Home Creators Posts Import Register Favorites Logout

Content

A new video is almost done. I'm looking to post this Friday or Saturday at the latest, just in time for everyone's favorite math-inspired day (and Einstein's birthday). Can you help me pick the title and thumbnail?

https://forms.gle/b2XDBjSCkVrkbbRH6

Why this is so important:

I've known for some time that titles and thumbnails are important, probably equally important to the quality of the video itself. But this was really hit home for me with the asteroids video. I posted it with the title: Asteroids: Earth's Biggest Threat

On the first few days, the video performed worse than 9 of my last 10. I tried slightly different titles and thumbnails but then on day 3 switched to: These Are The Asteroids To Worry About

After this, the video shot up in popularity and became the second most popular video of the previous 10. It went from a trajectory to get a bit over a million views to now 8+ million views.

The grey area shows the usual trajectory. The blue line is the asteroids video and you can clearly see when I made the change.

Here's where it is today. So I hope you can see just how important the right title and thumbnail are. It's funny that despite having a scientific background I've never tried to tackle this problem scientifically.

So with that in mind, I want to ask you to help me pick the next title and thumbnail. Just select your favorite here: https://forms.gle/b2XDBjSCkVrkbbRH6

If you have ideas about how I could be more systematic with this testing going forward, I would love to get your thoughts - please leave them in the comments.

And thank you!

Derek

Files

Can you help pick the next title and thumbnail?

Comments

Simon Marquardt

The more systematic approach would be real A/B testing, not sure wether YouTube supports that. For a questionnaire as a participant I like the kind of question where you get two random options and only have to decide between the two of them which one is better, and get a new combination to rate afterwards until you’ve rated all or lose your interest. It’s much easier to decide between two options than between eight. Tom Scott used this kind of question to determine the best thing: https://www.youtube.com/watch?v=ALy6e7GbDRQ&t=275s

Robert Blum

We’re always happy to be your test/ focus group.

Kimberly Green

To be systematic, I recommend start by analyzing other successful channels in your domain. Look at their view counts and find out which of their videos are doing the best. From there, look for common themes, patterns etc with both the thumbnails and titles. Also check to see if they are leaving searchable keywords, etc in their descriptions to improve searchability. Also, keep in mind that YouTube animates thumbnails now. Post the work-in-progress video to a hidden channel first and watch how the thumbnail gets animated. See if that makes it look more or less interesting & likely to draw view attention.

ReaperNX

This came into my e-mail as the thumbnail discussion came up in the video analyzing why the reservoir balls video went viral I happened to be watching. Weird.

Veritasium

Fantastic suggestion. YouTube has been exploring A/B testing but have not implemented yet. I think it would be a godsend to people like me who actually (empirically) don't have great instincts for picking titles and thumbnails. What you suggest - getting a bunch of people to do A vs B as long as they are interested is a really good suggestion.

Veritasium

Thanks Robert. I'm already enthralled by the results. I would not have ranked them the same way as they are currently being ranked.

Veritasium

Great points - Youtube is grabbing gifs from within the video and putting them with the thumbnail - sometimes they try to select something similar to the thumbnail, but other times they just pick a random portion of the video. Plus on phones, YT prerolls the video as you're scrolling so the beginning of the video is important.

Tristan Ursell

agreed, A/B testing ftw -- can you clone the video itself and post different thumbnails / titles and measure initial hit-rates? Then tournament select? This also brings up the interesting question: if title / thumbnail is so important, it might make sense to clone a video with different title/thumbnail and thus attract a broader audience ... (i.e. those that are attracted to A *and* those that are attcacted to B). But youtube might not allow that.

Simon Marquardt

Additional idea: to get a better focus group it might be useful to cooperate with other Youtubers and their fanbases and have thumbnail ideas for different topics compete within a broader audience. I guess other creators have similar problems, and to get a video viral(-ish) it’s probably not enough to resonate with just the nerds that made it into this patreon feed.

StealthSecrecy

I think I'm unlikely to notice any of the options in my feed because the colour scheme is just bland compared to other videos. When browsing for something to watch I usually just scroll until something bright catches my eye, and then I actually read the thumbnail and title. Out of all of the options H is the most appealing to me, but I think some colour could help it stand out. Never realized how hard designing a thumbnail was before

Matt Hart

i left a comment about potentially developing a mini a/b testing script, but looks like there are options that already exist. tubebuddy seems to be an option that people like

Dragi Raos

This "Asteroids" anecdote is just sad. FWIW, I prefer the first thumbnail. As for deciding what to watch, it is simple: for 20-ish creators I am a Patreon patron of, I watch everything. (the sam holds for further 20 or so for whom my Patreon budget was, sadly, exhausted.)

Quasitling

I mentioned this in the form, but it would help to see the thumbnails surrounded by other creator's thumbnails like they would be seen on Youtube. It's hard to tell which ones stand out when they are all compared to similar options all for the same video.

Quasitling

Cloning the video would look too much like a content farm post. Plus it could split the viewership between the videos, hurting all of them.

Chris Stone

The asteroids video is really interesting. I accidentally avoided it, seeing a typical "clickbait" style thumbnail that a information-sparce creator might use. The second one really appeals to my more scientific interests, I'm surprised your audience is comprised to heavily of people with that persuasion.

Veritasium

yeah, its utility mainly comes when a video is old, in the back catalogue - in the early days views are too variable hour to hour and day to day

Matt Hart

nothing will be ideal in the first few hours, so i guess the question becomes less "how do i find the best thumbnail/title pair as quickly as possible?" and more "is it worth it to get the marginal edge that i can from the limited data, at least weeding out some of the thumbnail/title pairs that are bad out the gate?"

Matt Hart

the best solution is for youtube to have a native solution that lets you test all your possibilities concurrently so they're subject to most of the same variables

Dion

Will you share the final ranking?

Some Dork

I’d love to see a video on just this topic tbh, keep collecting more data on thumbnails and you’d have some interesting stuff to talk about. Also I’m not the first to suggest this but I think you’d benefit from having a unique thumbnail style that makes each thumbnail instantly recognisable as yours, such as a border with a channel logo or a consistent style. Sometimes I don’t realise a video in my recommended is from you and I forget to watch it, even though I know every single video you make is a banger.

Mik Scheper

I can't imagine any subscriber clicking past any of your videos, so I wonder if this is really an exercise in attracting human attention, and not about influencing an algorithm somehow. Are you able to see how many people _saw_ the thumbnail? (And, just playing devils advocate, was the increase in the last video around the time those people in the UK found a meteorite in their driveway?)

Mik Scheper

This makes a lot of sense. Maybe just the logo on a white background. I can imagine my brain learning 'oooh, white square, must click ⬜'.

Marco Trevisan

Just a warning about the voting system: you provided so many options, that the winning option might not be giving the most successful title and thumbnail. Similar options will compete with each other (imagine that 70% of people in a group like classical music and 30% like pop, but you give 100 alternatives for classical music and just two for pop music... the winner will be a popstar, even though most would love Mozart). I suggest thinking about Single Transferable Vote to tackle this. This video explains it in one (very insightful) minute https://www.youtube.com/watch?v=ebqYVlB0UP0

Marco Trevisan

There are many options to make STV polls online, if you are interested. I like Demo Choice (I have no relationship with it, I am just a casual user), here https://www.demochoice.org

TheEdgeOfRage

I think that the first asteroids one looks *too* clickbaity, so people shied away from it. There's a fine balance between interesting and obnoxious and something as sensationalist as that completely unrealistically large asteroid is likely on the obnoxious part of the scale.

Dragi Raos

Yeah, you might be right.

VenkataSainath Ravikanti

"Ridiculous or Revolution? The story how Newton proposed 'Pi' calculation" I think this title is good and attention grabber.