Start from the tracking specification and test one representative journey at a time. Confirm event names, timestamps, user identity, property types, and expected absence as carefully as expected presence. Compare raw events with the product state, record evidence, and retest after fixes. Add automated schema checks where practical, but retain a short human journey test before major releases.
Put the idea to work
Write down the decision this analysis should change, the evidence that would challenge your current belief, and the person responsible for acting on the result. These three details keep exploratory work connected to a useful outcome.
When definitions change, record why and from which date. Consistency is valuable, but transparent evolution is more trustworthy than preserving a metric that no longer represents the product.