First, compare surfaces against a typical period
Resist the urge to look at the views chart alone. In Advanced mode, open the traffic sources table for the weeks of the drop and the same number of weeks from a period you consider normal, then lay the two side by side. Read views, impressions and impressions click-through rate (CTRShare of impressions that result in a click.) per source rather than for the channel as a whole. A fall that is concentrated in browse features or suggested videos is a recommendation problem; one concentrated in YouTube search is a demand or relevance problem; one concentrated in a single video is not a channel problem at all.
Then split the period by video. Sort by views lost and check whether one or two titles account for most of the gap. Finally, mark the calendar: when did uploads go out, which formats were they, and did anything change in the weeks before the decline. The traffic sources guide explains how the report attributes each view; this page uses it only to answer two questions, which surface lost views and whether impressions, the click or the watching broke. Work down the table in order, because the earlier rows are the quickest to confirm or rule out.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Views down mainly on the home feed or suggested; impressions there fell | Impressions by source for the drop against a typical period | Impressions fell on browse or suggested | Find the step that weakened on those surfaces and repair it there |
| Impressions held but views from them fell on recent uploads | Click-through of new videos against earlier videos at the same age, per source | Click-through on new uploads fell | Rework packaging on the affected uploads and test one change |
| Recent videos shown widely at first, then offered less | Retention curves of recent uploads against the channel's usual shape | Retention weaker, so recommendations narrowed | Fix the opening or structure in the next edits |
| Subscribers' feed views and returning viewers both falling | Formats and topics of recent uploads; gaps in the publishing calendar | Uploads drifted from the channel's promise, or cadence broke | Return to the core format and restore a sustainable rhythm |
| YouTube search views falling across older videos | Search terms per video across seasons; the same weeks a year earlier | Search demand fell or a season ended | Plan around the cycle and refresh evergreen videos |
| The fall sits almost entirely in one recent video | Views by video for the period; that video's daily curve | A breakout returning to its normal level | Set the baseline without the outlier and study what made it travel |
Fix, verify and prevent each cause
Impressions lost on the home feed or suggested videos
Browse features and suggested videos are where most channels earn the bulk of their views, and both depend on YouTube predicting that a viewer will choose and enjoy the video. When impressions on those surfaces shrink, the drop is a symptom of something upstream, so the job is to find which step changed for the videos that lost exposure.
Fix: filter the traffic sources table to the surface that fell, open the suggested videos list to see which neighbours the channel used to sit beside, and compare recent uploads with the videos that once earned those placements; the traffic sources guide sets out what each surface rewards. Repair the weakest step for that surface rather than everything at once.
Verify: impressions on the affected surface stop falling over the following uploads and begin to recover at the same video age.
Prevent: read impressions per surface every week, not total views, so a narrowing shows before the total does.
Click-through on recent uploads slipping
If impressions held and views fell, fewer people chose the videos they were offered. That points at the packaging on recent uploads: a new thumbnail style, a title pattern that reads differently, or topics framed in a way the audience does not recognise.
Fix: compare each recent upload's click-through with the channel's earlier videos at the same age and on the same surface, then rework the packaging on the clearest underperformer using the principles in the thumbnails and click-through guide; where the channel is eligible, YouTube's thumbnail testing compares variants on watch time.
Verify: click-through on that surface returns towards the channel's usual range while impressions stay level.
Prevent: keep a written visual system and review each new pair of thumbnail and title against the last few winners before publishing.
Weaker retention narrowing recommendations
YouTube's help material describes recommendations that learn from whether viewers watch and are satisfied. A run of recent videos that lose people faster than usual gives the system less reason to keep offering them, and the fall appears in impressions a little later.
Fix: open the curves of the recent uploads beside the channel's typical shape, find where each departs from it, and change the script or edit at that point, following the method in the retention curves guide.
Verify: the next uploads match or beat the channel's normal curve at the same age, and impressions on suggested stop shrinking after them.
Prevent: send one sentence about each recent curve to whoever writes and edits the next video.
Uploads that drifted from the channel's promise, or a broken rhythm
Subscribers follow an expectation. When recent videos change subject, format or tone, or when uploads stop for a while and resume irregularly, the subscriptions feed and the home feed both show the channel to its own audience less often, and returning viewers fall first.
Fix: list the last several uploads against the content pillars and series the channel set out, return to the format the core audience came for, and choose a cadence the team can sustain, as the channel strategy hub describes.
Verify: returning viewers and views from browse features recover across the next few uploads in the core format; the guide to reading YouTube audience reports shows how to read that split by format against the channel's own baseline.
Prevent: schedule experiments as a labelled series alongside the core output, never as its replacement. When the change of subject is deliberate because the business now needs a different audience, the guide to changing a YouTube channel's focus plans it instead, from a labelled pilot series to the channel page.
Falling search demand or a season ending
Search views follow what people type, and plenty of topics rise and fall with the calendar or fade as a product ages. A channel built on search can lose views on many older videos at once without anything being wrong with them.
Fix: compare search terms and search views per video with the same weeks of an earlier year, check the topic's interest over time with the research tools the video SEO hub lists, and plan the next uploads around the cycle while refreshing titles and chapters on evergreen videos that still match a live query.
Verify: search views follow the expected seasonal shape and the refreshed videos hold their share of what demand remains.
Prevent: keep a calendar of the channel's seasonal topics and publish ahead of each rise rather than during it.
A breakout video settling back
A single video that travelled far beyond the usual audience lifts the channel total for weeks and then fades, which makes an ordinary period look like a collapse.
Fix: remove the outlier from the comparison, rebuild the baseline from the rest of the catalogue, and study what made that video spread, using the per-video reports the YouTube Analytics hub describes.
Verify: with the outlier excluded, the rest of the channel is level with the typical period, or the real decline is now visible and can be diagnosed with the rows above.
Prevent: report the channel with and without its biggest video each month, so a breakout is never mistaken for the new normal.
Prevention: a weekly surface review
Most sudden drops were visible a week or two earlier on one surface. A short weekly review catches them. Open the traffic sources table for the past week against a typical week and note any surface whose impressions moved noticeably. For every upload since the last review, record impressions, click-through and the shape of the opening of the retention curve on each main surface, compared with recent uploads at the same age. Mark seasonal topics in the calendar a month ahead. Keep one indicator (KPIA metric tied directly to an objective.) that matches what the channel is for, such as views from new viewers or watch time from returning viewers, and read the surfaces beneath it rather than the total. Treat ReachUnique people who saw an ad. on each surface as an early warning: suggested and browse usually move before total views do. When something does move, write down the one change you will make and the date you will check it, then leave everything else alone until that date.
A cooking channel loses a third of its views in a month
A cooking channel with 120,000 subscribers sees weekly views fall from about 260,000 to 175,000 over four weeks in February. The traffic sources table against the same four weeks of the previous autumn shows YouTube search almost unchanged, suggested videos down 18% and browse features down 41%. By video, no single title explains the gap. The calendar shows the channel switched in January from weeknight recipes to a series of long restaurant reviews, and returning viewers fell from 46% to 33% of views. Retention on the reviews looks ordinary, and click-through among subscribers is lower than on recipes. The team restores two weeknight recipes a week, moves the reviews into a fortnightly labelled series, and checks again after six weeks: browse impressions have recovered to within 10% of the autumn level and weekly views are back above 240,000.
Common mistakes
Reading total views instead of views per surface. Comparing a new upload with videos that have been live for months. Swapping thumbnails across the catalogue when the problem is retention. Treating a breakout's decline as the channel failing. Deleting and re-uploading videos. Changing format, cadence and packaging in the same month, so nothing can be learned from the result.
