Documenting Assumptions and Open Questions
What you'll learn: How to systematically record assumptions, unknowns, and dependencies at project start so you can revisit and validate them as work progresses.
Why Document Assumptions?
Every data science project begins with incomplete information. You'll make educated guesses about data quality, user behavior, business rules, or model suitability. If these assumptions remain invisible—trapped in your head or scattered across emails—you risk building on shaky ground without realizing it.
Think of assumptions like placeholder variables in your project's foundation. You need to mark them clearly so you can replace them with validated facts later, rather than discovering critical misunderstandings during your final presentation.
What to Document
Create a living document (a dedicated markdown file, project wiki section, or notebook header) that captures:
- Assumptions: "Customer churn is defined as no purchase in 90 days" or "Missing values in age column represent unknown, not zero"
- Open Questions: "Does the marketing team track promo codes consistently?" or "Which model families are interpretable enough for regulatory review?"
- Dependencies: "Analysis requires Q3 sales data from finance team by March 15" or "Feature X depends on vendor API availability"
- Known Unknowns: "We don't yet know if historical patterns hold post-merger"
Making It Work
Review this document at each project milestone. Tag assumptions as "validated," "invalidated," or "still uncertain." When an assumption proves wrong, update your scoping document and communicate impact to stakeholders.
Key Takeaway: Explicitly recording assumptions and open questions transforms invisible risks into manageable project elements you can systematically address as information becomes available.