KDD 2019 IADSS Workshop - How Data Scientists can bridge the gap between Data and Business - Ary Bressane
0 likes374 views
This document discusses how data science projects often fail due to a lack of business adoption and the gap between data and business needs. It provides statistics showing that 85% of big data projects fail and 80% of AI projects do not scale within organizations. Common reasons for failure include solving the wrong problem, having the wrong data or skills, and not clearly defining the business purpose. The document then discusses how organizations are addressing these issues through design thinking, design sprints, and focusing on actionable insights and simple, prototype models. It provides an example of a predictive maintenance model that scheduled maintenance to detect HVAC failures and improve customer service.
KDD 2019 IADSS Workshop - How Data Scientists can bridge the gap between Data and Business - Ary Bressane
- 1. How Data Scientists can bridge the gap between Data and Business Ary Bressane, Head of Data Innovation Lab
- 2. About me Ary Bressane • Head of Data Innovation Lab at MNUBO • Lecturer at Concordia University on the Big Data Analytics program • Co-organizer of MTL DATA Meetup
- 3. Technical Challenges Approaches to tackle technical challenges Methodology, Best Practices, Infrastructure, Technology, Tools
- 4. Big Data & Data Science Projects Failure Rate GARTNER ESTIMATED 85% of big data projects fail (2017). The initial estimation was 60% (GARTNER 2016) THROUGH 2020 80% of AI projects will remain alchemy, run by wizards whose talents will not scale in the organization. (GARTNER 2018) THROUGH 2022 20% of analytic insights will deliver business outcomes. (GARTNER 2018) EXECUTIVE SURVEY 77% respondents say that “business adoption” of big data and AI initiatives continues to represent a challenge for their organizations (NEWVANTAGE PARTNERS 2019)
- 5. Big Data & Data Science Projects Why they Fail DATACONOMY (2016) 1. Solving the wrong problem 2. Mismatch of problem, technology and personnel 3. Data integrity HBR (2018) 1. You need to make your purpose clear 2. Choose the tasks you automate wisely 3. Choose your data wisely 4. Shift humans to higher- value social tasks KDNUGGET (2018) 1. Asking the wrong question 2. Trying to use it to solve the wrong problem 3. Not having enough data 4. Not having the right data 5. Having too much data 6. Hiring the wrong people … https://www.kdnuggets.com/2018/07/why- machine-learning-project-fail.html https://hbr.org/2018/07/how-to-make-an- ai-project-more-likely-to-succeed https://dataconomy.com/2016/06/3- reasons-why-data-science-can-fail/
- 6. Big Data & Data Science Projects Skill Gap
- 7. Big Data & Data Science Projects Skill Gap
- 8. How organizations are approaching the issue?
- 9. How organizations are approaching the issue? Using Design Tools • Focus on human values • Radical collaboration • Embrace experimentation • Bias towards action • Visual • Incentive variety and diversity • Defer judgement
- 10. How organizations are approaching the issue? Design Thinking Norman Tran
- 11. How organizations are approaching the issue? Design Thinking
- 13. Design Sprints Methodology SET THE STAGE • Choose a big challenge • Recruit the team • Lock five full days • Get sprint supplies SPRINT • Runt the Five days • Evaluate the learning
- 14. Design Sprints for Data Science Projects Lessons Learned
- 15. Think first how the data will be consumed (last mile problem)
- 16. Look for Actionable Insights
- 17. Start with a simple model (or not a model at all)
- 18. Lessons Learned Example of business outcome
- 19. Example of Business Outcome Predictive Maintenance FAILURE P(Failure) P(Failure)P(Failure) P(Failure)P(Failure) Failure Detect on HVAC systems Results
- 20. Example of Business Outcome Predictive Maintenance Schedule Maintenance Assistant WORKFLOW MOCKUP + PROTOTYPES Map 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 26 27 28 29 30 24 25 Sun Mon Tue Wed Thu Fri Sat Jan 2018 Schedule CUSTOMER MONITOR ASSETS REACTIVEPROACTIV E SALES TECHNICIAN SERVICE ENGINEERING CUSTOMER SERVICE INVESTIGATE ASSET NEW ISSUE ! KNOWN ISSUE X OPEN TICKET 1 2 3 4 5 The Customer reports a issue to the Sales Rep, the Technician or directly to the Service Team The Service Team investigates the problem with the specific asset In the case of a new issue, a ticket is opened and the Engineering Team informed In the case of a known issue, a ticket is opened or the Customer is contacted to solve the problem The Service Team monitors all the assets on the install base to identify the ones with low health
- 21. Thank you Ary Bressane, Head of Data Innovation Lab ary@mnubo.com