What belongs in an engagement rate report, and what to split
An engagement rate report is easy to produce and easy to make meaningless. Split interactions into likes, reposts and replies, and set the rate against reach.

An engagement rate report is easy to produce and easy to make meaningless. The difference is whether it splits the number or just states it.
What the report has to answer
Not "how much engagement did we get". That number alone cannot be judged. The three questions worth answering: did engagement rise or fall against the previous period, which posts and formats produced it, and did it move in the same direction as reach.
That last comparison is the one that catches problems. Engagement falling while reach rises means the content is being served to people who do not care.
Tell us what the report has to support and we will build it around that.
Tell us what the report has to support
The split that does the work
Total interactions tells you almost nothing. Broken into likes, reposts and replies, it starts describing behaviour. Likes are cheap approval. Reposts are endorsement, because the person attached their name to it. Replies are involvement, and sometimes they are complaints.
A month where replies jumped and reposts fell is a very different month from the reverse, and the total hides both.
Rate against volume
Report the absolute numbers and the rate, and say which is which. Interactions divided by impressions gives the rate. It rises when reach falls, so a campaign that reached far fewer people can post a better rate and a worse result. Showing both prevents that reading.
Apply a floor before ranking posts by rate, or the top of the table fills with posts seen thirty times.
Per format, per time, per author
Three cuts cover most of what anyone asks afterwards. By format: images, links, video, plain text, replies against originals. By time: hour of day and day of week. By author, when the report covers more than your own account.
In each one, put the share of volume next to the share of engagement. The mismatches are the findings.
What to do with the top posts
Take the ten with the most engagement and the ten with the least, and write one line on what separates them. That line is the only part of the report that changes behaviour. Everything above it is evidence for it.
Keeping it comparable
State the period, the query and whether reposts are included. An engagement report that changes its definition between months produces a trend line that means nothing. We deliver the report with the dataset behind it, gathered within the Twitter compliance rules, so any figure can be checked against the posts. Tell us the account or campaign and we will scope it.