Nothing in a dashboard is lying. The misreading happens between the number and the conclusion, and it happens the same three ways every time.
One: reading noise as a trend
A post did three times better than usual. That is one observation. Two more like it is a pattern; one is weather. Most content strategy changes are made on a single data point in a mood.
Two: comparing periods that are not comparable
This month against last month, when last month had a holiday, a launch and eleven posts against this month's six. The comparison needs the same length, the same posting rate and no unusual event, or it is comparing two different questions.
Three: measuring what is easy
Impressions are always available, so they get reported. They count the same person twice and reward nothing. The measures that matter — saves, shares, follows from a post — take one more click, which is why they show up in fewer reports.
Before opening the dashboard, write down the question. Numbers looked at without a question produce a story, and the story is always about whatever happened to be highest.