X (Twitter) analytics reporting for agencies, post by post

X counts five distinct things as engagement, and they say very different things about a post. A like costs nothing. A reply starts a conversation the account then has to staff. A repost (formerly a retweet) puts the post in front of an audience the account does not own; a quote post does the same while adding an opinion to it; and a bookmark is the quietest signal of all — somebody saving the post to come back to it later. The Engagement mix tab stacks all five per post, in volume or as a share of the mix, with a totals row underneath the table.

That breakdown changes what the next content plan should do. A month carried by likes looks identical, in a single headline engagement number, to a month carried by replies and reposts — and only one of those is actually spreading. Reading the stack post by post shows which piece of writing earned the reposts, which one started the conversation, and which one was quietly bookmarked by people who never engaged in public. Engagements are exactly the sum of those five kinds at every level of the report, so nothing is double counted, and the clicks and expands X also tracks are reported separately to keep the headline total reproducible.

The Links and profile tab answers the question a client really asks about organic X: did any of it move somebody off the timeline. Link clicks are reported for every post and then aggregated by destination, so a month resolves into a short list of paths with the clicks each one earned, its share of all clicks, and how that compares with the impressions it took to get them. Beside the table sit the counters that are easiest to lose — profile clicks from inside a post, profile visits from any route at all, detail expands, and hashtag clicks.

Those two profile numbers are not the same number, and the report refuses to merge them. A profile click is somebody tapping the handle inside a post; a profile visit is an arrival on the profile from anywhere, including search and a link somebody else posted. An account can earn a lot of one and very little of the other, and the gap tells an account manager whether the timeline is doing the work or whether people are arriving from somewhere else entirely. Neither is folded into the engagement total, which keeps the headline rate honest and keeps the click story separate from the conversation story.

Video on X is judged on completion, not on plays. The Media and video tab gives every clip a card with its length, its views and a quartile funnel — how many of those plays were still watching at a quarter, a half, three quarters and the end — drawn as labelled bars rather than compressed into one percentage. Average watch time sits beside it in seconds. A ninety-second clip that loses three quarters of its audience before the halfway mark is a different problem from one that holds them to the end and simply did not travel very far, and the two need different fixes.

Media views are reported separately from video views, because an image loads and a video plays and those are not interchangeable events. A photo post reports its media views; a video post reports both, with the plays always a subset of the loads. The table underneath lists every media post in the window with media views, video views, watch-through, average watch time, impressions and engagement rate, so the strongest and the weakest clip of the month sit in the same view — and next month's video brief has evidence behind it instead of a preference.

Text, photo, video and link posts do not earn the same things, and an average that mixes them hides which is which. The Post types tab rolls the window up by shape and reports posts published, impressions, engagements, engagement rate, link clicks and reposts per thousand impressions for each one, per post or as totals. The per-thousand figure is the comparable one: it refuses to reward a type simply for having been shown more often, so a text post that travelled on a small audience is not buried underneath a video the platform pushed harder.

The same shapes drive the report's primary filter, so choosing Text or Video re-bases the scorecards, the trend chart, the engagement mix and every table below them. A shape the month did not use says so with an empty state rather than a fabricated row — a month with no link posts reports no link posts, which is itself a finding when the retainer includes driving traffic. Over a few months this table becomes the fastest argument for changing the publishing mix, because it is built out of the account's own posts rather than a general claim about what performs on the platform.

The Followers and mentions tab reports the follower level day by day, walked backwards through the period's net growth so the figure on any given day is a level rather than a slice of a total, beside the mentions arriving, the replies received on the account's own posts and the median time the account took to answer them. Gains and losses sit above it with the net change, so a month that added four hundred followers while losing three hundred is never presented as a quiet gain of four hundred, and a shrinking month is reported as a shrinking month.

Reply speed is where an organic X retainer quietly fails, so it is reported as a number instead of an impression. Every day carries a median in minutes, a slow day is marked in the table, and the stat band names the slowest day in the window so nobody has to hunt for it. Alongside it, the Timing tab spreads engagements and impressions over each hour of the day and each weekday this account was read, marking the busiest and the quietest slot — X runs earlier than a feed-based network, and a publishing schedule argued from that is easier to defend than one copied off a generic calendar.