Scorecard is a table to compute, for example, credit score of a person. For example, add 10 if age < 30, add 20 if age < 40, add 30 if age <40, add 10 if he/she does not own a house, and add 20 if he/she owns a house. The grand sum of this process is ‘credit score’ which, as you can see in the below, is the logodds of on time loan payback. (logodds is log(p/(1-p)) where p is some probability.)

To get these ‘if’ conditions and the scores, we use logistic regression. The process is 1) select variable, 2) categorize continuous variable and get dummy variables for each category, 3) perform logistic regression, 4) take coefficients of dummy variables as scores, 5) scale scores if necessary.

Scaling is necessary as, otherwise, a score will be like 1.121 while we want to use 10 or 20 or 30 or etc.

Finally, summing up the scores according to one’s condition is actually adding logodds of on time loan payback. And, it’s actually is multiplying the odds, i.e., p/(1-p).

References)
1. http://www.plug-n-score.com/learning/logistic-regression-for-scorecard-calculation.htm.
2. 구자용, 박헌진, 최대우, 김성수, “데이터마이닝”, Knou Press.