A data science project delivers on the business case developed during the assessment. The gains are typically financial, but may also include important but hard-to-quantify benefits such as improved customer goodwill.
A modest data science project may only require a single data scientist for two or three months. A complex data science project might require five or six consultants for six to 18 months. A data science project constructs and possibly operationalizes a scientific model that helps optimize a set of business decisions, with the goal of realizing the business case identified in the assessment. A data science project always includes the following activities:
- data sampling
- variable selection
- development of the scientific model.
If the project calls for developing a decision support system (DSS) or decision automation system (DAS), the project may also include most of the following activities:
- data architecture
- ETL development
- solution development
- simulation development
- verification and validation
Data science project deliverables may include:
- a report or presentation summarizing scientific findings
- a production DSS or DAS
- training in proper use of the DSS or DAS
- metrics tracking progress towards realizing the business case
- periodic enhancements.