A practical guide on adapting Agile/Scrum methodologies for data science teams. Covers what works (periodic planning, retrospectives, demos) and what doesn't (estimation difficulty, engineering-like sprint expectations). Proposes a time-boxed iteration framework with four stages: feasibility assessment (2–4 weeks), proof of concept (4–8 weeks), production deployment (3–9 months), and ongoing maintenance. Also recommends sprint rituals like planning, prioritization, demos, and retrospectives, plus a project planning document template covering problem statement, intent, success metrics, deliverables, business value, dependencies, and constraints. Includes guidance on carving out dedicated innovation time (20% or quarterly sprints) within data science teams.
