This is one of the more correct ways to go. The methodology of the article is customer surveys and that is unreliable. There is also an ODB2 manufacturer that

This is one of the more correct ways to go. The methodology of the article is customer surveys and that is unreliable. There is also an ODB2 manufacturer that collects stats about ODB2 error code frequencies per model and manufacturer (they have an API their hardware uses for up-to-date info), and they correlated that to repair costs per error code to estimate the reliability and maintenance costs of vehicles. I can't remember who it was, though.