X (Twitter) Account Analytics

Turn any X (Twitter) account into a single, decision-ready analytics row. Give it a list of accounts and it reads their recent public posts and returns one report per account: average and median likes, reposts, replies and views, two separate engagement rates, posting frequency, a full day-of-week and hour-of-day breakdown, the content mix, the topics and sites the account links to, and its five best posts by reach and by likes. No X account and no login needed.

This is the report you would otherwise build by hand in a spreadsheet after exporting hundreds of posts — except the arithmetic is done honestly, with the number of posts behind every average stated on the row.

What you can do with it

• Benchmark competitors on one screen. Put twenty accounts in, get twenty rows out, sort by engagement rate and see instantly who is actually working.
• Find the best time to post. Every report carries a full posting breakdown by day of the week and by hour of the day, plus the day and hour where this account's own posts earned the most views.
• Qualify influencers and partners before you pay them. A large follower count with a 0.02% engagement rate tells a very different story from a small account at 3%. Both rates are on the row.
• Track your own account over time. Schedule a weekly run over your handle and watch averages, cadence and content mix move.
• Brief a content team with evidence. The content mix (photo/video share, link share, reply share, long-post share) plus the top hashtags, mentions and linked sites describe an account's strategy in one row.
• Screen a long list fast. Analytics for 200 accounts is one run and one spreadsheet, not 200 profile visits.

What you get

One row per account. This is a real row from a run over OpenAI, abridged:

{
  "inputProfile": "@OpenAI",
  "userId": "4398626122",
  "username": "OpenAI",
  "displayName": "OpenAI",
  "profileUrl": "https://x.com/OpenAI",
  "followers": 5087026,
  "following": 4,
  "postCount": 2048,
  "verifiedType": "Business",
  "joinedAt": "2015-12-06T22:51:08.000Z",
  "avatarUrl": "https://pbs.twimg.com/profile_images/1885410181409820672/ztsaR0JW.jpg",

  "postsAnalysed": 200,
  "ownPostsAnalysed": 180,
  "repostsAnalysed": 20,
  "dateRangeStart": "2026-04-30T19:08:00.000Z",
  "dateRangeEnd": "2026-08-04T21:05:30.000Z",
  "daysCovered": 96.08,
  "postsPerDay": 2.07,
  "postsPerWeek": 14.5,

  "totalLikes": 783571,
  "avgLikes": 4353.17,
  "medianLikes": 1639,
  "maxLikes": 47565,
  "avgReposts": 424.51,
  "avgReplies": 285.09,
  "avgQuotes": 254.63,
  "avgBookmarks": 985.21,
  "totalViews": 285178226,
  "avgViews": 1584323.48,
  "medianViews": 357408,
  "maxViews": 30967000,

  "totalEngagements": 1134470,
  "avgEngagementsPerPost": 6302.61,
  "engagementRateByViewsPct": 0.3978,
  "engagementRateByFollowersPct": 0.1239,
  "postsWithViewCount": 180,
  "postsWithoutViewCount": 0,
  "metricCoverage": {
    "postsAnalysed": 200, "ownPosts": 180,
    "likes": 180, "reposts": 180, "replies": 180,
    "quotes": 180, "bookmarks": 180, "views": 180,
    "engagementRateByViews": 180, "engagementRateByFollowers": 180,
    "contentShares": 180
  },

  "postsByDayOfWeek": {
    "Monday": 25, "Tuesday": 33, "Wednesday": 53, "Thursday": 57,
    "Friday": 29, "Saturday": 3, "Sunday": 0
  },
  "postsByHourUtc": {
    "00:00": 11, "13:00": 8, "16:00": 15, "17:00": 60,
    "18:00": 17, "19:00": 21, "20:00": 33, "21:00": 15
  },
  "mostActiveDayOfWeek": "Thursday",
  "mostActiveHourUtc": 17,
  "avgViewsByDayOfWeek": { "Monday": 1140741.85, "Tuesday": 1840813.03, "Sunday": null },
  "avgViewsByHourUtc": { "13:00": 2557859.57, "17:00": 1625783.81, "20:00": 1785844.52 },
  "bestDayOfWeekByViews": "Tuesday",
  "bestHourUtcByViews": 13,

  "pctWithMedia": 44.44,
  "pctWithLinks": 24.44,
  "pctQuotes": 4.44,
  "pctLongForm": 39.44,
  "pctReplies": 48,
  "pctReposts": 10,

  "topHashtags": [],
  "topMentions": [
    { "value": "AndrewMayne", "count": 5 },
    { "value": "Broadcom", "count": 2 },
    { "value": "Microsoft", "count": 2 }
  ],
  "topDomains": [
    { "value": "openai.com", "count": 26 },
    { "value": "open.spotify.com", "count": 5 },
    { "value": "podcasts.apple.com", "count": 5 }
  ],
  "languageMix": [ { "value": "en", "count": 174 }, { "value": "zxx", "count": 6 } ],

  "topPostsByViews": [
    {
      "postId": "2079658951264920020",
      "url": "https://x.com/OpenAI/status/2079658951264920020",
      "text": "We're partnering with @huggingface to investigate an unprecedented security incident…",
      "createdAt": "2026-07-21T20:05:06.000Z",
      "likes": 20823,
      "views": 30967000
    }
  ],
  "topPostsByLikes": [
    {
      "postId": "2074704958419792299",
      "url": "https://x.com/OpenAI/status/2074704958419792299",
      "text": "GPT-5.6 Sol, along with Terra and Luna, will launch publicly this Thursday. We’re expanding preview access globally now.",
      "createdAt": "2026-07-08T03:59:42.000Z",
      "likes": 47565,
      "views": 9867332
    }
  ],

  "unavailable": false,
  "scrapedAt": "2026-08-05T22:11:52.062Z"
}

How the numbers are worked out

Three rules decide every figure on the row, and all three exist because the obvious shortcut produces confidently wrong analytics.

1. An average is taken only over the posts that actually carry the value. X shows a view count on most recent posts and on fewer older ones. Where no view count is published, this actor leaves the field empty instead of writing a zero — and the empty ones are left out of avgViews and medianViews rather than dragging them down. postsWithViewCount and the metricCoverage block tell you exactly how many posts stood behind each average, so you can judge how much weight the figure deserves.

2. Figures about content come from the account's own posts. When an account reposts somebody else, the likes and replies on that post belong to the original author, and X reports the repost itself with no likes of its own. Mixing those in would pull every average toward zero — on a repost-heavy account it can understate average likes by 40% or more. The same is true of what a post contained: a repost does not carry the photo, video, link or long body of the post it passed on, so treating it as "published without a photo" halves the picture, and on a live account with a third of its feed reposted it did exactly that.

So avgLikes, avgViews, the medians, the engagement rates, the topic lists, the top-post lists and the content shares pctWithMedia, pctWithLinks, pctQuotes and pctLongForm are all built from ownPostsAnalysed — the count is repeated as metricCoverage.contentShares so you never have to guess the denominator. pctReposts and pctReplies are shares of everything analysed, because those two are true or false for every entry, and reposts still count in postsAnalysed and in the posting frequency, because the account really did publish them.

3. A pinned post does not stretch the window. Pinning keeps one post at the top of a profile no matter how old it is. dateRangeStart, dateRangeEnd, daysCovered, postsPerDay and postsPerWeek describe the continuous run of recent posts, so an old pinned post cannot turn a six-week window into a three-year one. It still counts in the totals and averages.

The two engagement rates. Engagements are

likes + reposts + replies + quotes + bookmarks

• engagementRateByViewsPct = 100 × (engagements ÷ views), summed across the account's own posts that carry a view count, with both sides of the division taken from those same posts. This answers "of the people who saw a post, what share interacted".
• engagementRateByFollowersPct = 100 × (avgEngagementsPerPost ÷ followers). This answers "how much interaction does a typical post draw, relative to the audience this account has built". It is the figure most influencer-marketing benchmarks use.

They are deliberately separate numbers and they are not comparable to each other. Both are percentages, so 0.3978 means 0.3978%.

Best day and hour. mostActiveDayOfWeek and mostActiveHourUtc are where the account posts most. bestDayOfWeekByViews and bestHourUtcByViews are where its own posts averaged the most views, and only time slots holding more than one post can win — a single lucky post at 3am is not a posting-time insight. All times are UTC, and the full breakdowns are on the row so you can convert and re-slice them yourself.

Input

Output fields

Pricing

This actor is pay per account report. You are charged once for each report row that lands in your dataset, whatever the size of the account or the depth of the analysis. There is no time-based charge and no charge for the individual posts read to build the report.

An account that turns out to be private, suspended or missing is still returned as a row so your list keeps its shape, and that row is charged the same as any other. A weekly competitive sweep over 25 accounts is 25 rows per run.

Limits & what this actor cannot do

• The report is built from an account's recent public posts, not its whole history. Deeper windows make the averages steadier but always describe recent behaviour rather than the account's lifetime.
• X publishes a limited window of recent public posts per account, so very large accounts return what X makes available rather than a complete history.
• Keyword search across all of X is not part of what X publishes openly. This actor works from the accounts you supply rather than a search term.
• Replies written by other people underneath a post are not publicly available, and neither are the lists of who follows an account. Follower and following counts are included.
• Protected (private) accounts are not publicly viewable. They are reported as unavailable rather than returned empty, and so are suspended, withheld and missing accounts — each with its own reason, so a suspended account is never confused with a handle that does not exist.
• Every engagement number is a snapshot at the moment of collection. A post keeps accumulating likes and views afterwards, so a report on very fresh posts will read lower than the same report a week later.
• X publishes view counts for most recent posts but not for every older one. Where X publishes none, the value is left empty rather than set to zero, and those posts sit outside avgViews, medianViews and the view-based engagement rate. The row tells you how many posts that was.
• Averages over a handful of posts are noisy. postsAnalysed and metricCoverage are on every row precisely so you can see when a figure is thin.
• All day and hour figures are in UTC.
• Speed depends on the size of the job and on X's own response times.
• X's terms govern automated access. You are responsible for using the data lawfully, in line with the source site's terms, and for handling any personal data in line with applicable privacy law.

FAQ

Do I need an X account? No. Nothing is required from you beyond the list of accounts you want reported on.

Does it need my login, password or a paid X subscription? No. None of those are used at any point.

Can I analyse accounts I do not own? Yes — that is the main use. Everything in the report is computed from posts that are already public on X, so competitor and influencer analysis works exactly like analysis of your own account.

Can I schedule it? Yes. Schedule it on Apify to run weekly or monthly over the same list and watch the numbers move.

Is the data complete? The report is complete for the window it covers, and the row states that window and how many posts stood behind every figure. It is not a lifetime history, and posts with no published view count are excluded from the view-based figures rather than counted as zero.

Why is the engagement rate different from the one my other tool shows? Because most tools quietly pick one of two definitions. This actor gives you both — one against views and one against followers — and states the formula, so you can match whichever your team already uses.

How do I get the results into my tools? Export from Apify as CSV, Excel, JSON, XML or an HTML table, push into Google Sheets, or pull them into your own system.