How to filter incomplete survey responses without deleting the original data
Apply a completion rule to the fields that matter, preview the effect, and keep the full merged dataset unchanged.
Choose the questions that should count
If you are evaluating Week 8 completion, count the Week 8 questionnaire fields, not demographics, researcher notes, or unrelated columns. Otherwise a participant can look more or less complete simply because the dataset contains extra information.
Use the threshold your study actually needs
There is no universal rule that says 70% or 80% is correct. Your protocol or analysis plan should decide that. The useful part of a completion filter is being able to state the rule clearly and see what it does before you export anything.
Look at the excluded IDs before you commit
A preview should tell you how many rows stay, how many are excluded, and which participants are near the cutoff. If a participant is at 69% when the threshold is 70%, you can verify that the result makes sense before creating the filtered copy.
Keep the unfiltered merge
Do not make the filtered dataset your only copy. If the criterion changes later, or a collaborator asks why a participant disappeared, the full merged result is the reference you will want.
A small example
Selected fields: Week8 Q1–Q10 Threshold: 70% P001 100% keep P003 70% keep P004 60% exclude P006 30% exclude
How RDM keeps the filter reversible
The completion filter creates a separate copy. You choose which columns count, preview the keep/exclude result, and leave the merged dataset untouched.