Propagating incremental changes and maintaining traceability are challenges for interactive model transformations, i.e. ones that combine automation with user decisions. After evolutionary changes to the source models the transformations have to be rerun. Earlier decisions cannot be used directly, because they may have been affected by the changes. Re-doing or verifying each decision manually is error-prone and burdensome. We present a way to model user interaction for transformations that are well (but not fully) understood. We model each decision as a set of options and their consequences. Also, we model the decision context, i.e. the circumstances (including model elements) affecting the decision. When a transformation is run, user decisions and their context are recorded. After a model change, a decision can be safely reused without burdening the user, if its context has not changed. The context maps source model elements to a decision, and thus provides traceability across the de...