology? • What are the Recency, Frequency and Monetary Value for a particular Customer and Technology?

Segmentation Evaluation Criteria • • • Identifiability • The ability to identify the customers in each segment using available variables. Substantiality • The size of each segment need to be a large enough portion of the market to ensure the profitability of target marketing programs. Accessibility • The ability to reach the targeted segments through promotional or distributional offers Stability • If the segments maintain their characteristics for a reasonable time horizon Responsiveness • If the segments respond uniquely to market actions targeted at them Actionability • Segments are actionable if their identification provides guidance for decisions on the effective specification of marketing actions. Source: Frank, Messy and Wind

Economic Analysis • Major Business Questions Answered • What are the economic drivers of bookings/revenue/TAM? Are there any leading economic indicators? • How does the macroeconomic environment impact the IT market, TAM, Cisco bookings/revenue? • What are likely economic scenarios and how are bookings impacted? • What is the forecast of macroeconomic activity by country over the next few years?

Forecasting Models • Major Business Questions Answered • What is the TAM for the next few years by country, technology, vertical, employee size class? • What is the outlook for global product bookings, product revenue, service revenue for the next two years? • What is the forecast for services bookings over the next few years? • What is the market share forecast over the next 5 quarters for Switching by competitor?

Optimization Models • Business Questions • What is the allocation of sales and marketing investments that maximizes Cisco bookings? • How much should Cisco invest in brand advertising in each country to maximize bookings? • What is the discount that maximizes the expected value of a non-standard deal? • What is the optimal WPL net price by sku for high-velocity products?

Advanced Analytics Models Classification Development Stage Production WW D&B Imputation WW Sampling WW Wallet WW P 2 B WW Bookings WW TAM Macro Scenario On Demand Funnel Stage Id First of a Kind Data Mining Forecasting Optimization Type

Lessons & Pitfalls • Outsourcing? • Competitive Advantage X Insurance Policy • Deployment Team • Data Quality • B 2 B Data sources • Optimization Paradox

Contact Information • If you have further questions or comments: Paulo Costa Advanced Analytics Director costa@cisco. com 954 -535 -5003 18