YouTube Algorithm Questions Answered

Avatar photo
April 15, 2020
YouTube: Algorithm Questions Answered

Have you been curious to learn more about how the “YouTube Algorithm” works? Todd and Rachel answer 6 common questions about the Search and Discovery system.

Question 1: Shouldn’t watch time averages be compared in % rather than actual time due to varying lengths of videos? (This is referring to the snapshot card in the Studio dashboard.)
Todd states that both absolute watch time and percentage viewed are used for the Search and Discovery system, but they obviously can’t fit everything in the snapshot card. He notes that there’s pros and cons to both approaches so both factors are considered for evaluation of performance. It’s easier to get viewers to watch a higher percentage of a short video, but that doesn’t make it weigh the same as those who watch that same percentage of a much longer video.
Question 2: Myth – better to post early in the day vs later at night?
Rachel says that creators shouldn’t focus so much on upload time. Chronological subscription feeds aren’t the primary consideration of when a viewer watches videos anymore. The homepage also doesn’t prioritize new videos that strongly.
(Epos’s notes, not Rachel’s answer: That being said, there’s definitely been strong evidence for the first way of “view velocity” impacting how well a video performs, even if in the short term, so IMO that should still be considered.)
Weirdly enough, the description has a note that says “When it comes to notifications, do try to upload when most of your audience is awake. Some users choose to mute notifications on their devices at night.” Which… completely contradicts the original answer given – notification CTR within some undisclosed block of time can definitely impact a video’s performance.
Question 3: Does YouTube consider a viewer finishing 1 1-hour video or 3 20-minute videos more? Is all watch time created equal?
Todd states that they don’t believe all watch time is created equal. Viewers appreciate different kinds of watch time differently, and more is not always completely better. Multiple 20-minute videos might attract more viewers and keep them around more, while a 1-hour long video is a bigger investment and can be a turnoff for some.
Question 4: Does sharing a video while it’s Unlisted before making it live hurt its performance?
No. Performance is only measured starting when the video goes public.
Obligatory mention that uploading as Unlisted for a couple hours before going live allows time for Content ID and the automated advertiser-friendly guidelines systems to scan your video and give you any possible flags.
Question 5: Myth – You need to upload daily or once per week?
There’s no “rule” in the search and discovery system that looks for frequency of posting. However, there is an inherent benefit to more frequent posting due to it providing more opportunities to reach viewers and to be found. Striking a balance of quality vs quantity and finding what makes sense for your content type is most important.
Question 6: Does changing video metadata (title, description, thumbnail) re-rank or reset the video’s performance? Is there a risk?
Videos are not manually re-ranked based on metadata changes. Creators are encouraged to tweak metadata of poorly-performing videos to try to give them a better CTR after they’ve been posted. Old videos can be revitalized.
However, keep an eye on a drop in audience retention after changing metadata, confusing viewers into re-watching (and then leaving) an older video they’ve already seen may have a negative impact.
Hope this helps you understand the “algorithm” a bit better!
Tech educator, ’90s and ’00s Nostalgia Nerd, Pixel and Framerate Junkie. 12 years on YouTube is a loooong time.