Description
Tests causal explanations within a single case by examining whether the predicted sequence of intermediate steps and evidence is present, using formal evidential tests.
Process tracing examines the causal chain linking a cause to an outcome within a case, asking what observable implications each step of a hypothesised mechanism would leave. Van Evera classified evidential tests by necessity and sufficiency: straw-in-the-wind (weak), hoop (necessary: failing eliminates the hypothesis), smoking gun (sufficient: passing strongly confirms) and doubly decisive. Bayesian process tracing formalises updating. In UX, single cases are common: one enterprise customer's failed rollout, one user's journey to churn, one incident. Process tracing turns case narratives into disciplined causal inference, with rival hypotheses tested against specific evidence.
Attributes
- Discipline
- Qualitative Analysis & Synthesis
- Loss type
- Unmigrated
- Stage
- Post-launch
- Data
- Qualitative
- Effort
- Weeks
Steps
- Define the outcome and candidate causal explanations, including rivals.
- For each explanation, specify the mechanism as a sequence of steps and the evidence each step should leave.
- Classify predicted evidence as hoop, smoking-gun, straw-in-the-wind or doubly decisive tests.
- Collect evidence: logs, documents, emails, interviews, timelines, with attention to source reliability.
- Evaluate each hypothesis against the tests, eliminating those that fail hoop tests.
- Write a causal narrative with the evidence chain, and state remaining uncertainty.
Sources
- George & Bennett, Case Studies and Theory Development in the Social Sciences (2005)
- Van Evera, Guide to Methods for Students of Political Science (1997)
- Beach & Pedersen, Process-Tracing Methods: Foundations and Guidelines (2013)
- Collier, Understanding process tracing (2011), PS: Political Science & Politics
Also known as
- Causal process tracing
- Within-case causal inference
- Van Evera's tests
Categories
- User Research
- Qualitative Analysis & Synthesis