What are the main causal pathways from an intervention to an outcome? We can trace chains of influence from a starting point like an intervention to a key outcome, revealing the step-by-step or branching logic described by the sources. We can even compare the strength of evidence for different pathways.

Path tracing is similar to Looking upstream and Looking downstream,: you can specify the number of steps, and you can apply the more conservative source tracing approach. The difference is just that you can specify both the source and the target factors.

Source tracing — What are the consequences of one or more factors, looking only at stories told in their entirety by individual sources ?

How to use

🔍 Start path tracing: Click on the Transforms Filters panel and select the 'trace paths' filter.

➡️ Define your path: Specify the starting and ending factors for the path you want to trace. You can leave the end field blank to show all paths leading from a factor, or leave the start field blank to show all paths leading to a factor.

🔢 Set the steps: Choose the number of steps (or links) to include in your traced path. Increasing this number will broaden the results by including longer paths. But keep in mind that, when you’re analysing interviews, people usually don’t report causal chains longer than 4 steps.

The factors which match the filter are shown with thicker borders.

Tips for Success💡:

🗺️ Simplify your map: Consider applying other filters beforehand to format and simplify your map before tracing paths.

⛓️ Avoid the transitivity trap: Be careful when drawing conclusions. The presence of links from A to B and B to C does not automatically mean that all respondents indirectly connect A to C: some may have mentioned only A to B and others only mentioned B to C. To avoid this trap, you can trace individual respondents’ threads within the paths, which filters to show only continuous chains of links from the same source to avoid the transitivity trap issue.

In most cases, we should always trace threads before any filter which changes labels: zooming, removing brackets, combining opposites and autoclustering. See this page for more information.

Remember that order matters: the order in which the filters are applied makes a difference.