Playing to Win: Turbocharged Tableau with a GPU Database
Download as PPTX, PDF1 like624 views
Caesars Entertainment, the world's largest gaming company, emphasizes a data-driven approach to decision-making, utilizing Tableau for analytics and performance tracking since 2011. The presentation outlines the evolution of Tableau's integration at Caesars, particularly focusing on the implementation of Kinetica for enhanced speed and efficiency in data analytics. With Kinetica, dashboard loads were accelerated, facilitating real-time decision-making for executive leadership and improving overall performance metrics across various operational areas.
Playing to Win: Turbocharged Tableau with a GPU Database
- 1. # D a t a 1 7
- 2. Welcome
- 3. Playing to Win: Turbocharged Tableau with GPU Database Gene Lee SVP, Chief Analytics Officer Caesars Entertainment Corporation # D a t a 1 7
- 4. Gene Lee
- 7. CEC is the Larg st Gaming Company in the World #1 #1 #1 #1 #1 #1 #1 #1 #1 #2 #2 Philadelphia Atlantic City Kansas City Tunica New Orleans Biloxi Shreveport / Bossier City Laughlin Las Vegas Reno / Lake Tahoe Council Bluffs Chicagoland Cincinnati Cleveland #1 Metropolis Elizabeth Windsor #1 52 casinos across 13 US states and 6 countries #1 and #2 market share in almost every US market Over 60 million Total Rewards customers #1 global publisher in the social casino-style games market EUROPE & AFRICA USA & CANADA
- 8. A Portfolio of Winning Brands
- 9. The Caesars Entertainment’s analytics team is the most talent-rich and data-driven group we have ever worked with. Combining that analytics talent with your unusually rich data, Caesars Entertainment is very enviably situated. -
- 10. I don’t want to act on hunches or on intuition, I want to gather evidence and think through a problem rigorously. That has served the company very well, not only in terms of good decisions, but also in avoiding situations where it was wise not to act.
- 12. Tableau at Caesars Today- Q3 2017
- 13. Tableau at Caesars - 2011 Impressed with external consultant deck (June 2011) • “That can’t be done in Excel…” Demoed Software (July - October 2011) • Partnered with supportive Tableau sales rep team to continue with license extensions despite no clear path for long-term purchase Attended Tableau Conference at Wynn Las Vegas (October 2011) • Attended with key partner in Caesars IT • Stole some great ideas for use in Caesars • Free Tableau Laptop Case Story Analytics and IT left conference excited • Added Tableau Server + Desktop Licenses to Next Year’s 2012 Big Data Budget
- 14. Tableau at Caesars - Q1 2012 to Q3 2014 Advanced Analytics team uses Tableau Desktop for variable discovery • Licenses heavily underutilized across enterprise Couple POC dashboards published on Dev Server • One dashboard view productionalized - inconsistent usage Tableau one among many BI tools around organization • 100 MB Excel is still king • Love/Hate Cognos • Explore Microstrategy • Explore Qlikview Q3 2014 – Bankruptcy rumors proliferate company • VP of Business Intelligence announces retirement • Gene’s role expands by taking over BI department
- 15. Tableau at Caesars – Q4 2014 CEOC Bankruptcy is imminent • CEOC will publicly file for bankruptcy Jan 15, 2015 • Income Statement / Operating Fundamentals are fine (healthy positive operating cash flow) • Balance Sheet unsustainable (too much leverage due to ill-timed LBO in 2007, can’t cover interest payments) Owners set aggressive 2015 EBITDA Targets • Must hit 2015 EBITDA targets for bankruptcy negotiations to proceed (still on-going today) • YoY Target growth had never been achieved before • “2015 Initiatives” (over 50 discrete initiatives) designed to waterfall and bridge gap between 2014 Act and 2015 Bud Analytics top priority becomes to micro-manage initiatives • Initiative performance tracking critical to hitting plan • Key stakeholders for initiative reporting feedback include external creditors, SMT / Board of Directors, Operators and functional leads • Tableau is chosen to be initiative tracking platform
- 16. Tableau at Caesars – 2015 to present Executive Dashboards rolled out (representative sample) • What is and how confident can we be with current Finance Forecast? • How are daily accounting metrics performing (cuts across time / region / filter exceptions)? • How are labor efficiency metrics performing (labor data merged with POS data)? • How are marketing efficiency trends performing (cut across time / region / segment)? Revenue Management Dashboards (representative sample) • How is my property/market pacing in room nights/revenue vs STLY? • How are my rates trending vs competitors? Any weird /incorrect public rate loads? • What is and how confident can we be with current forecast by channel by stay period? Marketing Dashboards (representative sample) • How are various segments performing (cuts across time/region/demographic/behavior )? • How are key marketing initiatives tracking/performing? • Explore segment elasticities (historic changes in interventions vs changes in play)? • Expense Exposure Insights (forward looking expectation on marketing expense)?
- 17. Tableau at Caesars – 2015 to present VIP Dashboards rolled out (representative sample) • Securely manage host book performance in real-time – how am I performing vs benchmark? • Who/when of my book is arriving at my hotel in real-time? Gaming Dashboards (representative sample) • Any interesting insights on structural gaming hold shifts across various segments? • Slot machine exceptions reporting Food & Beverage / Retail / Labor Dashboards (representative sample) • Heat map (when are certain outlets busy / not busy)? • Performance ranking by outlet (which outlets rank highest by difference metrics)? • Item Pricing Insights (any item pricing opportunities)?
- 18. Tableau at Caesars – 2017 – Need More Power Dashboards are slow when drilling into large datasets • Over 2 minutes for initial dashboard load based on dataset covering 50 million records Test alternative set ups to improve performance • On Premise – EDW, Microsoft SQL Server, Yellowbrick • Software – MemSQL, Kinetica • Cloud – Amazon, Microsoft, Google Key Considerations • Can solution improve performance to under 10 secs per load, per click • Relatively Inexpensive Cost– one time cost plus on-going support and maintenance cost • Existing Security framework hold (Tableau Server On Premise, Microsoft AD, Firewall) • Align with Caesars IT architecture and long-term roadmap • Scalability – how easy is it to expand as data scope increases • Reliability – how often does it crash and we need to open a help ticket to fix • Minimal Commitment – what if new cheaper / faster technology disrupts in 12 months? • Machine Learning model performance (bonus)
- 20. Business Objective Accelerate Tableau dashboards for faster customer 360 analytics Previous CPU-bound architecture did not allow executive leadership to make data driven decisions in real time during meetings
- 21. Competition and Considerations Why Caesars Chose Kinetica Tested alternative set ups to improve performance: Kinetica proved to be best solution • On Premise – EDW, Microsoft SQL Server, Yellowbrick • Software – MemSQL, Kinetica • Cloud – Amazon, Microsoft, Google Key Considerations • Can solution improve performance to under 10 secs per load, per click 5 seconds to load, 5 seconds to refresh based on new filters • TCO • Existing security framework hold (Tableau Server On Premise, Microsoft AD, Firewall) • Align with IT architecture and long-term roadmap • Scalability – how easy is it to expand as data scope increases • Reliability – how often does it crash and we need to open a help ticket to fix • Machine Learning model performance (bonus) • Improved performance to under 10 secs per load, per click 5 seconds to load, 5 seconds to refresh based on new filters • Affordability – one time cost plus on-going support and maintenance cost was within budget, less expensive than on premise databases • Works with IT security requirements, plus have row level security built into dashboards • Align with Caesars IT architecture and long-term roadmap & direction of migration to cloud • Scalability – Easy and predictable scale • Reliability – Hasn’t crashed since launched in July, dependent on Google Infrastructure • GPU Virtual machines powering existing MachineLearning (TensorFlow), added bonus
- 22. Accelerate Tableau and Power BI with Kinetica GPU Database New Capabilities Delivered • 24X faster dashboard loads • 3.5X faster slice and dice, drilldowns, filters Solution Overview • Tableau Server and Kinetica running on Google Cloud Platform • Kinetica accelerates EDW workload • Simply point to Kinetica using Tableau’s replace data source feature • Executive leadership can now make real-time decisions on live data • Analytics are very well served by having GPU powered Kinetica as the engine
- 23. Kinetica BI Acceleration – Add Speed Layer
- 24. Kinetica BI Acceleration – Combine AI & BI Workloads
- 25. Kinetica BI Acceleration – Real Time and Batch Directly
- 26. Caesars Analytics is Hiring!TheMoreYouGive…TheMoreYouGET. 100 Top MBA Employers
- 27. Q & A
- 28. Please complete the session survey from the Session Details screen in your TC17 app
Editor's Notes
- #13: In 2010, Caesars contributed over $58 million dollars to communities in which we operate. Get, Guide, Root Diversity & Inclusion Volunteering, philanthropy, sustainability, Engagement, Social Impact
- #24: Reducing overhead and complexity of current methods to speed up BI workloads, potentially replacing one or all the following: Cubes Pre-aggregated RDBMS views Tableau extracts, proprietary BI query files and/or flat files and their proliferation Localized RDBMS instances and their proliferation Maintaining and modeling more complex Hadoop SQL query layers More costly data appliances that are also more difficult to scale over time
- #25: The ability to leverage massive data sets for BI and ML workloads all within the same database, centralizing data science team use with business team use and the application of analytics: Minimizes the need for moving massive data extracts and their proliferation across separate and isolated ML platforms Enables data science teams to leverage SQL query, API, and BI tools for analytics and reporting on ML outputs Enables business users with the ability to run their own ML, freely iterate and explore ML scenarios with their preferred BI tool of choice
- #26: Kinetica’s powerful multi-thread, multi-head ingest is ideal for real time analysis of massively parallel, multi-source, streaming updates, and handles batch loads faster and easier than the traditional RDBMS technologies, as well as many of the newer, modern distributed data platforms: Introduce the ability for live query during real time, streaming updates to tables Reduce batch windows for refreshing analytic data sets Add new sources of data for analysis faster and easier, with less constraints and less additional tuning maintenance