Questions for Data Scientists in Software Engineering: A Replication

Conducted at ING—a software-defined enterprise providing banking solutions—this study presents 171 questions that software engineers at ING would like data scientists to answer. This study is a replication of a similar study at Microsoft. We found that the core software development challenges (relating to code, developer, and customer) remain the same. Subtle differences seen relate to the two companies' context and the time gap in between.

Transcript

1. Data in Software Engineering Photo by Markus Spiske on Unsplash
2. ING - Dutch Bank Software-defined Enterprise 600 software teams 2500+ deployments 750+ applications
3. Questions for Engineers Answers from Data Scientists
4. 5 years ago @ Microsoft
5. Impact of MS study 49% 35% 12% 4% Plain reference Example Partly answers Answers
6. Impact of MS study 34% 25% 20% 15% 5% Analytics Testing, quality Process Culture Mobile apps
7. “Do SE questions relevant for Microsoft apply to ING, five years later?”
8. ING and TU Delft
9. Design • Replication of MS study • Survey 1: find data science problems • Survey 2: rank questions on relevance • Comparison of MS and ING study • Reflections for research and industry
10. Similar to MS study
11. Similar to MS study
12. Top essential questions • Developer: Role of team happiness and pleasure at work on productivity and performance of DevOp teams • Best practices: How to build for reusability and scalability? • Bug: Compare efforts spend on fixing bugs and vulnerabilities to writing correct software from start
13. Top unwise questions • Customer and requirement: Software solution in one language, for all person and interests • Best practices: Convert PL1 code to Cobol with readability • Developer productivity: Compare performance of one department against another
14. Different from MS study • Compare broader themes • Word count • Map categories • Compare questions across themes
15. Different from MS study
16. “Core software development problems are same in ING and Microsoft.”
17. “Relevance and types of question change”
18. Context and time matters • Focus: MS on customer; ING at engineering team • Type of questions: Cloud-related questions at MS; deployment-related at ING • Relevance of questions: MS questions on adoption of agile and automated testing; ING at functional aspects • Eye on future: MS questions on legacy code; not at ING
19. Reflections on Replication • Policy influence data availability and design. • Companies have incomparable information. • A study independent of prior study is hard. • Free from corporate influence vs. self-censorship.
20. Software-defined Enterprise Your IT challenges are not different from software company.
21. Software Engineering • What can be answered? What should be answered? • Use new list of questions • Ask questions relevant to software-defined enterprises • Replicate often and in different contexts • Collect data for hard questions
22. Thank You!