THE IMPLEMENTATION OF LOGISTIC REGRESSION TO DEVELOP VENDOR RATING MODEL

• Main Author: Selviana
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/60942
• Institution: Institut Teknologi Bandung

A vendor rating model is an essential tool to rate vendor performance for successful vendor management. This research is conducted in an Indonesian leading palm oil company facing the problem of assessing the vendors' timeliness of delivery. In the past, the company had experiences of giving prepayment to several vendors and ended up the goods were delivered not in a timely basis; furthermore, some could not fulfil the agreement and went default. To address that issue, this vendor rating model is developed to help the company further assess the vendors’ timeliness of delivery based on the vendors’ scores for risk management purposes. This model is constructed by modifying the credit risk scorecard using logistic regression based on the company's historical vendors’ performance data and vendors’ profiles. Logistic regression is selected because of its excellent characteristics (e.g., robustness and transparency). There are 13 independent variables in this research, which are transaction value, transaction volume, transaction frequency, transaction existence per month, age of the company, age of parent company, digital existence, vendor relationship length of time, type of company, long-term contract agreement existence, RSPO (Roundtable on Sustainable Palm Oil) certification status of the company, RSPO certification status of the parent company, and foreign ownership of the company while the dependent variable is the timeliness of delivery. The results identified four variables that could be considered and used as not timely delivery probability predictors: age of parent company, digital existence, type of company, and long-term contract agreement existence. The model is tested using a cross-validation test, with the result as follows: the optimum is occurred in the fitted cut-off value of 0.53 with the accuracy rate, sensitivity rate, and misclassification rate, respectively, as for 100%, 70%, and 17.7%. This is the first research of developing vendor rating model by using credit risk scorecard and logistic regression to the researcher's best knowledge.