APPENDIX E: EVALUATION TEMPLATES

A SIMPLE ROUTINE FOR EVALUATION-DRIVEN IMPROVEMENT Run the whole dataset and save the report. Pick the most common failure and find its cause. Make one change that targets that cause. Run the whole dataset again and compare the new report with the old one. Keep the change only if the results improved and nothing else got worse. Add a new test item for every real failure you meet so it can never quietly come back.