I simply did percent_dead times num_dead/percent_cases times num_cases. That leads to ~0.5%. I agree with you that IFR is unknowable without very thorough sero-survey. So all we have is CFR to go on. The ratio depends on population age and test availability etc.