X shows everyone a view count on your videos, and it keeps the stage-by-stage playback numbers private to your account. Seeing those private numbers is what tells you where viewers stop watching.
On X alone: one public view count per post, with deeper rates spread across X's own dashboards and Media Studio. With Circleboom: Circleboom charts Video Started, 25%, 50%, 75% and 100% Watched for every native video on X through the official X Enterprise API, in one Video Analytics view.
→ Twitter video analytics
Two layers of video data, and only one of them shows where people leave.
What Twitter video analytics covers on X
Twitter video analytics is the set of numbers that describe how a video on X was shown and how far it was played. On X, those numbers sit in two layers with different owners and different audiences.
The public layer belongs to everyone who sees the post. The private layer belongs to the account that posted the video, and it holds the playback stages.
Good X video analytics work reads both layers in order: reach first, then holding power. Reading only one of them leads to the two classic mistakes, celebrating a clip nobody finished or abandoning a clip that simply was not shown to enough people.
What X shows about your videos by default
By default, X shows a public view count on each video post, and that number is the whole public layer. Anyone can see it, including your competitors, and it is the figure most people mean when they say a video "did well."
The view count has one property that surprises many creators. X's API documentation notes that video views are aggregated across all posts that contain the same video, so a clip reposted in several posts carries one total view count.
A view number under a post is therefore a property of the video file, not only of that one post.
What the public layer does not carry is any sense of progress. It cannot say whether people watched three seconds or three minutes, and it cannot separate an intentional play from a clip that started as someone scrolled.
That gap is the reason Twitter video analytics needs a second layer.
The view count says the clip was shown. It never says whether anyone stayed.
What X keeps private: the playback stages
X records how far each video was played and releases those numbers only to the account that owns it. The X API metrics documentation lists playback at 0%, 25%, 50%, 75% and 100% as non-public metrics. They are available with the owner's user authorization, not to anyone who looks at the post.
Circleboom's Video Analytics reads that private layer for your connected account and turns it into a horizontal bar chart per native video. The five bars are Video Started, 25% Watched, 50% Watched, 75% Watched and 100% Watched, and a date range selector narrows the view to the videos you want.
Access to owner-only data is exactly where trust matters. Circleboom appears on X's official Enterprise developers list. The playback stages come through authorized access, complete and policy-compliant.
Owner-only metrics never appear on the public page, so a tool that scrapes the page has no way to read them.
The view lives inside the X Post Planner menu, under Post Analytics.
When I review videos there, I read the chart first and the engagement figures second, because the chart tells me which part of the clip to rewatch. I learned that order the hard way, after reacting to reply counts on clips most people had abandoned early.
You can open the same stage chart for X videos for your own account, with no change to how you post.
Only one condition applies: the video must be uploaded natively to X. A post that links to YouTube or another host has no playback stages to read.
What the private layer can and cannot tell you
The stage chart is precise about some things and silent about others. It can tell you:
- How many plays began, and how many reached each quarter.
- Which interval lost the most plays.
- How two similar videos compare, stage by stage.
It cannot tell you:
- The exact second someone left.
- Why they left.
- Who they were.
- How they felt about the video.
- Whether they bought anything.
Circleboom's documentation draws the same boundary: the milestones identify broad intervals, and they do not establish sentiment or business conversions.
Two practical details round this out. The values reflect the moment Circleboom retrieved them from X, so a refresh pulls the latest counts. And Video Analytics only reports; editing, deleting and reposting all happen outside it.
Is a video start the same as a viewer?
No. A video start is a playback event, and it includes autoplay from the timeline, so it does not equal a person who chose to watch.
This distinction answers the question creators ask most about their video data: can I see who watched? Neither layer shows names.
The public view count is a single total, and the private stages are counts of playback reaching each point. Video Started does not identify distinct people, and no stage comes with a list of who reached it.
A start is something that happened on a screen; a viewer is someone who stayed past the opening.
Autoplay is the reason. A video that begins playing on its own as someone scrolls past registers a start even if that person never meant to watch. On a busy timeline, a clip can collect many such starts in a short time without a single person deciding to give it their attention.
That reframes how the top bar should be read. A large Video Started number tells you the clip was shown and began playing many times. It is context for the bars below it, not proof of interest.
The 25% Watched bar is the first one that reflects someone staying with the video, which is why the drop between the top two bars says so much about your opening.
The practical rule follows from it. Compare starts with engagement such as likes, reposts and replies to judge real interest, and compare stages with each other to judge holding power. Never treat starts as an audience size.
How the five playback stages read as one funnel
The five stages form a funnel in which each bar counts the plays that reached that point. The value is in the steps between bars, read as intervals:
| Stage | What it counts | A big drop right after it points to |
|---|---|---|
| Video Started | Plays that began, autoplay included | The opening quarter |
| 25% Watched | Plays that passed a quarter | The second quarter |
| 50% Watched | Plays that reached halfway | The third quarter |
| 75% Watched | Plays that passed three quarters | The final quarter |
| 100% Watched | Plays that reached the end | Nothing after it; it closes the funnel |
Read top to bottom, the table turns a chart into a to-do list: find the largest step, look up the quarter it points to, rewatch that quarter. That is the core move of Twitter video analytics, and everything else refines it.
Each step marks a stretch worth rewatching, not an exact second and not a confirmed cause. Two habits keep the reading accurate:
- Convert counts to percentages of Video Started before comparing videos, since raw completions favor whichever clip was shown most.
- Keep comparisons within one format, because video length changes how many seconds each quarter covers.
These Twitter video analytics supply the inputs; deciding what counts as a strong finish share is another question.
Where the public and private layers disagree
The two layers disagree most often on popular videos. A high view count with weak progression through the stages means the clip was shown widely and held few of its plays.
Circleboom's guidance for that pattern is to compare stage shares across similar videos before drawing conclusions, and not to infer autoplay behavior or viewer intent from totals alone.
The reverse also happens: modest views with strong progression, which usually means the right people found it and stayed.
When the private layer looks healthy and the view count looks small, reach is the question. The impression analytics view counts how many times the posts were shown, so you can tell limited distribution apart from weak holding power.
When both layers look healthy but the replies are quiet, check engagement analytics for the same posts. A video can be watched to the end and still collect few public responses.
Before changing a quiet video, ask whether its purpose required a public reply in the first place.
If you are weighing a separate analytics tool against X's native numbers, a roundup of the best Twitter analytics tools compares the options.
When X's own analytics page will not load: the alternative route to your post and video numbers.
How to use the weak interval in your next video
The weak interval tells you what to change before the next upload, and the next upload is where the analysis pays off. Circleboom's recommended next action when retention is weak is to revise the concept before scheduling:
- Tighten the hook when the drop sits before 25%.
- Shorten or clarify the stretch where a later step falls hardest.
- Change the visual framing or thumbnail when attention dies early.
Once the revision is ready, schedule it from X Post Planner. The workflow to schedule a tweet with a photo, video or GIF covers the posting side step by step.
An illustrative case shows how this plays out. Say a product demo lost most of its plays before the 25% mark while the later bars held steady. The revision cuts the slow introduction and opens on the result.
If the next version's first step down is smaller, the opening was the problem; if it is the same, the topic or the thumbnail is the next suspect.
Timing deserves its own check when a revised video still starts slowly. X Post Planner's best-time slots handle the scheduling directly, while When Followers Are Online shows the follower activity graph behind those slots.
Upload the new clip natively so it lands in the same chart. Then you can set the old and new versions side by side and see whether the step you targeted got smaller.
Strong performers are also worth saving to Collections, so the hook or script that held viewers can be reused.
Twitter video analytics checklist for your next upload
Both layers of video data are yours to read; only one of them explains drop-off. Before your next video goes out:
- Upload the file natively to X, never as a YouTube or Vimeo link.
- Read the five-stage chart of your last similar video.
- Mark the interval with the largest step down.
- Revise that stretch, then schedule the new version.
- Compare the old and new charts once the new one has a few days of data.
To begin, open Circleboom Twitter Management with the account that posts your videos, then go to X Post Planner and Post Analytics.
→ Read the private layer with Circleboom's video analytics for X
What to know about Twitter video data
Is X Premium worth it just for video analytics?
Not for the stage chart, since Circleboom shows starts and the four watch milestones for your native videos without X Premium. Premium bundles Media Studio with other features, so whether Twitter Premium is worth it depends on the rest of what you would use.
Why do my older videos show missing or partial stages?
X may no longer return granular playback data for older posts, so their stages can come back partial or empty. That is a platform data limit, not a Circleboom error, and the same kind of gap explains many cases of Twitter analytics not working.
Can I see the watch stages of a competitor's video?
No. X releases playback stages only with the owning account's authorization, so Circleboom charts them for the X account you connect and nobody else's. For a competitor's clip, the public view count is the only video number on offer.
){kind=link}
