Akindi uses the point-biserial correlation coefficient (rpb) to calculate a discriminatory score for each question in your assessment, and for your assessment as a whole.
The discriminatory score shows the correlation between students correctly answering each question and performing well on the assessment as a whole; in other words, it shows how well each question discriminates between students who did well on the assessment and students who did poorly.
Akindi considers questions with a discriminatory score above 0.20 to be well-constructed: high-performing students generally answer them correctly, and low-performing students generally answer them incorrectly. Give yourself a pat on the back for these questions: they are probably doing a good job of accurately assessing your students.
Questions with a discriminatory score between 0.00 and 0.20 are considered candidates for improvement: high and low performing students answer correctly with approximately the same frequency. It might be worth double checking these questions to see whether they could be clarified, or ambiguity could be removed.
Questions with a discriminatory score below 0.00 are considered problematic: the highest performing students are getting them wrong and the lowest performing students are getting them right! It is definitely worth reviewing these questions to see if they can be clarified, or (more frequently) the answer key was entered incorrectly (Akindi will even tell you which of the "incorrect" answers were chosen most frequently).
The assessment's overall discriminatory score is the average score of the questions in the assignment, excluding questions with a weight of 0.
When calculating the discriminatory score, questions weights are not considered except for zero-weighted questions which are not considered when calculating the overall assessment discriminatory score.
GIF: Best constructed questions and questions which could be improved breakdown.