Big Data and advanced analytics
Big Data is plenty big, and it's getting bigger. By using advanced analytics, companies are figuring out how to turn that data into value, sales, and growth to beat out the competition.
Transcript
- 1. MARKETING AND BIG DATA Big Data, Better Decisions Tim McGuire, McKinsey & Co. NCDM Conference, Florida, Dec. 3 – 5, 2012
- 2. 295 exabytes If you stacked a pile of CD-ROMs on top of one another until you’d reached the current global storage capacity for digital information …
- 3. … it would stretch 80,000 km beyond the moon
- 4. Every hour, enough information is consumed by internet traffic to fill 7 million DVDs
- 5. 84o,56o m – side by side, they’d scale Mount Everest 95 times
- 6. The world’s 5oo,ooo+ data centers are large enough ...
- 7. … to fill 5,955 football fields The world’s 5oo,ooo+ data centers are large enough ...
- 8. The Big Data market is set to increase from USD 3,2oo,ooo,ooo to USD 16,9oo,ooo,ooo 2010 2015
- 9. By 2020 one third of all data will be stored or have passed through the cloud and we will have created 35,ooo,ooo,ooo,ooo,ooo,ooo,ooo bytes of data
- 10. 10x more servers 75x more files 50x more data By 2020 IT departments will be looking after
- 11. We have produced more data in the last two years … 3,5oo,ooo,ooo,ooo,ooo,ooo,ooo 2010 2012
- 12. Bang … than in all of history prior to that
- 13. Advanced Analytics
- 14. 1986 Data volume 3 exabytes of data Data variety 99% analog Computing capacity 0.001 bn of MIPS
- 15. 2007 Data volume 295 exabytes of data Data variety 94% digital Computing capacity 6,380 bn of MIPS
- 16. Sophisticated analytics Predictive modeling Neural networks Social network mapping Visualization tools
- 17. Tremendous opportunities Creating transparency Enabling experimentation Segmenting populations Replacing/supporting human decision making Innovating new business models, products, and services
- 18. Capturing value from Advanced Analytics … Predictive and optimization models Big data Organizational transformation … is based on 3 guiding principles Decision backwards Step by step Test and learn
- 19. example Direct mailing campaign
- 20. The traditional approach Direct mailing based on generic segmentation
- 21. Regression + clustering based on transaction history Predictive modeling based on customer preferences > 100 campaigns per year > 2 million customers > 10 million transactions
- 22. Customer value modeling Intelligent segmentation for direct mail campaigns
- 23. Average profit from promoted products Average profit from whole basket + 288% Impact + 147% Control group Test group EUR 0.31 EUR 0.79 EUR 0.08 EUR 0.32
- 24. example Next product to buy
- 25. 80 million consumers 100 million transactions
- 26. Multivariate statistics Association rule analysis
- 27. Are being offered the most probable recommendation Get recommendations to generate maximum margin Receive recommendations from other categories to broaden their purchase behavior "infrequent shoppers" "frequent shoppers" "site lovers"
- 28. USD 1 billion identified USD 300 million already realized within 6 months Impact
- 29. example Assortment optimization
- 30. The traditional approach
- 31. The traditional approach Generic allocation of limited shelf space SKUs ranked by sales No substitution of SKUs considered Limited granularity
- 32. The Big Data approach
- 33. Terabytes of data Multi-year transaction data Consumer panel data Loyalty card data
- 34. Advanced statistical methods Stochastic switching model (entropy calculations)
- 35. Advanced statistical methods Stochastic switching model (entropy calculations) Hierarchical clustering (dendograms)
- 36. Advanced statistical methods Multidimensional scaling (consumer decision tree) Market Segment 1 Segment 2 Brand A Brand B Type 1 Type 2 Type 1 Type 2 Flavor 1 Flavor 2 Size 1 Size 2 ▪ Actual behavior (switching, walk rates) ▪ Statistically relevant ▪ Optimal SKU selection per store ▪ Predictive sales forecast
- 37. Revenue growth more than double the category growth in the market Impact
- 38. example Optimizing branch networks
- 39. Segment customers based on channel behaviour and willingness to travel Define branch con-cepts (e.g., advice branch) in line with multi-channel strategy 0.08 FTE Determine required capacity by customer and plot capacity within micromarket using geo-marketing methods
- 40. 0.08 FTE 9.2 9.4 4.3 14.1 6.4 2.4 Assess footprint risk profile and adjust if risks are too high
- 41. 9.2 9.4 4.3 14.1 6.4 2.4 Optimize locations to set up branch for success
- 42. Carefully plan and execute micromarket transition
- 43. Multiple scenarios based on 90 processes 7 million customers 1,000 branches
- 44. 40% cost reduction < 1% revenue at risk Impact
- 45. A new approach
- 46. Data Models Transformation Data consultancies IT software vendors Management consultants
- 47. Delivering end-to-end capabilities Data Models Transformation Data Center infrastructure to support Tb+ datasets Analytics Center of Competence with state-of the-art statistics Solutions Office (MSO) selects, modifies, builds and implements software IT – BTO with experience in implementing sustain-able solutions in existing IT landscape Process – broad expertise in ops, technology, merchandizing, etc. People – capability building, transformational change, organizational design Strategy – decades of experience in developing and implementing strategies
- 48. Additional Reading from McKinsey Advanced Data and Analytics Big Data & Advanced Analytics: Success Stories from the Front Lines http://www.forbes.com/sites/mckinsey/2012/12/03/big-data-advanced- analytics-success-stories-from-the-front-lines/ Making Advanced Analytics Work For You http://cmsoforum.mckinsey.com/article/making-advanced-analytics- work-for-you The One Tool You Need to Make Big Data Work: The Pencil http://www.forbes.com/sites/mckinsey/2012/10/09/the-one-tool-you- need-to-make-big-data-work-the-pencil/ Getting Beyond the Buzz: Is Your Social Media Working? http://cmsoforum.mckinsey.com/article/getting-beyond-the-buzz- is-your-social-media-working Simplify Big Data - Or It'll Be Useless for Sales http://cmsoforum.mckinsey.com/article/simplify-big-data-or-itll-be- useless-for-sales
- 49. tim_mcguire@mckinsey.com McKinsey Consumer Marketing Analytics Center thank you Vicky & Tanja inspired by Gerhard Richter