Data Analytics 201: Adding Value with Modeling Techniques
Download as PPTX, PDF1 like293 views
The document discusses a webinar focused on data analytics in asset management, highlighting strategic challenges faced by managers, including trust issues and budget constraints. It presents the findings of a study on third-party data usage, indicating key applications such as data cleaning and lead generation. Additionally, it covers methodologies for customer segmentation and measuring success in analytics strategies while emphasizing the importance of blending data science with business insights.
Data Analytics 201: Adding Value with Modeling Techniques
- 1. www.nicsa.org | #WebinarWednesdays Data Analytics 201: Adding Value with Modeling Techniques October 18, 2017
- 2. www.nicsa.org | #WebinarWednesdays Pct. of Asset Managers Rating Strategic Hurdles Somewhat to Very Challenging Strategic Challenges Managers >$100B AUM Managers <$100B AUM Loss of wholesalers' trust when data aren't comprehensive 63% 57% Budget constraints prevent fully executing on data strategy 50% 82% Limitations in quantity or experience/skills of data personnel 50% 50% Salespeople are resistant to greater adoption of data analytics 50% 48% Poorly defined policies on who owns various data responsibilities 38% 54% Unclear how to show that data analytics value justifies cost 33% 54% No clear strategy for what data analytics are supposed to do 6% 50% Source: Ignites Research
- 3. www.nicsa.org | #WebinarWednesdays MODERATOR: David Lieberman Vice President, Product Development Albridge Analytics Jackie Noblett Senior Reporter Ignites Lyndsay Noble Lead Analytics Consultant DST Systems PANELISTS: Greg Piaseckyj Head of Sales SalesPage Technologies LLC Deep Srivastav Head of Client Strategies & Analytics Franklin Templeton
- 4. www.nicsa.org | #WebinarWednesdays Asset Managers' Top Uses of Third-Party Vendors' Data Uses of Vendor Data All Managers Managers >$100B AUM Managers <$100B AUM Data cleaning and enrichment 89% 100% 82% Territory management 75% 94% 64% Lead generation 73% 81% 68% Advisor segmentation 70% 94% 57% Marketing campaigns 70% 75% 68% Matching products to advisors 66% 88% 54% Source: Ignites Research
- 5. www.nicsa.org | #WebinarWednesdays Methodology: RFM segmentation A data driven extension of the segmentations that many firms use right now Can be applied to • purchases • engagement • portfolio diversity • any concept for which these variables make sense ecency requency agnitude
- 6. www.nicsa.org | #WebinarWednesdays Gather & Prepare Your Data Purchases Engagement Portfolio diversity Recency and Frequency Date of every purchase Date of every touchpoint Date of first purchase for every product Magnitude Dollar amount of every purchase Type of touchpoint and importance AUM for every product How far back? 2-3 years Up to 1 year 3-5 years + 0 5000 10000 15000 20000 25000 30000 35000 40000 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 260 270 280 290 300 310 320 330 340 350 360 370 Create Logical Cutpoints Recency Example
- 7. www.nicsa.org | #WebinarWednesdays Create Your Segments
- 8. www.nicsa.org | #WebinarWednesdays Best Practices • Be conservative with the number of cut points on each variable • Include your business experts in the decisions • Don’t go too far back in time
- 9. www.nicsa.org | #WebinarWednesdays Recency Frequency <= 15.00 15.01 - 20.00 20.01 - 25.00 25.01 - 30.00 30.01+ 1 9,728 11,597 8,456 3,786 3,727 2 6,609 12,962 10,166 4,152 2,942 3 - 4 7,689 19,528 16,023 6,177 3,333 5 - 7 5,610 16,247 14,277 5,060 2,145 8+ 6,830 18,204 15,294 4,699 1,761 1 8,457 13,450 10,088 4,405 4,503 2 6,721 14,322 11,571 4,881 3,479 3 - 4 7,084 19,113 15,999 5,915 3,490 5 - 7 4,716 13,281 11,062 3,671 1,755 8+ 4,010 10,591 7,568 2,318 990 1 10,235 14,984 11,197 5,031 5,140 2 9,865 19,063 14,886 6,302 4,675 3 - 4 8,705 20,946 16,216 6,276 3,447 5 - 7 4,714 12,276 9,262 3,167 1,419 8+ 3,244 7,870 5,486 1,664 628 1 7,092 14,711 9,673 4,655 5,088 2 7,939 18,005 13,420 5,796 4,580 3 - 4 6,059 16,873 12,774 4,947 2,876 5 - 7 3,030 8,915 6,626 2,240 1,097 8+ 2,148 5,494 3,912 1,129 487 1 6,842 10,100 6,631 3,125 3,227 2 6,811 12,597 8,970 4,007 2,861 3 - 4 5,206 11,212 8,214 3,073 1,787 5 - 7 2,265 5,647 4,150 1,305 663 8+ 1,451 3,397 2,313 645 292 22 weeks+ Ave_Value 0-4 weeks 4-10 weeks 10-16 weeks 16-22 weeks Example – Buying Pizza
- 10. www.nicsa.org | #WebinarWednesdays Methodology: K-Means Algorithm Clustering is grouping of data or dividing a large data set into smaller sets of similar classes • Define the analytics roadmap for deliverables with business and data science teams • Identify data sources and perform data quality • Analyze variables (Correlation?) and identify key variables for model inclusion • Standardize variables and/or create derived variables if necessary • Execute algorithm to create the final K segments • Validation of segments - review significance statistics • Generate summary statistics for each segment • Generate profiles of each segment
- 11. www.nicsa.org | #WebinarWednesdays Sourcing the ‘Right’ Data Develop structured approach to data acquisition, cleanliness, and analytical insights • Industry Providers and Third Party participants • Internal systems – CRM sales activity, AUM, product usage, or content preferences • Distributors Directly Quantitative Measures • Financial Statistics – AUM, AUM per Advisor, Sales, or # of fund companies • Business Mix/Asset Allocation – Market Share by asset class or product category • Risk and Return Sensitivity • Relative Benchmarks • Cost Determinants Qualitative/Categorical • Communication Preference • Industry Designations or Ten Information Classification: Confidential
- 12. www.nicsa.org | #WebinarWednesdays Use Case - Define, Design, and Integrate • Objective: To create a unique quantitative segmentation of 4k+ branches that comprise the Albridge Analytics’ national broker dealer database • Approach: Combine the domain distribution knowledge of the business with the statistical expertise of data scientists • Analytical Model: To create segments where homogeneity within a segment is maximized, and heterogeneity between segments is maximized • Integrate Results: Utilize the developed segmentation to provide deeper analytical insights − Identify the most profitable branch segments and advisors − Develop more targeted marketing and communication strategies − Efficiently allocate of resources and make strategic deployment decisions − Understand ownership patterns and buying behaviors Information Classification: Confidential
- 13. www.nicsa.org | #WebinarWednesdays Segment Profile: Modest Size/Performance Driven • Segment represents ~250 branches, and holds 4% of mutual fund assets and experienced 5% of mutual fund sales. • Displays greater propensity to own and purchase high performing mutual funds. However, the segment exhibits a lower sensitivity to purchase and own high performing ETFs as categorized by Lipper Leader measures. • This segments displays a higher than average orientation to MF Large Cap investment styles/ • Mutual Fund velocity is lower in this segment, which signals a lower than average Sales to Asset ratio. • For Alternatives, the segment tends to have a lower concentration Alternative styles, and also experiences a lower sales growth rate. This segment represent moderately sized branches with a higher MF Equity asset class orientation. The segment displays a higher propensity to own and purchase high performing MFs, while less sensitive to ETF return volatility. There is relatively lower demand for Alternatives and other niche investment styles. Key Statistics MF ETF SMA Average AUM $189M $35M $66M Average Sales $29M $14M $42M Information Classification: Confidential
- 14. www.nicsa.org | #WebinarWednesdays Measuring Success & Practical Applications The 3 stages of developing an effective strategy to measure success • Stage 1: Setting the proper foundation • Stage 2: Utilizing an MDM to provide a “single source of truth” • Stage 3: Measure success and answer some important questions: − Are we targeting the right people? − What data are we missing? − What is the value of the data that we are buying? − Is our sales and marketing strategy effective? Practical Measures − Sales Lift − Marketing message effectiveness − Advisor scoring − Advisor segmentation − Going beyond quantity: Now a factor in measuring the quality of the engagement of an advisor Information Classification: Confidential
- 15. www.nicsa.org | #WebinarWednesdays Measuring Success & Practical Applications Combining the Science and the Art • Scoring to drive a strong segmentation strategy • Predictive models • Advisor scoring system − Measuring the value of interactions that is qualitative rather than just quantitative. Common Pitfalls • Keep it simple!!! • Do it in bit sized chunks • You can only do it for the first time once. Information Classification: Confidential
- 16. www.nicsa.org | #WebinarWednesdays How to Show Value • Pilots are crucial • Help get the concept right • Allow ‘test and learn’ vs having to overanalyze early on • Help blend the ‘art and science’ • Put the distribution teams in charge • Indicative- Intuitive results help early on • Feedback/surveys with distribution teams • Early wins; success stories • Finding advocates helps
- 17. www.nicsa.org | #WebinarWednesdays How to Show Value • Eventually we need more rigor in analyzing impact • Control groups need to be narrowly defined and broadly accepted • Need to account for similar sales, engagement and profiles • Test for statistical significance for results • Share results and analysis with multiple leaders • Build credibility and ask for more!
- 18. Q&AQUESTIONS & ANSWERS SESSION www.nicsa.org | #WebinarWednesdays
- 19. www.nicsa.org | #WebinarWednesdays WEBINAR SPONSORED BY: