Cultural Content - Good YouTube, Bad YouTube

If you're using YouTube mainly as free storage space, read on...

Good morning partners in content,

Today I’m writing up some thoughts off the back of a really interesting project we’ve wrapped up with the National Archives on YouTube strategy.

As many of you will be aware, we at One Further are not videographers. We do however have a lot of data, and like to think we know how to use it.

So, here’s our 1,000 word summary of what we found out working with cultural sector YouTube data, the Wayback Machine, and Our Thinking Caps.

---

YouTube probably ticks your organisation goals

A lot of organisational audience (and digital) strategies at the moment want:

• MORE AUDIENCES

• MORE DIVERSE AUDIENCES

• MORE MEANINGFUL ENGAGEMENT

YouTube is a pretty good channel for all of these reasons.

Some lesser known facts about YouTube:

• Like TikTok, YouTube is built off of a recommendation engine rather than a social graph (c.f. New Yorker interesting article about the difference)

• This means that most people who see your content won’t follow your page (most cultural sector YouTube’s get upwards of 96% of their views from non followers). Instead rather an algorithm will try and pair content you create with audiences around the world that it thinks might be interested in it

• Both TikTok and YouTube (the recommendation engines) are better than traditional social media (Facebook, Twitter, Instagram) at getting your content out in front of a global audience who hasn’t necessarily already heard of you

• …which make it a good fit with strategic objectives of audience growth and diversity

It’s also a good fit for more meaningful engagement; if you spend 30mins watching a video you’re very likely to come away with a higher level of intellectual/emotional engagement than if you scrolled past a pretty picture/ hackneyed meme in a feed.

---

But the reality is most cultural organisations use YouTube a bit like Sharepoint…

Whilst YouTube has the potential to grow you diverse audiences and meaningful engagements, that’s not going to happen overnight, particularly if you use YouTube in a similar way to a floppy disk…

Most cultural organisations use YouTube a bit like a video version of Flickr, it’s a repository and handy space for hosting a grab bag of video content from different departments online.

Suffice to say, this approach is not going to make you a YouTube superstar

But why not?

---

How does the YouTube algorithm work?

At the end of the day, YouTube is a business, and one that’s been growing and evolving for over 16 years.

It’s in the business of monetising your attention. The longer you’re on YouTube, viewing ads, the more money it gets.

The algorithm wants to understand what the content creators make is about, so that they can match it with users who have similar interests.

YouTube did great numbers between 2020 and 2021:

Source: Statista

So, how does YouTube do it? How does it keep people in the platform wanting more?

To explain this section we’re going to go back in time and look at YouTube back in the day and how it worked, and how it works now

2005-2012: All About Views

Look at this adorable picture of YouTube c. 2005 (thanks WayBack Machine).

What’s interesting here is that in 2005 YouTube looked a lot more like Google. Most of the landing page is taken up by a search bar, and – much like Google, it relied a lot more on the user’s input to surface content.

In 2005-2012 the algorithm largely prioritised videos with the most views, this incentivised creators to make videos that look interesting (and it didn’t hugely matter if those videos didn’t deliver)… which resulted in a lot of clickbait… which in turn frustrated users… leading to less time in the platform… and so less revenue for YouTube… 👎

2012: Watch Time

Around 2012 YouTube changed its algorithm to prioritise watch time.

These algorithmic changes led to creators prioritising content with a higher watch time. This was better from YouTube’s perspective (in terms of keeping people in the platform longer), but there was still a problem with retention at the end of a video…

2016: Machine Learning

In 2016 YouTube published a paper on how it was using Machine Learning to update its algorithm. It contains this somewhat opaque diagram on how they choose a particular video for a particular feed:

Essentially, from 2016 Machine Learning and neural networks were fed into the algorithm to create a recommendation engine for subsequent video views.

What that means today is that (if you’re logged in to YouTube) your ‘feed’ - what you see when you land on YouTube - will differ person to person:

Side by side view of YouTube from my personal gmail, and from the One Further gmail account

YouTube today recommends videos based on what it thinks the user is interested in – based on a user’s history and similar content.

What this means for your YouTube content creation

YouTube factors in rankings like the watch time, impressions CTR and view count of your videos to give it a sense of how ‘good’ your content is for a general audience.

It’s got to try and work out what subjects your video covers, so make sure you’re using well optimised keywords that correspond with terms people are actually searching for. It’s also really helpful if your description field gives the full context of the video and doesn’t just assume you’ll already know what - for example - a particular promotional trailer refers to.

Formats are great YouTube content because they lend themselves to YouTube recommending a series of the same content type after the first one is watched (giving each video in the series a cumulative halo effect).

Formats

Here’s a couple of examples of YouTube formats. By ‘format’ I mean a recurring video type; with a custom branded thumbnail, recurring title text and narrative structure to the video.

A great many YouTubers - like Amelia Dimoldenberg - have launched very successful YouTube formats from the ground up. Legacy print publishing houses (like Vanity Fair) have upped their game to match; designing unique YouTube formats they can create to bolster their presence amongst a YouTube audience.

Hot Ones is an interesting case in point, at the point I took this screenshot (below), I’d never sat down and watched a Hot Ones (hat tip to

Chris Unitt

for putting me onto it), but what’s interesting is that every other video in the recommended sidebar on the right is a Hot Ones video. So, although I’d never watched one, evidently lots of people have – and it seems likely that lots of people don’t just watch one, but then go on to watch a string of other Hot Ones videos…

This is why formats are such a good fit with YouTube’s current algorithm. YouTube is trying to nail what you might want to watch next, to keep people in the platform longer. Formats are gold dust in this regard as YouTube has very good criteria for recommending you something similar to watch afterwards. This also means that all videos in the format have a halo affect of being associated with something of a similar type.

How can YouTube understand what your content is about?

YouTube has a couple of different data sources that you can readily control:

• Title

• Description

• Tags

• Thumbnail

If you’re not optimising your videos for the right keywords, no one will find you.

There are a couple of free sources of keyword data. YouTube provides you with the terms your viewers currently search for and the top performing keywords people have entered to land on your content. However both of these let you see what’s working most amongst what you’re currently doing – neither gives you granular enough data to rearchitect your strategy around high volume low competition terms that correspond with the expertise your institution holds.

Competitive search intelligence tools (we use SEMRush, but others exist) give you this data (but come with a price tag).

In terms of tags, YouTube themselves say they’re not that interested in tags, if there are alternate spellings of keywords in your title and description, it’s worth putting them in tags, but in general YouTube isn’t paying this content field a tonne of attention.

Thumbnails

There is a lot written on the internet about YouTube thumbnails. The most followed YouTuber in the world is Mr Beast. Specifically, there is a lot written on the internet about how much Mr Beast spends on his heavily stylised and saturated thumbnails:

Upwards of $10,000 for each video (for the thumbnail alone) is the general consensus…

This video looks at 200 Mr Beast thumbnails to determine what he’s doing that’s working so effectively at converting impressions to clicks…

The video essentially concludes that thumbnail best practice is:

• 0-4 words as text in the thumbnail itself

• The image should communicate a single idea, which – in tandem with the title – should hint at intrigue/ suffering /a big reveal, but not quite communicate the outcome behind the premise

• It does seem to help to have a recognisable ‘face’ to your YouTube content

• Services like TubeBuddy allow you to A/B test different thumbnails

Interestingly these thumbnail findings overlap with a really interesting Twitter thread on how Netflix has optimised its thumbnails for impressions Click Through Rate.

Essentially Netflix found that the ‘greatest lever’ in converting browsing users to viewers was how compelling the content thumbnails were.

Netflix originally used promotional metadata images, frequently repurposes from show posters and billboard signs, but what they found was that these didn’t work well in a streaming context.

So they built a ginormous algorithm to score millions of frames within video content and score each frame for things like brightness, nudity potential, composition – and used this to generate their own thumbnails for shows

Ultimately what they found was that Expressive Traits, Main Characters and Brightness were winning traits (so quite similar to the Mr Beast Thumbnail analysis findings)

To sum up….

• YouTube accounts that are growing diverse global engaged audiences exist in the intersection here:

• But most of the cultural sector is here:

• YouTube is a smart and frequently underused platform in the sector, it’s great for growing audiences, reaching a diverse global audience for the topics you can cover, and improving the depth of engagement with those audiences

• Formats are a really good fit with YouTube’s current algorithm as they give the platform a really good steer as to what to recommend someone views next (another video in the same series)

• Much more work can and should be done in the sector in terms of working out where search opportunities (amongst the topics you have expertise in) exist and using that data to build a video strategy

• It’s not uncommon for cultural sector videos to use a video still as a thumbnail, this will almost definitely be killing your impressions Click Through Rate, building an arresting thumbnail is a really important part of the appeal and storytelling of YouTube content (like a book cover)

Want to know more? How does this chime with your experiences of running a YouTube platform? Get in touch (georgina@onefurther.com), or comment below