Your Video Died at 40 Views Because the Metadata Was Written for 2018

Forty views is not an editing problem, it is a metadata problem

The shape is always the same in Analytics. Around 900 impressions in the first 72 hours, a click through rate near 1.8 percent, and an average view duration of 1 minute 10 on a nine minute video. Almost nobody watched long enough to dislike it. Almost nobody was shown it at all.

What failed is the package around the video, and it usually failed because it was built to a 2018 checklist. Thirty tags scraped from a keyword tool, a description with the target phrase repeated eleven times, a title assembled to match a search volume number. YouTube moved most of that weight elsewhere and never sent a memo.

So here is the inventory with the thresholds attached. What the ranking systems actually read sits in the list below, roughly in order of how much it decides. Four of the six are things you write or say on purpose, and the one everybody obsesses over sits at the bottom.

• Spoken content, indexed through the caption track, the largest text block your video owns
• The title, and inside it the first 40 characters more than the rest
• The description opening, the 157 characters that become the snippet
• Chapter timestamps, which let one video answer several different queries
• Click through rate, judged against its traffic source, never a universal average
• Retention at the 30 second mark, then average percentage viewed across the video
• Tags, which YouTube documentation itself calls minimal in discovery

The title: 100 characters allowed, about 60 anyone will read

The hard cap has been 100 characters since 2012. What changed is where the cut lands. Desktop search shows roughly 60 to 70 characters, with 62 to 66 fitting cleanly on a 1920 by 1080 screen. The iOS app cuts near 48 to 52, Android 50 to 55, mobile web around 55 to 60, and Google results truncate close to 60.

That geometry writes the rule for you: the whole promise goes inside the first 50 characters, and anything past 60 is a bonus a phone will delete. Front load the query the way your audience types it, then place the differentiator after it. People scan the first three words and decide there.

A search title and a high click through title are also different games. A video that lives on search should carry the query nearly verbatim, because that query will still be typed in three years. A browse video needs tension and specificity instead. Decide which door the video enters before publishing, then open Traffic Sources at day 14 and protect whichever one delivered.

• Cap is 100 characters, but plan for 60 on desktop and 48 to 55 on a phone
• Target query inside the first 40 characters, spelled the way people type it
• One promise per title, because two promises read as none
• Do not repeat the words already printed on the thumbnail
• Rewrite candidates: over 1,000 impressions, under 2 percent clicks, older than 30 days
• Never rewrite a video that is climbing right now
• Change one variable and wait 14 days before judging the result

The description: 5,000 characters allowed, 157 that decide the click

The field holds 5,000 characters and YouTube shows about 157 of them on desktop, closer to 100 on a phone, before Show more swallows the rest. That same opening is what Google lifts as the snippet when your video ranks in web search. So the first two sentences are the only ones written for a stranger, and spending them on a channel greeting is the most common unforced error in video SEO.

Chapters are the highest return item in the description and most channels still skip them. The requirements are strict and simple: at least three timestamps in ascending order, the first at 00:00, each chapter 10 seconds or longer, and the list in the description rather than a pinned comment. The payoff is key moments in Google search, where one fifteen minute video becomes several separate results.

What no longer works is the 300 word keyword dump. Metadata stuffing is a stated policy violation and can pull a video out of search entirely, a real risk for a fake reward. The ranking systems already read your spoken transcript, so repeating the phrase eleven times adds nothing. Two hundred to four hundred structured words beat the wall every time.

Writing that opening well on every single upload is a discipline problem rather than a knowledge one, which is exactly the kind of thing worth handing to a pipeline. In FalconVid the video SEO is produced together with the video: the title, the description and the tags are written from the actual script, so the opening 157 characters describe what was really said instead of repeating a phrase, and the chapter list is cut from a transcript that already exists rather than typed from a stopwatch. Nothing gets stuffed, because the text comes from the content and not from a keyword tool.

• First 157 characters: the topic in plain words plus the promise, no greeting
• Chapters need three timestamps minimum, the first at 00:00, each 10 seconds or longer
• Chapters go in the description, never in a pinned comment
• Aim for 200 to 400 real words, not 5,000 characters of keyword soup
• Two or three hashtags, and YouTube ignores all of them once you pass 15
• The first three hashtags also render above the video title
• Keyword stuffing in metadata is a policy violation, not a grey area
• Links go after the chapters, where a human can still find them

Tags: low weight, not zero weight, and what replaced them

The tag field holds 500 characters in total, roughly 8 to 12 useful entries, and YouTube help documentation states plainly that tags play a minimal role in discovery and exist mostly to catch common misspellings. The Creator Liaison has repeated the same thing publicly. Channels that strip tags off an entire library report no measurable change in performance.

Low is still not zero, so here is the honest use. Tags catch spelling variants of a brand or a name, and they disambiguate a term with two meanings, the mercury planet against the mercury element. That is worth five to eight tags and about two minutes. Thirty near duplicate tags crammed into the 500 character budget is not a strategy.

The signals that took over the job are the ones you control anyway: the caption track carrying every spoken word, the title, the description opening, the chapter labels, and early engagement. That is the ranking committee now. Four of those five are decided on purpose before you upload, and the fifth is decided by your first two days of viewers.

• The tag field holds 500 characters, roughly 8 to 12 usable tags
• YouTube documentation calls the role of tags minimal, mostly a misspelling catch
• Use them for spelling variants, brand names and ambiguous terms, then stop
• Hashtags are capped at 15, and past that count YouTube ignores every one
• Hashtags are a topic and browse signal, not a search ranking lever
• Budget two minutes on tags, and the eighteen you saved go to your first 30 seconds

Captions and transcript: the biggest block of indexed text you own

Every video with speech gets an automatic caption track and YouTube indexes that text. In a documented experiment, a phrase that appeared only inside the captions, nowhere in the title, description or tags, was enough for the video to rank fourth for that query. A ten minute video carries roughly 1,400 to 1,600 spoken words, around ten times your description and twenty times the tag field.

Which is why leaving automatic captions untouched is a quiet leak. Auto transcription drifts exactly where it hurts, on proper names, product names, jargon, numbers and accents outside the training middle, and those are the words you wanted to rank for. Correcting the transcript and uploading an SRT takes about ten minutes and fixes indexing and accessibility at once. Translated tracks open queries in languages your channel does not speak.

There is a retention effect on top of the search effect, and the two come from different files. Burned in on screen captions hold viewers watching with the sound off, a large share of mobile and feed viewing, while the closed caption file is the one that gets crawled. Read the auto transcript once after upload, because it is the fastest audit of what YouTube currently thinks your video is about.

This is the item most channels never fix, and it is the one a pipeline fixes for free. When the narration is generated instead of recorded, the caption track is not a machine guessing at what was said, it is the script itself, so names, numbers and jargon land correct on the first pass and there is no SRT to repair. FalconVid also burns karaoke style captions on screen, with more than 15 styles, so the file that holds retention and the file that gets crawled are both handled in the same run. And because narration runs in 63 languages, you can duplicate a project into another language paying only the difference and start collecting queries your channel could not previously spell.

Click through rate and the first 48 hours, where ranking is decided

The platform wide average click through rate sits between 4 and 5 percent, the good band runs from 4 to 10, and above 6 percent is genuinely strong. Those numbers are useless without the traffic source attached. Search runs 8 to 15 percent on a well optimized video, browse and suggested sit at 3 to 7, gaming often hits 8 to 15 while educational content lands near 4.5.

The thumbnail is the lever that moves that number and it deserves a guide of its own, so one sentence here is enough: the thumbnail is an input into ranking, not decoration. Perfect metadata on a video nobody clicks produces no watch data, and a video with no watch data has nothing to rank with. That is why a checklist about text still has to mention an image.

Then come the first 24 to 48 hours, when the decision actually happens. YouTube serves the upload to a small sample: your subscribers, the home feed of people who watch your topic, and a few suggested slots beside related videos. If click through and average view duration hold, the sample widens. If they do not, impressions decay within a day or two and restarting the same video later is very hard.

Which is worth knowing if you are not designing a thumbnail every week. In FalconVid the thumbnail is generated inside the same pipeline as the video and upscaled to 4K, and it lands attached to the upload with the title, description and tags already written for it. The video then goes out at the exact slot on the calendar you approved, instead of whenever the export happened to finish, and a fixed slot is the difference between a test you can read and a coin flip. Nothing here rescues a video nobody wanted, but it does mean the 48 hour window opens under conditions you chose.

• Platform average click through is 4 to 5 percent, and above 6 is strong
• Search traffic: 8 to 15 percent is the band for a well optimized video
• Browse and suggested: 3 to 7 percent, judged separately or you fix the wrong thing
• Retention target: 70 percent of viewers still there at 0:30
• Average percentage viewed of 40 to 50 percent on an eight to twelve minute video
• Be present for the first hour, early engagement is part of the sample
• If impressions flatlined after 72 hours, metadata edits will not rescue that upload

The checklist to run before you click publish

Done in the right order this takes about twelve minutes per video, and the order matters because each step feeds the next. Decide the query, write the title from the query, write the description opening from the title, cut the chapters from the transcript, and spend whatever attention is left on tags. Reverse it and you end up naming videos after a keyword tool.

It is worth being honest about what this does. A checklist does not rescue a video nobody wanted, and no arrangement of text makes a boring topic interesting. What it does is guarantee that a video people do want is actually findable, which is a different and far more common failure. It also compounds, since a searchable video with chapters keeps collecting impressions for months.

Paste the list below into your upload template and run it every single time. It works not because any one line is clever, but because the six or seven signals that still matter in 2026 are all covered, none of them are left to chance, and the twenty minutes that used to disappear into tags are back in your pocket.

• Target query decided before recording, and said out loud in the first 30 seconds
• Title: query in the first 40 characters, whole promise inside 60
• Description: the first 157 characters say what the video answers
• Chapters: minimum three, first at 00:00, each 10 seconds or longer
• Caption file: a corrected SRT uploaded, not the automatic track as it landed
• Tags: five to eight for misspellings and ambiguity, two minutes maximum
• Hashtags: two or three in the description, never anywhere near 15
• After publishing, watch click through by traffic source and the 30 second retention line

Doing this by hand on 30 videos a month is where the channel breaks

Put a clock on the checklist and the problem becomes obvious. Query research runs about 10 minutes, the title 5, the description opening plus chapters 15, correcting the caption file 10, tags and hashtags 5. That is roughly 45 minutes of metadata per video, after the editing is already finished. At 30 videos a month it is 22 hours, most of a working week.

That is the part FalconVid was built to remove. The pipeline produces the title, the description and the tag set for each video, and the caption track falls out of the narration itself, so the chapter list is a paste from a transcript that already exists. The thumbnail is generated in the same run, and it then publishes on the calendar you approved, across YouTube, Instagram, TikTok, Rumble and Facebook, in any of 63 narration languages.

What you approve is the calendar, which days, how many videos per day and the exact time of each, never a script every morning. The numbers deserve the quality mode attached, because the mode moves the bill far more than the plan does: a 12 minute long video costs 1,008 credits in economy, 8,676 in balanced and 26,760 in premium. Starter is $47 a month with 15,000 credits, which is 14 videos if you stay in economy the whole month, or exactly one if you shoot everything in balanced, so the realistic working plan is around 10 to 12 videos a month mixing economy with one in balanced. Pro is $97 with 30,000 credits, 29 economy videos, 5 channels and 5 concurrent generations, up to Scale at $997 with 317 economy videos a month, 50 channels and 50 concurrent generations, and every creation feature ships on every plan with a 7 day guarantee.

So the honest ending carries two numbers rather than one. Done by hand, those 45 minutes of metadata per video are themselves the ceiling: three videos a week costs you most of a working day every month, and 30 videos a month costs 22 hours, which is why almost everyone quietly stops correcting caption files somewhere around video fifteen. When the metadata is produced alongside the video, and the videos are generated in parallel instead of in a queue, 2 at a time on Starter and up to 50 on Scale, the checklist runs exactly the same on video 3 and on video 300. It is a channel pipeline and not a manual editor, so choosing the topic stays yours. The 22 hours of typing do not have to be.