Why AI/Data Science Projects Fail (Synthesis Lectures on Computation and Analytics)

Recent data shows that 87% of Artificial Intelligence/Big Data projects don’t make it into production (VB Staff, 2019), meaning that most projects are never deployed. This book addresses five common pitfalls that prevent projects from reaching deployment and provides tools and methods to avoid those pitfalls. Along the way, stories from actual experience in building and deploying data science projects are shared to illustrate the methods and tools. While the book is primarily for data science practitioners, information for managers of data science practitioners is included in the Tips for Managers sections. Preface 13 Introduction and Background 15 Project Phases and Common Project Pitfalls 19 2.1 Tips for Managers 24 Five Methods to Avoid Common Pitfalls 27 3.1 Ask Questions 28 3.2 Get Alignment 28 3.3 Keep It Simple 28 3.4 Leverage Explainability 29 3.5 Have the Conversation 29 3.6 Tips for Managers 30 Define Phase 33 4.1 Project Charter 33 4.2 Supplier-Input-Process-Output-Customer (SIPOC) Analysis 37 4.3 Tips for Managers 42 Making the Business Case: Assigning Value to Your Project 43 5.1 Data Analysis Projects 44 5.2 Automation Projects 45 5.3 Improving Business Processes 46 5.4 Data Mining Projects 47 5.5 Improved Data Science 47 5.6 Metrics to Dollar Conversion 47 Acquisition and Exploration of Data Phase 49 6.1 Acquiring Data 49 6.2 Developing Data Collection Systems 49 6.3 Data Exploration 51 6.4 What Does the Customer Want to Know? 51 6.5 Preparing for a Report or Model 52 6.6 Tips for Managers 52 Model-Building Phase 55 7.1 Keep it Simple 55 7.2 Repeatability 56 7.3 Leverage Explainability 56 7.4 Tips for Managers 57 Interpret and communicate phase 59 8.1 Know Your Audience 60 8.2 Reports 61 8.3 Presentations 63 8.4 Models 65 8.5 Tips for Mangers 66 Deployment Phase 67 9.1 Plan for Deployment from the Start 67 9.2 Documentation 68 9.3 Maintenance 69 9.4 Tips for Managers 70 Summary of the Five Methods to Avoid Common Pitfalls 73 10.1 Ask Questions 73 10.2 Get Alignment 73 10.3 Keep It Simple 74 10.4 Leverage Explainability 74 10.5 Have the Conversation 74 References 77 Author Biography 79 Table 1.1: Five project pitfalls 16 Table 1.2: Alignment between data science project phases and Lean six sigma DMAIC framework 17 Table 3.1: Connection between the methods to avoid pitfalls and the five project pitfalls 27 Table 3.2: Questions to ask at retrospectives 31 Table 4.1: Key components of a project charter 34 Table 5.1: Deliverables and metrics for various types of data science projects 43 Table 5.2: Example calculation for time saved 45 Table 5.3: Types of waste with manufacturing and office examples 46 Table 5.4: Common metrics and dollar conversion 47 Table 8.1: Data science project types and typical final deliverables 59 Table 8.2: Data visualization reading list 62 Blank Page 2 La 4è de couv. indique : "Recent data shows that 87% of Artificial Intelligence/Big Data projects don't make it into production (VB Staff, 2019), meaning that most projects are never deployed. This book addresses five common pitfalls that prevent projects from reaching deployment and provides tools and methods to avoid those pitfalls. Along the way, stories from actual experience in building and deploying data science projects are shared to illustrate the methods and tools. While the book is primarily for data science practitioners, information for managers of data science practitioners is included in the Tips for Managers sections"