Guide

Reading Your Analytics: The 6 Metrics That Matter for Story Channels

The six metrics that actually matter for a faceless story channel are click-through rate, three-second hook hold, average percentage viewed, the shape of the retention curve, rewatch rate, and follows per 1,000 views. Every one of them is a rate, which is the point: rates tell you whether the video worked, while totals like views and watch time mostly tell you how much the platform decided to test it. Read them in funnel order, find the first stage that is failing, change one thing on the next video, and re-measure. That loop is the whole discipline. Everything below is how to run it.

Why most creators read analytics wrong

Open any creator dashboard and the biggest number on the screen is views. It is also the least actionable number on the screen. Views are the product of two things: how good the video is, and how many people the platform chose to show it to. The second part swings wildly and is largely outside your control, especially on short-form, where one video can be tested to 400 people and the next to 40,000 for reasons that have nothing to do with quality. If you judge yourself on views, you will conclude that a good video was bad and reshape your format around noise.

Rates do not have that problem. Click-through rate, hook hold, percentage viewed and follow rate all normalise by the size of the test, so a video shown to 500 people and a video shown to 50,000 can be compared honestly. That is why every metric in this guide is a rate, and why the first habit worth building is simple: never open a video’s analytics without a comparison. One number alone means nothing. The same number next to your last ten videos means everything.

The second habit is scale discipline. Below roughly ten videos in a consistent format, or a few thousand views on any single upload, the variance between uploads is larger than the effect you are trying to measure. Publish a real series before you start optimising, or you will spend a month chasing statistical noise.

The funnel order: fix the first broken stage

The six metrics are not a checklist, they are a funnel, and their order matters. A viewer has to be shown the video, click or stop scrolling, survive the first three seconds, stay through the story, want it again, and only then follow. If an early stage is broken, everything downstream is polluted — a video with a weak hook will always show poor completion, and no amount of ending-tuning will fix it. So read top to bottom and stop at the first stage that is clearly below your own median. That is the only stage worth changing this week.

This is also what stops the most common failure mode in creator analytics: changing four things at once. If you rewrite the hook, swap the thumbnail, shorten the video and change the narrator voice in the same upload, the numbers move but you have learned nothing. One variable per upload, held for at least three videos, is slower and it is the only thing that actually compounds.

1. Hook hold: the first three seconds

Hook hold is the percentage of people still watching at the three-second mark. On short-form it is the single most predictive number you have, because the platform is deciding within those seconds whether to keep testing your video. Every platform reports it slightly differently — YouTube shows it as the leftmost point on the retention graph, TikTok reports watched-full-video and average-watch-time alongside a retention curve, Instagram surfaces it as the drop in the first seconds — but you are always looking at the same thing: how many people did not swipe away.

Some early drop is normal and universal; a chunk of any audience bounces before the video has really started. What you care about is your relative drop. Pull the three-second figure for your last ten videos, take the median, and treat that as your line. A video far below it has a hook problem, and hook problems are almost always one of three things: the opening line does not deliver on the promise the title or thumbnail made, the first frame is visually static, or there is throat-clearing before the story starts — a logo, a greeting, a “in today’s video”. Cut all three. The story should be running on frame one. See How to Hook Viewers in the First 5 Seconds for the specific openings that hold.

2. Retention curve shape, not the average

The retention curve is the most information-dense chart on your channel and most creators glance at it for two seconds and read only the average. The average is a summary of the curve; the curve is the diagnosis. A video that averages 45 percent because it declined smoothly is a completely different problem from a video that averages 45 percent because it held beautifully and then fell off a cliff at 0:38.

Read the curve for three features. Cliffs are vertical drops at a specific timestamp: go to that timestamp and watch. Nine times out of ten there is a concrete cause — a long establishing shot, a sentence that explains instead of advances, a music change that reads as an ending, a scene where the character’s face visibly shifts. Slopes are steady decline with no single culprit, which usually means the pacing is uniformly slow rather than any one moment being bad. Bumps are rises, where retention goes back up: viewers scrubbed back or rewatched, and whatever is at that timestamp is the strongest thing in your video. Do more of it.

3. Average percentage viewed

Average view duration is reported everywhere and it is the wrong metric to optimise, because it rewards length rather than quality. A ten-minute video with 90 seconds of average duration looks “better” than a 45-second Short watched to the end, and it is plainly worse. Average percentage viewed normalises for length and lets you compare a Short, a five-minute story and a compilation on the same axis.

Sensible working floors, judged against your own channel rather than any published benchmark: for a 60-second vertical story, finishing above roughly half the audience is a workable base and the strong videos will run much higher; for a mid-length story of five to ten minutes, 40 to 50 percent is usually healthy. The number falls as length rises — that is expected and not a failure. What matters is the direction across your last ten uploads, and whether a specific format change moved it.

One trap worth naming: percentage viewed can be gamed by simply making videos shorter, which raises the rate while lowering total watch time and often the follow rate too. If percentage viewed climbs while follows per 1,000 views falls, you have made the video easier to finish and less worth following. Check the two together.

4. Click-through rate on impressions

Click-through rate is a long-form metric: the share of people who saw the thumbnail and clicked. It is meaningful on YouTube’s home and suggested surfaces and largely irrelevant on a swipe-based feed, where the thumbnail is barely a factor. Treat it as the packaging score, and read it strictly alongside retention.

The two-by-two is the useful part. Low CTR, good retention means the video is fine and the packaging is failing it: rework the title and thumbnail, and consider re-uploading a new thumbnail on the existing video. High CTR, poor retention is the dangerous quadrant: the packaging over-promised and the video did not pay it off, which teaches the platform to stop recommending you. Low on both means the topic itself is not landing with this audience. High on both is the video to study and reproduce. Our guide to thumbnails and titles for faceless story channels covers the packaging side in detail.

Watch the trap here too: CTR is inflated on videos with small, subscriber-heavy impression sets and falls as the platform pushes a video to colder audiences. A dropping CTR on a video that is gaining views is often a sign of success, not failure.

5. Rewatches and loops

On short-form, retention above 100 percent is possible, because the video loops and viewers replay it. Rewatch rate is the clearest signal that the ending worked — that the last line made someone want the first line again. Story channels have a structural advantage here: a twist, an unanswered question or a line that reframes everything you just watched all produce replays in a way a straightforward explainer never will.

If your rewatch numbers are flat, look at the final four seconds. The most common killer is an ending that trails off, or a call to action that lands after the story has already resolved, breaking the loop. The fix is to end on the story beat, not the housekeeping: put the twist last, and make the final frame flow into the first frame so a loop feels intentional. Shares and saves belong in the same family of signals — both mean someone valued the video beyond their own viewing, and both tend to move together with rewatches.

6. Follows per 1,000 views

Total subscribers is a vanity number; follows per 1,000 views is a diagnosis. It answers the only question that determines whether you have a channel or a series of lottery tickets: having watched this, does someone want more of it? Divide the follows a video earned by its views, multiply by 1,000, and track it as a rate per upload.

A video with strong retention and near-zero follow rate is a specific, common and fixable failure: it entertained but it did not communicate a promise. The viewer enjoyed a story and had no idea what the next one would be. That is usually a branding problem rather than a story problem — the channel has no repeatable premise, the videos vary too much in format, or nothing at the end signals what following gets you. A consistent premise, a recognisable narrator and a visual identity that repeats every video are what convert a good watch into a follow. Our post on naming and branding a faceless story channel is the companion here.

The weekly review, in ten minutes

Analytics only pay off as a routine. Once a week, fill this in for the videos you published, compare against your rolling median, and pick exactly one change. Keep the log — the value is in the history, not any single row.

Three uploads is the minimum before you judge a change, because a single upload is well inside normal variance. If the metric moves in the right direction across all three, keep the change and move to the next stage in the funnel. If it does not move, revert and try a different lever on the same stage rather than jumping ahead.

Three curves and what each one means

Most story-channel retention problems reduce to three shapes, and each has a different fix. A cliff in the first seconds is a promise problem: the packaging said one thing and the opening delivered another, or the video takes too long to start. Rewrite the first line so it starts mid-situation, and cut anything before it. A steady slide with no cliffs is a pacing problem: nothing is bad, but nothing is urgent either. Tighten the script, cut the connective sentences, and make sure a new piece of information or a new complication lands every few seconds — the beat sheet for shorts is the practical version of this. A late cliff, where retention holds and then collapses near the end, usually means the story resolved before the video did; trim everything after the resolution.

One more pattern worth recognising: a curve that is strong for a specific segment and weak overall. If your platform lets you break retention down by traffic source or by new versus returning viewers, do it. A video that returning viewers finish and new viewers abandon is not a bad video — it is a video that assumes context newcomers do not have.

What to ignore (for now)

• Total subscribers. A stock, not a flow. Follows per 1,000 views is the version you can act on.
• Likes and comments in isolation. Nice, but they trail retention rather than predict it.
• Total watch time. Useful for monetisation thresholds, useless as a quality signal, because it scales with reach.
• RPM and revenue, before you have volume. Below meaningful view counts these numbers are pure noise.
• Day-one performance. Short-form especially can be re-tested days or weeks later; do not bury a video after 24 hours.
• Any single video. Read the set. One upload is an anecdote; ten in one format is data.

The reason most creators do not act on their analytics is not that the numbers are unclear — it is that acting on them means re-editing. When a re-cut hook or a tighter middle is a re-generation rather than an afternoon in an editor, the measure-change-remeasure loop actually runs. That is what Taleframe is for: describe the story once, get a finished narrated video, and make the next iteration cheap enough to be worth doing.

FAQ

Analytics tell you what to change; the rest of the channel decides how fast you can change it. If your bottleneck is output rather than insight, read Posting Schedules: How Often a Faceless Channel Should Upload, and From Idea to Upload: A Faceless Video Workflow for the production loop these metrics feed back into.

Further reading: the platform’s own explanation of how the retention graph is built and what each dip means — Audience retention report (YouTube Help).