I create a crosswalk between U.S. and Canadian HS8 product codes using machine learning methods to improve precision in studies documenting the impacts of trade policies and tariff wars. The Harmonized System (HS) provides a standard system for tracking the flow of goods in international trade at various degrees of granularity ranging from two to six digits. To track goods at a finer level of disaggregation, countries have developed their own system of HS8 and HS10 product codes. However, these country-specific HS codes are not directly comparable. Tariff implementation in the U.S. and Canada is based on the HS8 codes. The lack of comparability between the U.S. and Canadian HS8 codes precludes observing which products leaving the U.S. are subject to tariffs when they enter Canada. The crosswalk I create solves this problem. I leverage similarity based on trade values and product descriptions between the two systems to generate matches using competing machine learning models by leveraging 263 product matches from Khushboo (2025) as “ground truth” data. I find that a semi-supervised Self-Training model based on the Random Forest Classifier performs the best in terms of key evaluation metrics. This is the first work documenting cross-country HS8 product codes. Although the motivation for this project comes from the desire to study the effect of tariff wars in Canada, the methods documented can be used to create an HS8 concordance between any two countries.