AI Briefing
KO

Kurly Freezes Address Refinement 1.0 Based on MOIS API Due to Stability Issues

·2025.01.20 00:00

Key point

Kurly operated Address Refinement 1.0 using the Ministry of the Interior and Safety (MOIS) API, but decided to freeze the system and redesign it due to API response delays and coverage limitations.

Details

Kurly designed Address Refinement 1.0 using the MOIS Open API with the goal of reducing external vendor API call costs, and landed it in the production environment in August 2023. The initial goal was cost reduction rather than terminating the external vendor contract, and the logic was structured considering the unlimited traffic of the road name address lookup API and the limited usage of the coordinate API.

Initial Design and Results

Kurly cached about 5 years of accumulated address refinement data from the external vendor and utilized it. Through this, address refinement for single buildings was prioritized, and after landing in the production environment, the company achieved a reduction in external vendor API calls to less than half of the previous volume. During development, stability was verified through canary deployment and monitoring.

Risks Encountered During Operation

During operation, the MOIS road name address lookup API experienced 500 Bad Gateway errors or timeouts for 5-10 minutes, resulting in about 50 errors. In response, a hotfix was applied to conservatively adjust timeouts and fall back to the external vendor's logic, but the fundamental stability issues of the MOIS API persisted.

Sober Assessment and Future Plans

While there were cost-saving effects, it was determined that terminating the external vendor contract was not realistically feasible due to the risk of response delays from the MOIS API and limitations in new address coverage. The increase in management points to two—the external vendor and the MOIS API—also became a burden. Ultimately, Kurly decided to freeze the current system and explore new address refinement methods.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.