A client asked how the account had done over the previous quarter, and I sent back the five best posts. The list was accurate and close to useless, because five winners out of two hundred published posts describe a highlight reel rather than a quarter.
What she wanted was the record. Everything we published, and what came back from each of it.
A ranking cannot produce that. It answers a different question, and it answers it so confidently that you stop noticing the question you actually asked.
How do you see your interacted posts history on X?
Put every post into one row of a table with a column for each type of response, bound that table to a date range, and read the whole period instead of the top of a sort. Circleboom lists the posts your X account can still return with likes, reposts, replies, bookmarks, profile clicks and link clicks beside each one.
→ see your interacted posts history on X
How I see my interacted posts history on X, quarter by quarter
On screen: what a per-post impressions and engagements read looks like once every response type sits on the same row.
https://www.youtube.com/watch?v=fqMJIGZg-ug
The pass I run now, step by step.
Connect the account that holds the record
- Log in to Circleboom Twitter, authorizing whichever X account holds the period under review.
- Open X Post Planner from the menu, then pick post analytics. One published post fills one row; one response type fills one column.
- Set the date range to the stretch you are actually reporting on, so the table holds a bounded period instead of everything at once.
Read the period before you rank inside it
- Sort by the column that matches what those posts were for. Replies for conversation, link clicks for traffic, reposts for distribution, bookmarks for saved value.
- Scroll to the bottom of that sort and read the posts that came back with almost nothing, because that end of the list is where a diagnosis is possible.
- Open any row worth acting on and reschedule it, queue it for a repost cycle, or rewrite it with AI before it goes back out.
The order carries the whole thing. The date range decides which posts are being compared against each other, and the sort only decides how they stack up inside that comparison. Rank before you bound and you are quietly measuring a launch week against a quiet August.
Short version: connect, bound the period, sort by intent, read the bottom, act from the row.
Why a ranking is not a record
A sort is a claim about order. A record is a claim about coverage.
Both are useful, and they are not interchangeable. When someone asks what the account did, a leaderboard hands back the five posts that happened to travel furthest, and it hands them back stripped of everything they happened next to.
You lose the shape of the period that way. A quarter where nine posts out of ten landed flat and one caught fire reads identically to a quarter where every post did respectable numbers, because both quarters produce a top five.
The fix is not a better sort. The fix is bounding the set before you rank it, so the ranking describes something you chose rather than something the algorithm handed you.
That is also the difference between reviewing content and planning it. If you want the ranking to feed the next month of posting, using past post analytics to decide what to tweet picks up exactly where the ranking stops.
Once you have the period in front of you, the useful move is to stop treating it as a contest. Pull one month of your interacted posts history for a stretch you remember well, and check whether your memory of that month matches its rows.
What the rows with nothing in them are telling you
The posts that returned almost nothing are the only rows in the record that can tell you a post missed its job.
That sounds backwards until you try to diagnose something from the top of a list. A winner tells you it won. It does not tell you why the other forty posts did not, and it cannot, because it has no counterexample attached to it.
The bottom of the sort does. A post built to send traffic that collected impressions and almost no link clicks is a diagnosis, not a disappointment. So is a question post that gathered likes and no replies.
In both cases the number is not low across the board. It is low in exactly the column the post was written for.
A post fails specifically, and only the record shows you where.
That is the reading that changed how I run a review. I stopped asking which posts did well and started asking which posts did the thing they were written to do, and the second question needs the whole period on screen.
The posts that never do their job in any column are a separate decision, and eventually a cleanup one. Deleting low engagement tweets from mobile covers that call when the pattern holds across a year rather than a week.
Two bounded periods beat one leaderboard
A record you can bound is a record you can compare against another one.
That is the move a ranking cannot make. All-time top posts are a single list, and a single list has no baseline, so you can stare at it for an hour without learning whether the account is getting better.
Two date ranges of similar length fix that in about a minute. Read March, then read the March before it, and the difference is legible without any statistics.
The arithmetic stays simple on purpose. If you published 40 posts last quarter and 12 of them cleared 50 replies, that is 30 percent of the quarter doing conversational work. Run the same count on the previous quarter and you have a number that means something, because both halves were built the same way.
Totals will not give you that. A total moves when your posting volume moves, so a busy quarter beats a quiet one on almost every column regardless of whether the content improved.
Rates and shares survive that problem. Counts do not.
Does interacted posts history mean your posts or the ones you liked?
It means both, depending on who is asking, and the two answers live in completely different places.
If you mean the posts you interacted with, that is your likes tab, your bookmarks and your reply history, and X keeps that as content rather than as measurement.
If you mean the posts that were interacted with, meaning yours, that is a performance record with a number in every cell.
This article is about the second one, because it is the harder of the two to assemble and the more useful of the two to own. The first one is browsable. The second one has to be built.
Both readings are legitimate searches, which is why the phrase gets used for opposite things in the same thread. When it is the content side you want, searching your Twitter history is the route that answers it.
What X hands you natively, and where that stops
X does give you per-post numbers, and for a recent stretch they are enough.
The constraint shows up when you want a period rather than a moment. X's own Post and Video Activity Dashboards documentation puts a ceiling on the export. You can pull up to 30 days of data at a time, with a cap of 3,000 posts per CSV file.
Thirty days is a fine unit for a monthly check-in. It is an awkward unit for a year. A year of history assembled that way is at least a dozen separate pulls that you then have to stitch together yourself, in a spreadsheet, without introducing an error.
That is the real reason most accounts never look at their own record. Not disinterest. Assembly cost.
The native path is worth understanding before you rule it out. How to see tweet analytics covers what X displays, and which screen each figure lives on.
What the record is good for once you have it
The record earns its keep in three places, and none of them is the leaderboard.
It grounds a plan, because a period of posts with their response columns attached tells you which formats your audience actually answers rather than which one got lucky.
It grounds a report, because a client question about a quarter deserves a quarter-shaped answer.
And it grounds reuse, since the posts worth putting back into circulation are visible as a set rather than as a single obvious winner.
Circleboom is an official X Enterprise Developer company. The numbers in that table are the ones X reports for your account, not a reconstruction of whatever happens to be publicly visible. On a reporting question that distinction is the whole ballgame.
The action layer sits on the same row as the number. A post you decide to reuse can be rescheduled, queued for a repost cycle, or sent through rewrite old tweets with AI without opening anything else. Any rewrite stays a draft until you publish or schedule it deliberately.
The one column that flatters a weak post
Engagement rate is the exception to everything above, and it needs a warning label.
The calculation divides engagements by impressions, which means a post that reached almost nobody can top the column outright. The rate is arithmetically correct and it describes twelve people.
Pair the percentage with the raw counts on the same row before you quote either. For the wider metric set, tweet activity and the metrics behind it is the reference I keep going back to.
When you are ready to run this on a real period, open your X interaction record and start with the last full month.
Where the record ends
Standard account access reaches back across your 3,200 newest posts and no further, which gives the record a back edge that keeps moving.
Every post you publish pushes an older one past it. For an account posting twice a week this is not a constraint you will meet for years. For a brand account running replies at volume, the oldest reachable row may sit inside the last few months.
Neither case is a defect. It just means the phrase "all time" describes a window, and it is worth knowing where yours falls before you quote a total that sounds permanent. Twitter post analytics is the hub the per-post table sits inside, alongside the other reporting views.
The part of all this worth keeping
Five best posts is a highlight reel. Two hundred rows with their response columns attached is a quarter, and only one of those can be reviewed.
Bound the period, sort by the intent the posts were written for, then read upward from the bottom, where the misses explain themselves. The winners were never the informative part.
→ Run it on your last full month: see your interacted posts history on X
Questions about reading your X interaction history
Can I keep a copy of the record outside the tool?
Yes. Post-level data can be exported for campaign reports, and on a busy account the habit earns its keep quickly, since publishing steadily pushes the oldest reachable posts out of range. A snapshot taken while a campaign still sits inside the window answers questions you will only think to ask a year later.
How long should a reporting period be before the numbers mean anything?
Long enough that one strong post cannot carry it. A single week on most accounts is dominated by whichever post happened to travel, so the shape you read is luck rather than pattern. A month is usually the shortest period where format and topic differences start showing up as something repeatable.
What do I do with a period where nothing performed?
Read it by column rather than by total. A flat period where link clicks held steady and replies collapsed is a different problem from one where reach fell across everything, and the two need opposite responses. Flat totals hide that split, which is exactly why the per-post record beats a summary number.
