How to Stop People Scrolling Past Your Video.

Most videos are not skipped because they are bad.

They are skipped because the opening frame gave the viewer nothing to decide with, and scrolling costs nothing.

A viewer meets your video with one frame, half a second of audio and whatever text sits on screen. If none of those name something worth staying for, the thumb moves before the content starts.

So the fix is rarely a better hook in the abstract. It is three concrete things: what is visible in the first frame, what the first line actually says, and whether the video keeps moving forward every few seconds after that.

The useful part is that you do not have to guess which one is broken. Every major platform now reports where viewers left, and the shape of that drop-off tells you whether you have a packaging problem, an opening problem or a pacing problem.

What a retention curve actually is

A retention curve plots the percentage of viewers still watching at each second of your video. It begins at 100 percent and falls, and the speed and position of the fall is the diagnosis.

YouTube reports this as audience retention in Studio and marks the moments it considers notable: intro, spikes, dips and top moments. Its intro measure is specific, the percentage of your audience still watching after the first 30 seconds.

Instagram shows a retention chart on individual reels alongside a skip rate for the opening. Labels differ by platform, so read your own tools rather than a benchmark borrowed from someone else’s account.

How to read the shape of the curve

Four shapes cover most of what you will see, and YouTube publishes what each one means.

• A flat line means viewers are watching that part of the video from start to finish.
• A gradual decline means viewers are losing interest over time, which is normal. The gradient is the thing to judge.
• A downward dip marks a moment that was skipped, or where viewers stopped watching.
• An upward spike marks a moment that was rewatched or shared, which can also mean it was unclear the first time.

Read the curve with the video open beside it. A timestamp on its own tells you nothing. The two seconds before the drop tell you everything.

The first second is its own decision

A counted view does not mean anybody watched. YouTube reports two numbers on a Short, views and engaged views, and defines engaged views as how many times viewers stayed to watch past the initial seconds, not including loops.

The gap between those two numbers is your bounce. Nothing you do after the opening changes it.

Instagram names the same moment from the other direction. Skip rate reports the share of views from people who skipped inside the first three seconds, so it counts the people who left rather than the people who stayed. Reading it beside share rate is the fastest way to separate an opening problem from a middle problem.

The three drop-offs, and what each one means

Nearly every retention failure sorts into one of three, and they need different fixes.

• The instant bounce, at frame one. This is a packaging problem rather than a content problem. The video can be excellent and nobody will find out.
• The early skip, inside the first few seconds. The opening was seen, understood and judged not worth the wait, usually because it took too long to name its subject.
• The mid-video slump. The viewer committed, then the video stopped moving forward and gave them a moment to leave.

The tail is a fourth pattern and it is not a fault. A fall near the end is normal, so judge it against your own median video rather than against 100 percent.

Why holding people to the end decides your reach

TikTok’s own description of the For You feed says that a strong indicator of interest, such as whether a user finishes watching a longer video from beginning to end, is given greater weight than a weak indicator.

Instagram describes ranking reels as a set of predictions, one of which is how likely you are to watch a reel all the way through.

The practical read is that the last few seconds are not a formality. A video that holds people to the end is being scored by systems looking for exactly that, and one leaking viewers at second two is asking to be shown to fewer people.

None of that needs spend behind it. It is the same organic system working better or worse.

How to fix an immediate bounce

Everything here is packaging, and all of it happens before anyone hears a word.

• Open on the most visually loaded moment you have, not the setup that leads to it.
• Put the subject in the frame. If a viewer cannot tell what the video is about with the sound off, they will not wait to find out.
• Cut intros, logos and greetings. They are the most reliable way to lose people who were willing to watch.
• Add on-screen text of roughly five words that names the payoff rather than the topic.
• Frame for the crop the platform applies, so the subject survives the safe area on every placement.
• Make something move inside the first second. A static opening frame reads as nothing happening.

How to fix a mid-video slump

Find the timestamp of the drop, then watch the two or three seconds in front of it. The cause is almost always sitting there.

The usual suspects are a shot held too long, a sentence that adds no new information, an explanation delivered before the thing it explains is shown, and a visual that repeats what the previous one already said.

The fix is usually a cut rather than an addition. Remove the seconds before the drop, bring the payoff earlier, and give the viewer a new visual beat every few seconds so the video keeps signalling that something is still happening.

That is editing work more than filming work, which is why video production and editing tends to move retention further than another shoot day.

Dead frames are the usual culprit

A dead frame is any second in which nothing new appears, moves or is said. Strung together, they are the most common cause of a mid-video slump.

The cheapest test costs you one viewing. Scrub through with the sound off and look for a stretch where the picture barely changes. Every one of those stretches is an invitation to scroll.

Sound off is a retention problem, not an accessibility footnote

Feeds start muted and people watch in public. If your message travels only through audio, a silent viewer sees someone moving their mouth and nothing happening, so they leave before the point arrives.

Captions carry the words, on-screen text carries the promise, and the visual carries the story. Building sound off friendly content usually holds better even when the sound is on, because it forces the story into the picture.

The diagnostic sequence to run this week

1. Take your last ten videos and rank them by average watch time, never by views.
2. For each one, write down the opening number your platform gives you, engaged views against views, or skip rate.
3. Open the retention curve for each and note the timestamp of the first steep drop.
4. Sort those drops into the three buckets: instant bounce, early skip, mid-video slump.
5. For every mid-video drop, watch the three seconds before it and write down literally what is on screen.
6. Compare the openings of your best three against your worst three, and name the difference in one sentence.
7. Change only the opening on your next batch, then read the same numbers again.

Step seven is the one teams skip. Change three things at once and you learn nothing, which is how a brand ends up with a year of posts and no idea what works.

What good looks like

A healthy curve is recognisable at a glance.

• The first seconds are flat rather than a cliff, so the opening earned the watch.
• The decline is a slope, not a step, because nothing in the middle stalls.
• The largest drop sits at the end, where it belongs.
• Spikes land on the moments you intended people to rewatch.
• The same opening pattern shows up across your strongest videos, which means it is repeatable rather than lucky.

The organisational version matters as much. A team that can say why video nine held better than video eight will keep improving, and a team that cannot will keep producing at random.

How NBK thinks about video retention

Retention is a feedback loop, not a talent. The brands that improve it are the ones where the person editing the next video has already seen the curve from the last one.

We treat it as an operations problem: a fixed review of drop-off points, a note of which opening pattern held and which leaked, and a small set of formats repeated until the curve flattens.

The habit comes from publisher work. Across the team’s time at UNILAD, LADbible Group, SPORF and Social Chain, and at 46K+ posts shipped and 600M+ views a month, volume only compounds when every video teaches the next one something.

Next step

Read the curves on the ten videos you have already posted before you plan another one. The pattern in your own drop-offs is worth more than any benchmark from an account that is not yours.

Then change one variable at a time. Opening frame for one batch, first line for the next, pacing for the third, with the numbers kept side by side so you can see which change moved them.

If your video output feels busy but not effective, an NBK audit will find the constraint in the system rather than in the last edit.