Home / Blog / Retention on a ranking short

Guide

Retention on a ranking short

2026-09-23 · Dylan Yu

Retention on a ranking short is not a mystery and it is not a hack. People stay when the next number might be better, and they leave when the current number already bored them. You cannot see that in a template. You can see it if you watch your own countdown once, honestly, at the point where you want to skip.

This is not a report with view counts. We do not have your analytics, and we are not going to invent a percentage. The notes below are structural. They match how the format is built in the editor: high number first, title on the hold, hold length you chose.

The first clip is the ad for the rest

#5 or #10 is not the throwaway. It is the proof. If the series title says the video is fails, the first clip must be an obvious fail within the first second of the hold. If it is a menu, a walk to the arena, or a face talking about the fail, you spent the proof on a preface.

Trim until the action is the first frame of the file, then set the hold. A strong first clip buys patience for a stranger second clip. A weak first clip does not get a second chance just because a sting said “#4.”

Do not open with #1. The structure note explains why. The retention version is shorter: you spent the peak, and the graph of interest only falls.

Sameness is the midpoint drop

Countdowns die in the middle when ranks 6 through 3 are the same shot with different titles. The viewer learns the ceiling of the video and waits only if they believe #1 breaks the pattern. If they can predict the pattern, they leave before you break it.

The fix is to cut, not to decorate. Remove two similar clips. A tighter Top 5 will hold better than a padded Top 10. The Top 10 page has the runtime table. Use it. If you cannot describe how #7 is different from #6 in five words, #7 should not be in the file.

Variety can be angle, stakes, or outcome. It cannot be a synonym. “Another miss, but from the left” is still another miss unless the left angle shows something the first miss hid.

Titles: short, specific, unspoiled

A title is read in the time the hold is on screen, often without sound, often while the viewer is also looking at the action. Long titles lose. Cute titles that do not describe the clip lose. Titles that spoil the outcome can work as a joke and can also remove the reason to keep watching the hold.

Prefer the setup. “One bullet left” keeps the next second alive. “Dies to the last player” ends the clip in the text. Use the spoiler title only when the picture is confusing without it.

Read all the titles in order. If you stumble, the viewer will not work harder than you did. The plate wraps to two lines and then cuts the rest off with an ellipsis. If you are relying on the third line, you do not have it.

Hold length is a retention control

Too long: the action finished and the number has not changed. The viewer feels the template waiting.

Too short: the title is still being read when the sting already fired. The viewer feels rushed and a little cheated, especially early in a Top 10 where they are still learning the rhythm.

Match the hold to the action, then check the total. A Top 5 can afford four seconds. A Top 10 usually cannot afford four seconds on every card. Let one card be long if it earns it. Do not let the slider sit at the maximum because the maximum exists.

Watch the countdown on a phone, not only in the desktop preview. The workbench preview is large on purpose. The post is not. If you cannot read the title on a phone, shorten the words or accept that the title is for you, not for them.

The penultimate rank

#2 is where arguments start, if you ordered well. It should be almost the winner. If #2 is a slump, the last sting feels like relief rather than a reveal, and people who already left during #2 never see #1. Swap #2 with a middle rank before you touch the font.

#1 should pay the series title off. If your title promised fails, #1 is the least excusable fail, not the coolest highlight you wanted an excuse to include. Category drift at the end is a retention problem disguised as a finale.

Captions and covers

The on-screen title is not the post caption. The caption can repeat the promise and invite the argument: “#2 should have been #1.” That invitation is optional. It also trains people to scrub. If you list every title in the caption, you have published the ranking as text. Some accounts want that. Some want the wait. Choose.

The cover frame should show a number. A number is the format at the size of a grid. A pretty frame from the middle of a clip, with no number, looks like every other clip in the grid.

What not to add

Do not add a fake percentage. Do not add “97% fail this.” Do not add a leaderboard you did not measure. The format is already a claim. Stacking a false statistic on it makes the claim cheap. A clean countdown of real clips is more interesting than a countdown wearing a lab coat.

If the video is still losing people and the structure is honest, the clips are the problem. No template fixes clips you would not watch yourself. Film or select again. Then come back and order them.

Questions

Does a bigger number mean better retention?

No. Top 10 often loses people in the middle because the holds are either too long or too same-y. Top 5 is the safer default.

Should I spoil #1 in the caption?

Only if you want comments more than you want the wait. The video’s structure assumes you did not.

What is the first clip’s job?

To prove the series title in one glance, so the viewer trusts that the rest of the countdown is the same kind of thing.

Keep going