Operating in a Big Data World. Thinking about ROI

Size: px
Start display at page:

Download "Operating in a Big Data World. Thinking about ROI"

Transcription

1 Operating in a Big Data World Thinking about ROI Vincent Dell Anno Managing Director December 9, 2013

2 Big Data in the Enterprise What are we seeing? Approach Just Go For It Characteristics - Big data technologies are well understood - Traditional approaches are failing and measured risk is appropriate - Leapfrogging the competition is a strategic imperative CIO Agenda - Scale-up challenges driving increased costs - End-of-life for current state technology environments - Reducing cost of operations, improving technology flexibility are strategic imperatives - Savings used to fund business innovation Digital Enterprise Agenda - From product focus to focus on customer experience - Focus on issues to outcomes - Big data and big data technologies as enablers - Data discovery environments and methods to enable deep insights Copyright 2013 Accenture All rights reserved. 2

3 Big Data in the Enterprise Five Common Themes The ability to fully leverage data internal and external to the enterprise pose opportunity and challenge for the CIO and business leaders. CIO Agenda Digital Enterprise Agenda Copyright 2013 Accenture All rights reserved. 3

4 The CIO Agenda Design for Analytics Data as a Platform Data Acceleration Leveraging big data technologies to reduce time and costs and increase flexibility of mobilizing, managing and accessing data across the enterprise Developing robust methods for analyzing technology TCO and tradeoffs and accurately estimating big data project activities Leverage experience and deploy savings as a self-funding pool for additional big data projects and activities Typical Projects - Platform right-sizing - End-of-life technology architecture re-design - Hadoop as data staging and data integration layers - Cassandra as online OLTP engine - Tactical reductions in software licensing Copyright 2013 Accenture All rights reserved. 4

5 Big Data in the Enterprise Fabric The CIO Agenda The Challenge: Where to start To reach a transformational future state, the dialogue must evolve from technology feature function debates to what technologies will enable rapid differentiation atscale (and tighter collaboration with the business). Enterprise Adoption Agile Analytics through Big Data Agile BI through Big Data Departmental Strategy Extended Pilot Pilot Proofs of Concept Awareness Big Data Adoption How do I mobilize all my data and operationalize predictive analytics at scale? How do I implement machine learning and traditional analytics to improve my predictive capabilities?? How do I free more data to analyze to those who consume it? How do I leverage these technologies to transform a functional area of my business? How do I extend this pilot and implement it into production? What business problem can we solve for? Can I see how these technologies work? What is this stuff? Level of Sophistication Transformational Future State Value generated from born data and existing data where insights are operationalized at the right speed, anywhere and with no interruption. Data is liberated to enhance decision-making performance and drive to the right outcomes. Focus on driving differentiated results at scale leveraging predictive analytics. Faster access to insights, quicker response to market shifts, and more innovation. Near Term Focus What are big data technologies and how can they solve old problems more effectively and drive new opportunities from which the enterprise can differentiate? What do I need to be thinking about when I deploy these technologies? How will big data change the way my IT functions; how will my business users approach IT for services and interact with enterprise data? Copyright 2013 Accenture All rights reserved. 5

6 The Digital Enterprise Agenda The Strategic Challenge: Transforming the Customer Experience Copyright 2013 Accenture All rights reserved. 6

7 The Digital Enterprise Agenda The Issues to Outcomes Challenge: Where to focus The new normal for the digital enterprise how to combine data and analytics at speed to drive outcomes. Copyright 2013 Accenture All rights reserved. 7

8 Copyright 2013 Accenture All rights reserved. 8

9 The Digital Enterprise Agenda Strategic Digital Objectives Differentiating Tactical Functions Focusing on strategic objectives and where big data, big data technologies and analytics support a differentiating experience Developing clear models for measuring business value drivers and evaluating how social media data, sensor data, and other contextual data drive value a model where more data and better insights create differentiated value Engaging proactively with big data through data discovery at scale insight supported by Hadoop - Predictive maintenance Typical Projects - Social media + - Fraud detection & analysis - Credit risk modeling - Entitlement management - Contextual personalization Copyright 2013 Accenture All rights reserved. 9

10 Being Productive in a Noisy Data World Practicing Data Science Copyright 2013 Accenture All rights reserved. 10

11 Being Productive in a Noisy Data World What s different? Scale Speed Factor Change of Focus More Variables Different Tools Broader Talent Pool Scale Working Directly with Raw Data Shipping to Production Using all Data Without Sampling Example Exploring and determining the right questions to ask, based on the reality found in the data landscape in the organization Dozens or hundreds of variables is no longer a limit; some models allow for millions of variables, allowing for detailed long-tail predictions Hadoop, Mahout, R, Vowpal Wabbit, Python libraries (numpy, scipy, pandas, scikit-learn, nltk, graphlab, sagemath) Data engineers and data scientists are generalists, able to write pipelines, collaborate on analytics jobs, and write services to expose results to client applications 10s of TBs and up of raw data Hands-on pipelining and workflow, being able to run queries across all data sets Web services serving results, running models directly in production, measuring performance in production and bringing results back into analytics jobs Using Hadoop and other parallelizing frameworks, do computation over all data to get fuller picture and lower risk of over-fitting; only need to sample when desired for specific approaches Copyright 2013 Accenture All rights reserved. 11

12 Data Discovery New Customer Segments From BIG DATA To BIG OUTCOMES 200TB raw data 77M records Discover segments using behavioral similarity combining data from multiple channels to enable an enhanced customer experience Weekly learning cycles Discovery of customer groups different from demographic segments Will lead to targeted service and better experience Combine transaction data, web logs, consumer data, call center logs, banking center data, surveys Use of Machine Learning techniques. Cluster customers using more than 400 variables to discovery new focus segments Refine and Learn through quick and agile iterations Copyright 2013 Accenture All rights reserved. 12

13 Data Discovery Experiences Retail Revenue From BIG DATA To BIG OUTCOMES Tracking 100GB logs per day near real time Fast insights for recommending appealing products and services to individual customers First 3 months of recommendation platform Leading to: Enhanced contextaware recommendations Multi-million dollar boost in revenue Combine data from multiple sales channels including mobile devices Log processing and recommendation engine design Large scale machine learning Integrated web application Copyright 2013 Accenture All rights reserved. 13

14 Explore more Journey to Analytics ROI Analytics in Action: Breakthroughs and Barriers on the Journey to ROI High Performers in IT: Defined by Digital Putting Data to Work (video) Data Monetization in the Age of Big Data Data Visualization Turns Raw Data Into Meaning and Meaning Into Understanding What s Your Data Worth? Big Accenture View in slideshow mode to enable links Copyright 2013 Accenture All rights reserved. 14

15 Thank You