How do examiners reach joint decisions when they grade oral examinations? While government and policymakers provide general frameworks about grading decisions, we know little about how they are actually accomplished in interaction, particularly when examiners initially disagree. We scrutinized 29 video-recorded grading conversations between secondary school examiners using conversation analysis. Results showed that proposing and deciding grades involved a stepwise calibration through which examiners adjusted their individual positions. While in most cases examiners expressed and sought agreement, we also investigate cases where examiners initially disagreed but eventually reached a joint decision. The paper contributes insights into decision-making in institutional interaction, as well as to our understanding of whether and how guidance is implemented in 'live' assessment situations. Our findings suggest adjustments are needed both to practice and assessment policy. Data are in Norwegian with English translation.