First, read the opening of the curve by traffic source
Open the retention report for the video in Advanced mode and look only at the opening stretch, roughly the first minute of a long-form video. Overlay the curve on the channel's recent videos of the same type and length at the same age, so the comparison is fair. Then apply the traffic source filter and read the opening for each main source in turn: YouTube search, suggested videos, browse features, and the Shorts feed or related-video link if the video receives viewers from Shorts. The retention curves guide explains the report itself; this page uses it to answer one question, which group of viewers is leaving and when.
Two readings settle most cases. If every source falls at the same second, something in the video at that moment is the cause, and the timeline will show what is on screen. If one source falls far more steeply than the others, the viewers from that source expected something else, and the fix lies in the packaging or in how they were routed. Set the click-through rate (CTRShare of impressions that result in a click.) for each source beside its curve, because a strong click followed by a cliff is the signature of a broken promise.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| All sources slide steadily through the opening, then level off when the content starts | What is on screen during the slide: logo, greeting, outline | Intro delays what the title promised | Open on the promised result; cut the preamble |
| A sharp cliff in the first seconds; click-through healthy or above baseline | Thumbnail and title against the first frames | Packaging promises something the video does not deliver | Show the promised thing immediately or make the packaging match |
| One source falls much faster than the others | Retention per source; the suggested list or search terms for that source | Viewers arriving from a surface that expected something else | Adjust packaging or metadata so that surface sends the right viewers |
| A dip shortly after the start that only partly recovers | The timeline at the dip: recap, sponsor read, tangent | Recap, sponsor read or tangent placed early | Move it after the first payoff or remove it |
| Viewers from Shorts leave almost at once; other sources hold | Retention filtered to Shorts-related traffic against the rest | Shorts viewers sent to long-form they did not choose | Link Shorts only to long-form that continues the same idea |
Fix, verify and prevent each cause
An intro that makes the viewer wait
Branded animations, greetings, a spoken outline of the video and early requests to subscribe all sit between the click and the reason for it. The curve slopes downwards through that stretch and flattens once the real content begins, which is how this cause differs from a mismatch.
Fix: cut everything before the first moment of value and open on the outcome the title named, as the retention curves guide describes for intro drops; if the branding matters, place it after the first payoff.
Verify: on the next video, compared at the same age and with the same source filter, the opening slope is shallower and the curve settles higher.
Prevent: write the opening line of the script as the first thing the viewer will see and hear, and make "what is the payoff in the opening seconds" a question in the edit checklist.
Packaging the video does not honour
A thumbnail and title that promise a result, a reveal or a verdict earn the click; if the opening shows something else, the viewer leaves the moment they notice. The tell is a healthy click-through rate paired with a cliff, not a slope. Swapping the thumbnail for a weaker one would mask the problem rather than solve it.
Fix: compare the first frames with what the packaging claims, then either show the promised thing at once or change the packaging to describe the video that exists, following the honesty rule in the thumbnails and click-through guide.
Verify: the cliff softens on the next comparable period, and click-through stays close to where it was.
Prevent: before publishing, watch the opening straight after looking at the thumbnail and title, as a stranger would.
A surface that sends viewers expecting another video
Suggested videos can carry a video to audiences watching something adjacent but different, and a search term can match a title while wanting a different kind of answer. Those viewers leave early even when subscribers stay.
Fix: read which neighbouring videos or search terms supply the source with the steep fall, using the traffic sources guide, and adjust the title, description or thumbnail so the video is offered to viewers who want what it contains; leave the edit alone if subscribers hold.
Verify: the opening for that source moves closer to the others, or that source's share of views shrinks while better-matched sources grow.
Prevent: at each weekly review, read the opening per source for new uploads, not only the overall curve.
A recap, sponsor read or tangent before the first payoff
Material the viewer did not come for, placed early, invites them to leave before they have been given a reason to stay. A recap of the previous episode, an early sponsor segment or a personal digression shows as a dip shortly after the start that never fully recovers.
Fix: locate the dip on the timeline and move the segment after the first payoff, shorten it, or replace a recap with a chapter or a link in the description, as the retention curves guide suggests for recurring dips.
Verify: the next video shows no dip at the old position, and the curve after the opening sits higher than before.
Prevent: fix the running order in the script template, so the segment cannot drift back to the front in later episodes.
Shorts viewers routed to long-form they did not choose
A Short that links to a related long-form video can send viewers who enjoyed a quick moment into a long video that opens differently, covers more, or is about something else. They did not pick that video from a thumbnail, so many leave at once.
Fix: link each Short only to the long-form video that continues its exact idea, end the Short by naming what the viewer will get, and open the long-form video close to where the Short left off, following the bridging advice in the Shorts versus long-form comparison.
Verify: the opening for viewers arriving from Shorts moves towards the other sources at the same age.
Prevent: choose the related video when the Short is scripted, not after it is published.
Prevention: scripting the opening to the promise
Early drops are mostly written into the script before the camera rolls. Start every script from the packaging: write the title and describe the thumbnail first, then write an opening that delivers the start of that promise in the first moments on screen. Keep branding, sponsor reads, recaps and housekeeping out of the opening, and put them after the first payoff. Decide who the video is for and which surfaces should bring them, and make the title and description say so, so that the people who arrive are the people the opening was written for. When a video is linked from Shorts or from an end screen, open it in a way that follows on from the video that sent people there. After publishing, read the opening per source against the channel's recent videos at fixed ages, and pass one sentence about it to whoever edits the next video. Judge the result on the channel's chosen indicator (KPIA metric tied directly to an objective.), such as watch time from returning viewers, rather than on a single curve. Wider ReachUnique people who saw an ad. follows from videos people keep watching, so the opening is where distribution is earned. The tutorial on writing a YouTube video brief gives the opening its own field, planned after the packaging and before the chapters, so the script starts from a written promise.
A fitness channel separates its intro problem from a routing problem
A home-fitness channel's latest workout video keeps only 52% of viewers after the first 30 seconds, against 68% on its previous five workouts at the same age. Filtered by source, browse and search both slide steadily from the start to 0:25 and settle around 64%; viewers arriving from a linked Short fall to 31% within 10 seconds. The timeline shows a 20-second montage and a spoken outline before the first exercise. The Short that sends traffic shows a single stretch for desk workers, while the linked video is a full-body strength routine. The team cuts the montage to a single line over the first exercise, and relinks the Short to a desk-stretch video. On the next workout, compared at seven days, 67% remain after 30 seconds across browse and search, and viewers from the relinked Short hold at 58%.
Common mistakes
Swapping the thumbnail when the curve shows a slow intro. Reading the overall curve without the source filter. Comparing with videos of a different length or age. Moving the sponsor read without changing the template, so it returns. Linking every Short to the newest long-form video rather than the one that continues it.
