How can we ease the confusion of engineers when comparing data from Grafana against Kibana?

Spawn from gitlab-org/gitlab#440723 (comment 1795781565).

Other engineers are often confused when comparing data from Grafana against Kibana. Many times the expectation is that Grafana will show exact numbers instead of samples and approximations.

How can we approach this to align expectations with the users or make it easier for them?

Should we educate them better? I've been doing that every time the question surfaces. Though it's efficient on 1-on-1, I wouldn't say it scales well.