Totally depends on the effect size and the threshold at which you're evaluating statistical significance. If their churn were 10x higher than Google's, 20 employees over 10 years would probably be enough to confirm the difference at a high level of confidence. But if their "true" long-run churn is "only", say, 10% higher than FAANG and friends, you would need a much larger sample to identify that.
N.B.: Even a 10% increase in employee churn is expensive in general, and even more so when your recruiters make as much as your senior engineers.
N.B.: Even a 10% increase in employee churn is expensive in general, and even more so when your recruiters make as much as your senior engineers.