X (Twitter) Metrics and KPIs in 2026: What to Track, Formulas and Benchmarks

The X (Twitter) metrics worth tracking in 2026 are impressions, engagement rate per impression, replies, reposts and quotes, bookmarks, profile visits, link clicks and net follower growth. Pick three or four KPIs that match your goal: reach, conversation, traffic or followers. Review them weekly against your own median, not against other accounts or a single viral post.

That answer is short because the hard part is not the list. The hard part is knowing which number leads to a decision. X puts a view count on every post and gives a full analytics dashboard to Premium subscribers, and X's own ranking code now rewards replies, dwell time and profile clicks far more than likes. If you track the wrong thing, you optimise for the wrong thing. This guide covers the official metric definitions, the formulas, what the latest benchmark studies say, how the right KPIs change by goal, account type, content format and account size, and a step-by-step review method you can run every Monday.

X (Twitter) Metrics and KPIs: The Short Version

A metric is any number X reports. A KPI is the handful of metrics you have decided will judge whether your X account is working. Most accounts need one primary KPI and two or three supporting ones. A founder building an audience might use net follower growth as the primary KPI, with profile visits and replies as supporting metrics. A publisher might use link clicks, with impressions and engagement rate behind it. A support team might use response time, with mentions and replies behind it.

Everything else is diagnostic. Likes, for example, are useful for spotting which topics land, but they carry the least weight in X's published ranking weights, and they rarely map to a business outcome on their own. Treat them as a clue, not a goal.

Key Findings

• Impressions are not unique people. X's developer documentation defines an impression as each time a post appears on a screen, so one person seeing a post twice counts as two, according to the X API metrics reference.
• Private metrics expire after 30 days in the API. Link clicks, profile clicks and engagements are "non-public" and only available for posts created in the last 30 days through the API, per the same reference, so export or log them monthly.
• Replies outweigh likes by a wide margin in X's published weights. The April 2023 ranking weights on GitHub gave a reply 13.5, a reply the author engages with 75, and a like 0.5, according to the open-sourced Heavy Ranker README.
• Premium changes your baseline. Buffer's analysis of 18.8 million posts from about 71,000 X accounts found Premium accounts got about 10 times the median reach per post of regular accounts, in its X Premium review (observational data, not proof of cause).
• Text posts lead on engagement rate. Buffer's State of Social Media Engagement 2026 found median X engagement rates of 3.56% for text, 3.40% for images, 2.96% for video and 2.25% for links.
• Industry-wide engagement per follower is tiny. Rival IQ's 2025 benchmark report, which divides engagements by followers, found X engagement fell 48% year over year, the largest drop of any platform it measured.
• X's audience is narrower than it looks. Pew found 10% of US adults use X daily, in its 2025 social media use report (5,123 adults surveyed in February and March 2025).

What's Changed in 2026

Four changes in the last eighteen months affect how you should read your X numbers.

X published a new ranking system

In August 2026 xAI published the code for the For You feed in the xai-org/x-algorithm repository. The new ranker, a transformer model called Phoenix, predicts the probability of a long list of actions for each viewer: likes, replies, reposts, quotes, shares by DM and copy link, clicks on the post, profile, link and photos, video watch time, dwell time, profile visit time and follows. It also predicts negative actions: "not interested", mute, block, report and "not dwelled". Each prediction is multiplied by a weight and summed. The practical lesson for measurement is that X itself treats profile clicks, dwell and follows as signals, so you should too. A post with modest likes and strong profile visits is doing more for you than the like count suggests.

The 2023 weights still tell you the direction

The newer repository does not publish the exact weights in the same simple form. The older Heavy Ranker README, dated April 5, 2023, did: like 0.5, repost 1.0, reply 13.5, profile click that leads to engagement 12.0, a "good click" into the conversation 11.0, reply that the author engages with 75.0, negative feedback minus 74.0 and report minus 369.0. Those numbers will have changed, but the shape is consistent with the 2026 model: conversation and curiosity count for much more than a tap on the heart, and one block or report can cancel out a lot of positive engagement.

Premium is now part of every benchmark

Buffer's engagement report found Premium and regular accounts moved together until January 2025 and then diverged, with Premium engagement rising and the median for regular accounts falling to 0% in the most recent months it measured. That finding appears in both the 2026 engagement report and the X Premium review. If you compare your numbers with a benchmark, check whether it mixes Premium and free accounts, and compare yourself with your own history after any change in subscription.

Analytics access depends on your plan

Sprout Social's March 2026 guide to X analytics notes that "you must have a subscription to view analytics for multiple posts", and Hootsuite's X analytics guide says the full dashboard is a Premium feature on desktop, while free accounts still see post-level numbers in the app. That makes a simple spreadsheet, or a scheduler with its own analytics and CSV export, more important than it used to be.

Every X Metric, Defined

Here are the metrics X reports, what they mean and where the definition comes from. Where X's own documentation is silent, we say so.

The public versus private split comes from the X API metrics reference, which lists reposts, quotes, likes, replies, impressions and bookmarks as public, and URL clicks, profile clicks and engagements as non-public. It also notes that video views are aggregated across every post that contains the same video, which matters if you reuse a clip.

Impressions

Impressions are the number on every post. X counts each time the post appears on a screen, and the developer documentation is explicit that impressions are not unique. That makes impressions a measure of distribution, not of audience. A post seen three times by your most loyal follower and a post seen once by three strangers both show 3. Use impressions to spot which posts the algorithm pushed, then look at what people did with them.

Engagements and engagement rate

X counts every interaction as an engagement: likes, replies, reposts, quotes, link clicks, profile clicks, media expands, follows from the post and more. Engagement rate in X's native analytics is engagements divided by impressions. Sprout Social and Hootsuite both use this definition in their guides. Benchmark studies vary. Buffer divides likes, reposts and comments by reach or impressions. Rival IQ divides interactions by followers. These produce very different numbers, so never compare a rate from one method with a rate from another.

Replies, reposts, quotes and bookmarks

These are the four "heavy" public signals. Replies matter most because they start the conversations X ranks highly, and because your own reply back adds the biggest positive weight in the 2023 table. Quotes carry more information than plain reposts: someone thought your post was worth commenting on. Bookmarks are the quiet signal that a post was useful enough to keep, and they are common on how-to threads and reference posts.

Profile visits and new follows

Profile visits measure curiosity: someone saw a post and wanted to know who wrote it. The profile-to-follow rate tells you whether your profile closes the deal. If a post drives 400 profile visits and 4 follows, the post did its job and your bio, pinned post or header did not.

Link clicks

Link clicks are the only native metric tied directly to traffic. They are private, and in the API they are only available for posts from the last 30 days. Pair them with UTM tags so you can see what visitors did on your site. Google's Analytics help says you "should always use utm_source, utm_medium, and utm_campaign", and strongly recommends adding utm_id and utm_source_platform too, in its guide to campaign URLs. Our free UTM builder makes the tags in seconds, and the UTM parameters glossary entry explains each field.

The Formulas

Write these into your spreadsheet once and you will never argue about definitions again.

Per-1,000 rates are easier to read than tiny percentages for replies and bookmarks. "Two replies per thousand impressions" is a number your team will remember; "0.196%" is not. The X engagement calculator does the impression-based rate for you, and the engagement rate glossary entry covers the other methods.

What the Benchmarks Say, and How to Use Them

Benchmarks are useful for one thing: telling you whether your own numbers are in a sensible range. They are not targets. Here is what the current studies report, with the method each uses.

Notice how far apart these are. Buffer's 2.83% is engagements per impression from accounts that publish through a scheduler. Rival IQ's figures are interactions per follower from established brands, which is why they are much lower. Buffer's own report warns that its jump in X engagement probably reflects changes in who is posting and how metrics are defined, not better performance. The lesson: pick one method, stick to it, and measure yourself against your own median from the previous 30 or 90 days.

The Right KPIs by Goal

The first breakdown is the most important one: what are you trying to get from X? The same post can be a success for one goal and a failure for another.

Goal 1: Awareness and reach

Primary KPI: impressions per post, measured as a median rather than a total so one viral post does not distort the month. Supporting: amplification rate (reposts and quotes per 1,000 impressions) and the share of impressions from non-followers if your analytics shows it. Reach comes from being shown to people who do not follow you yet, and in the 2026 code out-of-network posts are discounted relative to in-network ones, so amplification by people who do follow you is the main way to get out.

Goal 2: Conversation and community

Primary KPI: replies per 1,000 impressions. Supporting: quotes, and your own reply rate (the share of replies you answered within a day). This goal maps most directly to X's ranking signals. If you run an engaged founder account or a niche community, this is the KPI that grows everything else.

Goal 3: Website traffic and leads

Primary KPI: link clicks, and then sessions and conversions in your web analytics using UTM tags. Supporting: link CTR and the engagement rate on link posts. Expect link posts to have lower engagement: Buffer found links had the lowest median engagement of any format, at 2.25%. Many accounts get better results by putting the link in a reply or by alternating link posts with native posts that build the audience. Our guide on X posts that drive traffic covers the formats in detail.

Goal 4: Follower growth

Primary KPI: net new followers per week. Supporting: profile-visit rate and profile-to-follow rate. If profile visits are healthy but follows are not, fix the profile: bio, header, pinned post. If profile visits are weak, the posts are not making people curious. Our guide on how to get more X followers covers the tactics.

Goal 5: Customer care

Primary KPI: median first response time. Supporting: mentions handled and the share resolved in public versus moved to DM. Sprout Social's statistics roundup reports that 18% of users consider X a top customer service destination, citing its 2026 content strategy report, in its X statistics page. If you serve customers on X, response time matters more than any engagement rate.

KPIs by Account Type

The second breakdown is by who is posting. A personal account, a B2B brand and a consumer brand use X for very different jobs.

Founders and personal brands

For a founder, X is a reputation and network channel. Track profile visits, follows and replies, and the number of meaningful DMs or inbound conversations a month (log these by hand). Impressions matter less than who is replying. Twenty replies from people in your market beat two thousand impressions from a trending topic you will never sell to.

B2B companies

B2B brands should track link clicks to content, demo or newsletter pages, plus profile visits and quote posts from people in the industry. Rival IQ's benchmark noted that Financial Services doubled its X engagement rate while nearly doubling post frequency, while Media posted heavily without higher engagement. Volume helps some industries and not others, which is why you should test frequency on your own account. If you split your time between X and LinkedIn, our comparison of X and LinkedIn for B2B and our LinkedIn metrics guide use the same KPI framework so the numbers line up.

Consumer brands and ecommerce

Consumer brands usually care about reach, replies and link clicks to product pages. Watch the format mix closely: images and short video tend to earn more engagement than bare links. Sprout's roundup reports that 37% of users say they are most likely to engage with short-form brand video, and 36% with text posts, citing its 2026 content strategy report.

Publishers and media

Publishers live and die by link clicks and repost velocity in the first hours after a story breaks. Track clicks per post and the time from publish to peak impressions. Pew's 2025 study of news influencers found 82% of the 500 news influencers it tracked still had X accounts in early 2025, and 83% of those posted at least four days a week, so news competition on X is intense.

Audience age matters for every account type

Before you set targets, check who is actually on X. Pew's social media fact sheet, based on 5,022 US adults surveyed between February and June 2025, found 33% of adults aged 18 to 29 use X, compared with 25% of 30 to 49 year olds, 16% of 50 to 64 year olds and 10% of those 65 and over. If your buyers skew older, a modest impression count may still reach most of the people who matter, and link clicks or replies from the right people are a better KPI than raw reach.

Creators

Creators who earn from X need to track the metrics the payout program uses. Under the Original Content Rewards program that replaced Creator Revenue Sharing in September 2026, the key number is Home Timeline impressions from verified users on your original posts, with replies excluded. Our guide to how much X pays creators explains the eligibility thresholds and what counts.

KPIs by Content Format

The third breakdown is by what you post. Each format has a different "job", so judge it by a different number.

Text posts

Text posts had the highest median engagement rate in Buffer's 2026 data, and in its Premium analysis Premium text posts reached nearly 0.9% median engagement by mid-2025. Judge text posts by replies and quotes, since those are what text is good at provoking.

Image posts

Images earn bookmarks and reposts when they carry information: a chart, a checklist, a screenshot with a point. Make sure they are sized so the preview is not cropped awkwardly. Our guide to images on X covers sizes, and the X image resizer fixes them in the browser.

Video

The X API reports playback quartiles: the number of unique users who played through 0%, 25%, 50%, 75% and 100% of a video. That is the retention curve, and it is far more useful than the view count. If most viewers drop before 25%, the opening is the problem. If they drop at 75%, the video is too long for its payoff.

Link posts

Link posts consistently show the lowest engagement. In Buffer's Premium analysis, links for regular accounts sat at 0% median engagement from March 2025. That does not make links useless. It means you should judge them by clicks and on-site conversions, not engagement rate, and you should balance them with native posts.

KPIs by Account Size

The fourth breakdown is size. What "good" looks like at 300 followers is very different from what it looks like at 300,000.

Under 1,000 followers

At this size, most of your impressions come from replies you post on other people's threads and from the small boost new authors can get. The 2026 code includes a "new-author boost" that lifts posts from authors with few impressions toward a target position. Track replies sent, replies received, profile visits and follows. Ignore engagement rate: with 200 impressions, one extra like swings it by half a percentage point.

1,000 to 10,000 followers

Now your own posts start to reach people reliably. Track median impressions per post, replies per 1,000 impressions and profile-to-follow rate. This is the stage where a consistent schedule shows up in the numbers, because the algorithm has enough history to learn who your posts are for.

10,000 to 100,000 followers

At this size, the averages stabilise and you can run proper tests: one variable at a time, two weeks per test, judged against your trailing median. Track amplification rate and link CTR closely. Watch negative signals too: if your follower count stalls while impressions grow, you may be reaching people who mute or ignore you.

Over 100,000 followers

Large accounts should segment their metrics by content pillar and by format, and watch quote-post sentiment. At this scale a post's reach can be driven by people arguing with it. Track replies and quotes, but read a sample of them before you call a post a success.

Where to Find Your X Metrics

You have four places to get your numbers. Most teams use two of them.

The post itself

Every post shows views publicly. On your own posts you can open post analytics from the post to see impressions, engagements, profile visits, link clicks and follows for that single post. Hootsuite's guide notes that free accounts still see this post-level view in the mobile app.

The analytics dashboard

Sprout's guide says the account-level dashboard lives at x.com/i/account_analytics and is reached from the Premium menu on desktop, with a Content tab for individual posts. Access depends on your subscription, account type and region, and both Sprout and Hootsuite note it may change.

The API

If you have developer access, the API returns public metrics for any post and private metrics for your own posts from the last 30 days. That window is the reason to export monthly. If you wait three months, the link clicks and profile clicks for the first month are gone from the API.

Your scheduler

Scheduling tools pull post metrics into their own reports. In Postprism, the analytics view shows how each post performed next to your calendar, and you can export to CSV for your spreadsheet or deck. Our roundup of social media analytics tools compares options if you need deeper reporting.

A 15-Minute Weekly X Review: Step by Step

This is the method we recommend. It works with or without Premium, and it turns numbers into next week's posts.

1. Choose your primary KPI and write it down. One goal, one KPI, from the goal table above. Add two supporting metrics. Put them at the top of your sheet so nobody drifts back to likes.
2. Set up the log. One row per post with date, time, format (text, image, video, link), content pillar, impressions, likes, replies, reposts, quotes, bookmarks, profile visits, link clicks and follows. Add formula columns for the rates in the formulas table.
3. Collect the week's numbers every Monday. Open post analytics or export from your scheduler. Do it weekly so you never fall outside the 30-day window for private metrics.
4. Calculate your trailing medians. Median impressions, median engagement rate and median link CTR across the last 30 posts. Medians resist viral outliers; averages do not.
5. Flag winners and losers. Mark any post that beat the median on your primary KPI by 50% or more as a winner, and anything below half the median as a loser. Most posts will be neither, and that is fine.
6. Read the winners. For each winner, note the format, the pillar, the first line and whether you replied to commenters. Look for a pattern across two or more winners before drawing conclusions.
7. Read the replies on your best and worst posts. Numbers tell you what happened; replies tell you why. Two minutes reading replies is the most useful part of the review.
8. Pick one change for next week. One variable only: more text posts, a different posting time, a stronger first line, a reply in the first hour. If you change three things, you learn nothing.
9. Schedule next week's posts with the change built in. Put the week's posts on your content calendar so the test actually runs. Buffer's analysis of about 8.7 million posts found weekday mornings from 9 to 11 a.m. were the most reliable window, with Tuesday at 9 a.m. the single best slot, in its best time to post on X study, while Sprout's study of about 2 billion engagements points to Tuesday to Thursday, 12 to 6 p.m., in its 2026 X timing analysis. Test both against your own audience. Our best time to post on X guide goes deeper.
10. Roll up monthly. At month end, total your primary KPI, compare it with last month and write two sentences: what worked and what you will change. That is your monthly report.

A worked example

Consider a hypothetical B2B analytics startup, Northwind Data, posting 10 times a week from the founder's account. The goal is newsletter sign-ups, so the primary KPI is link clicks, with profile-visit rate and replies per 1,000 as supporting metrics.

In week one, the trailing medians are 1,900 impressions, 1.4% engagement rate and 0.5% link CTR. Two posts beat the click median by more than 50%: both were short text posts sharing one chart, with the newsletter link in the first reply. Three link-first posts landed below half the median. The one change for week two: move links into the first reply on four posts and keep them in the body on four, so the two approaches can be compared fairly. After two weeks, link clicks per post are 40% higher for the reply versions, so that becomes the default. Every number in this example is invented to show the method; your own results will differ.

Common Mistakes With X Metrics (and the Fix)

1. Treating impressions as people. Impressions are not unique. Fix: describe them as "times shown", and use follows and profile visits when you mean people.
2. Judging every post by likes. Likes carry the least weight in X's published ranking weights. Fix: make replies, profile visits or link clicks your primary KPI, depending on your goal.
3. Mixing engagement rate methods. Rate by impressions and rate by followers can differ several times over. Fix: pick one method, label it in your sheet and only compare like with like.
4. Comparing yourself with industry averages. Benchmarks mix Premium and free accounts, big and small brands, and different methods. Fix: compare with your own trailing median.
5. Letting private metrics expire. The API only returns link clicks and profile clicks for posts from the last 30 days. Fix: log weekly or export monthly from your scheduler.
6. Not tagging links. X link clicks stop at the click. Fix: add UTM tags to every link so you can see sessions and conversions.
7. Calling a test after one post. One post is noise. Fix: run each change for at least two weeks and compare medians.
8. Changing several things at once. Fix: one variable per test, written down before the week starts.
9. Ignoring negative signals. Mutes, blocks and reports carry heavy negative weights. Fix: read replies on high-reach posts, and watch for follower growth stalling while impressions rise.
10. Forgetting to reply. In the 2023 weights, a reply the author engages with scored 75, the largest positive weight in the table. Fix: block 15 minutes after each post to answer replies.
11. Judging link posts by engagement. Links have the lowest median engagement. Fix: judge them by clicks and conversions.
12. Posting inconsistently, then blaming the algorithm. Gaps make your numbers impossible to read. Fix: schedule a steady rhythm in advance so the data reflects content, not chaos. Our guide to staying consistent has a simple system.

How to Track X Metrics in Postprism

Postprism is a scheduler first, so the measurement workflow starts with the calendar.

• Plan the test on the calendar. Drag next week's X posts into the slots you want to test, in month, week or day view, on the content calendar.
• Write the X version once. Postprism publishes single X posts with up to four images. If the same idea goes to LinkedIn or Bluesky, the editor keeps a separate version for each network, with optional AI adaptation to match each platform's style. The X scheduler page shows the details.
• Load a month at once. Use bulk scheduling to upload a CSV with content, scheduled_at and optional media_url columns, check the preview table and schedule the lot.
• Review and export. Check how posts performed in analytics and export to CSV to drop into the weekly sheet described above.

Frequently Asked Questions

What are the most important X (Twitter) metrics?

For most accounts: impressions, engagement rate per impression, replies, reposts and quotes, bookmarks, profile visits, link clicks and net follower growth. Choose one primary KPI based on your goal (reach, conversation, traffic or followers) and two supporting metrics, and review them weekly.

How do you calculate engagement rate on X?

X's analytics divides engagements by impressions and multiplies by 100. Some benchmark studies divide interactions by followers instead, which gives much lower numbers. Use one method consistently and compare against your own median.

What is a good engagement rate on X?

It depends on the method. Buffer's 2026 report found a median of about 2.8% per impression for 2025, with text posts at 3.56%. Rival IQ's follower-based method gives far lower figures. The best benchmark is your own trailing median over the last 30 posts.

Do I need X Premium to see analytics?

You can see post-level numbers for your own posts without Premium. Sprout Social and Hootsuite both report that the full account dashboard requires a subscription. A scheduler with analytics and CSV export, or the API, are alternatives for tracking over time.

Are X views the same as impressions?

Yes. The view count on a post is its impression count. X's developer documentation defines impressions as the number of times a post appears on a screen, and they are not unique, so one person can add several.

Which X metric matters most for the algorithm?

X's published ranking code predicts many actions, including replies, reposts, profile clicks, dwell time and follows, plus negative actions like mutes and reports. In the 2023 weights, replies and replies the author engaged with carried far more weight than likes. Conversation is the strongest signal you can influence directly.

How often should I review my X metrics?

Weekly for decisions, monthly for reporting. A weekly 15-minute review keeps you inside the 30-day window for private metrics and gives you enough posts to spot patterns without overreacting to single posts.

How long should I run a test on X?

At least two weeks, with one variable changed. Compare the median of the test posts with your trailing median, not with your best post. For small accounts, run tests for four weeks because the numbers are noisier.