YouTube SEO for channels where search is 15% of views

On the channels we audit, YouTube search delivers between 8 and 20 percent of views. Suggested and Browse deliver most of the rest. That ratio sets the priorities: metadata is worth getting right and it will not rescue a channel, so the bulk of an engagement goes to the two things that move Suggested — what a video promises at the packaging level, and whether the people who clicked are still watching at second thirty. What we look at first: The full catalogue through the YouTube Data and Analytics APIs, with every metric split by traffic source. Channel-average click-through rate is a number that means nothing. Four percent from Browse and four percent from search are different diagnoses requiring different fixes. Then the retention curves, aligned at zero, first thirty seconds only. In gaming and fitness the videos that fail there fail for a reason we can name out loud: the thing promised in the title arrives at 4:20 and nobody waits for it. Packaging: Titles and thumbnails, tested rather than argued about in a meeting. Where the channel qualifies for Test and Compare we run three thumbnails against one title and read it at significance, not on day two. Where it does not qualify, we change one variable and hold everything else still for fourteen days. Old videos get retitled deliberately and rarely, because a title change resets how the recommendation system understands a video, and doing forty at once destroys the data you would have learned from. The part that genuinely is search: Chapters, written by hand, because they produce key moments in Google's video results and they are the only mechanism that makes a video addressable at a timestamp. Uploaded SRT subtitles rather than automatic captions, which matters more in Swedish than in English and matters most for product and character names the recogniser mangles. The first 157 characters of the description, since that is what gets shown. Playlist architecture built so a series has one entry point rather than eleven equally weighted episodes competing with each other. Demand comes from YouTube's own autosuggest and the search terms report inside Analytics, not from third-party volume estimates, which for YouTube are guesses dressed as data. Not included: We do not design thumbnails or edit video. We art-direct and reject, and we will tell your editor exactly which frame in the first thirty seconds is losing the audience, but production stays yours. We do not buy views, run engagement pods, or sell a tag audit. Tags have been close to irrelevant for years and anyone charging for them is charging for theatre. We do not run community management, and Shorts get a recommendation rather than a separate product. Who this is not for: Channels publishing less than twice a month. There is nothing to optimise; publish more and call us in six months. Brand channels that use YouTube as ad storage, where the retention problem is the creative brief and no amount of metadata will fix a commercial. Anyone who wants subscriber count as the headline metric — we report watch time, returning viewers and the search-versus-suggested split, because those predict the next video's reach and subscribers do not. What you get: A catalogue verdict, video by video: which are worth retitling, which are worth reworking, and which to leave alone. A packaging test plan with a schedule and fixed read dates. A chapter and description standard your editor follows without us in the room. And a monthly read on whether the changes moved the specific traffic source we said they would, which is the only defensible way to know if any of it worked.