Virtual segment brief
Download as PPTX, PDF0 likes11 views
The document outlines a framework for assessing customer attributes and their impact on customer value through a model analogous to electric circuit resistance. It discusses strategies for optimizing market penetration, implementing automated analysis, and leveraging big data for enhanced business insights. Additionally, it emphasizes the importance of data governance and agile planning for successful implementation.
Virtual segment brief
- 2. • Current feature assessment The effect of already existing attribute on the overall base customer value • New feature assessment Ingesting new attribute and see how it can change the fabric of and play manner • New acquisitions profiling What kind of market you need to penetrate to optimize the customer sized sample ingestion • Recommendation (Next best action) Which segment to target and tune the top 3 attributes to tune to one level up Summary Model Output
- 3. • Uncertainty: Each feature or attribute at the same time is contributing in resisting the overall customer potential and increasing the customer overall potential as well. • Increasing new parallel resistance will decrease the overall resistance • Increasing new series resistance will increase the overall resistance Model Analogy Electric circuit resistance analogy (Assumptions)
- 4. • Parallel resistance : 1 𝑅 𝑇𝑜𝑡𝑎𝑙 𝑝𝑎𝑟𝑎𝑙𝑙𝑒𝑙 = 1 𝑟1 + 1 𝑟2 + 1 𝑟3 + ⋯ + 1 𝑟 𝑛 • Series resistance 𝑅 𝑇𝑜𝑡𝑎𝑙 𝑠𝑒𝑟𝑖𝑒𝑠 = 𝑟1 + 𝑟2 + 𝑟3 + ⋯ 𝑟𝑛 • Parallel resistance quotient: 𝑅 𝑇𝑜𝑡𝑎𝑙 𝑃𝑎𝑟𝑎𝑙𝑙𝑒𝑙 = 1 𝑖=1 𝑛 1 𝑟 𝑖 • Series resistance quotient: 𝑅 𝑇𝑜𝑡𝑎𝑙 𝑠𝑒𝑟𝑖𝑒𝑠 = 𝑖=1 𝑛 𝑟𝑖 • 𝑉𝑇𝑜𝑡𝑎𝑙 𝛼𝑅 𝑇𝑜𝑡𝑎𝑙 𝑃𝑎𝑟𝑎𝑙𝑙𝑒𝑙 • 𝑉𝑇𝑜𝑡𝑎𝑙 𝛼𝑅 𝑇𝑜𝑡𝑎𝑙 𝑆𝑒𝑟𝑖𝑒𝑠 • 𝑉𝑇𝑜𝑡𝑎𝑙 = C𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝑥 𝑖=1 𝑛 𝑟 𝑖 𝑖=1 𝑛 1 𝑟 𝑖 Equation Analogy V Customer potential r Customer feature C Business Domain/Market constant
- 5. Virtual Segment Data Set Customer Attribute Is Categoric al? Calculate Attribute (Frequency per dataset) Numerical Calculate the normalized, standardized, scaled unitless value from 1 to 100 Per customer (Customer Value) Unsupervised clustering with 10 centers over the customer value Get the maximum difference value between 2 consecutive clusters Query to get the customer base with the lower boundary Tweak or tune the original attribute Calculate the Rank per customer Yes No New Attributes Remove correlation, redundancy & causality Get the least 3 value attributes (X per this cluster Recommendation 𝑖=1 𝑛 𝑟𝑖 𝑖=1 𝑛 1 𝑟𝑖
- 6. Sampling Criteria (example to generate sample virtual customers) Customer Total Customer revenue from activation to date 5% Random Sample from the total base Calculate the average and Standard deviation for the total revenue 100 times Calculate the average of means, final standard deviation of standard deviations Select count (distinct customer) from base where total revenue between average_of_means and ±5 final_standard_deviation * average_of_means
- 7. • Before big data you cannot ingest massive amount of columns with ease to a dataset, columnar databases made that feasible • The model is independent of the number data sources, you can add new data sources instantaneously . • Processing the model with scalability and efficiency on big data platforms. Utilizing Big Data - MPP
- 8. SWOT Strength Weakness Opportunity Threat • Automating the costly manual analysis to next best action • More certainty on variable tuning (ie age, plan) • Predict success rate of campaigns with minimal cost • Identify which variable to target and test campaign with virtual segments immediately • Identifying the correlation for each added attribute against existing attributes, can be costly and complicated • Some times would need manual consultancy or advise from domain experts • It is complicated to perform the same analysis on cross net base customers
- 9. What’s Next? Leveraging on Big Data High Level Example Source 1 Source 2 Source 3 Source 4 Source 5 Subscriber Hive- Hbase Golden Record per customer from activation date (delta Monthly Data aggregated) Spark QL Hive Link to Data Model Final stored data Model execution Because of Big Data: • 100k+ columns can be processed per dataset • Billions of records • Commodity of hardware • Scalability • Reliability and supporting business continuity Subscriber Subscriber Subscriber Subscriber
- 10. Virtual Segment • Approachable sales/marketing method • Time to market will decrease • Better traffic/capacity projections • Better target customer group • More data driven organization – centralized data team
- 11. 11 Thank you!
- 12. Implementation Readiness • Plan -- Gantt Chart • Project Stakeholders • Cost Benefit Analysis • 6 month break even agile sprint • Data governance plan • End to end data architecture • Hardware and resource planning