YouTube has been recently experimenting with a discovery queue option of sorting by color. The earliest bubbles I can find for this is 6 months ago (Sept 1 2023). Interestingly, it’s resurfacing again, but with more pop:

Brushing past the whole YouTube PMs are smoking something entirely alien, it helps to take a step back and try to guess why this is happening.

What we know

Let’s establish a couple of things:

First, the scale of YT is enormous — 2.5B active users over the entire year means that a 1% user base is still 25M, which might be the total customer base for most of us here.

Second, Google/YT has been known to favor profit-enhancing features in lieu of user-friendly features. I cannot quote this, but a gander at the mid-tier YT content producer channels and their frequent issues with YTs platforms is enough to come to this conclusion. Add to the fact that they have to make money (more money than yesterday), it is obvious that such strategies are bound to be prioritized, if not outright mandated.

Still, we won’t know what happens in their MBRs/QBRs.

This, however, is interesting. I can think of two separate reasons why this whole ‘color’ based discovery queue is done. I’m also adding a popular internet theory.

Theory 1 — Occam’s Razor

One — and this is the Occam’s Razor at work — is that they have a metric problem. All other discovery improving methods are either too time consuming or have been utilized. This explains how they can, probably, run this as another way to improve the discovery to views metric.

Breaking this down in my favorite problem space — solution space mapping:

Problem Space

• Discovery — View CTR is low.
• Bounce from Home Screen is high.
• Discovery — View — Abandoned rate is high (people abandon the video that they started watching from the home screen)
  All of the above can be summed up with a nice Quality of Content score on the homepage, but I’d wager that they’re already doing something similar.
• Constant refreshes of the home page (which shows new content) without clicks is increasing.
• Scroll beyond a threshold (50% of the page or N pixels) is increasing.

One can argue that these are symptoms and not the core disease — which is content quality. But that’s another team’s KPI no?

Solution Space

• Improve discovery queue to match the users' current needs.
  Con: Heavy on server side — need to personalize based on time and behaviors — could be a hit or miss.
• Provide easy way to navigate to the next video if the first video is not meeting requirement
  (e.g. if a user wants to watch the latest Helldivers 2 update, but the video they clicked on is old, maybe find the latest one? OR find similar behaviors)
  Again, time consuming — the patterns are not easily visible and requires a tremendous amount of computing to achieve this.
• Provide new ways to filter from within the discovery queue (they already have some of it, so add new ways to filter/prioritize your discovery queue)

Within this solution space, using colors as a way to sort/filter the discovery queue might help the user discovery new content that they usually wouldn’t interact with… and that seems plausible — you’re giving the user a new way to engage with new content, giving them a sense of accomplishment that they found a new YouTuber or learned something new.

If the engagement through this discovery queue moves the metrics for the selected users, YT might be tempted to move it up to the general audience.

Theory 2 — Least Common Denominator

Second explanation is a simple one — No one opposed it. It’s straight from the left field, seemed like an easy thing to do, and hell, it made logical sense. It is likely that multiple approaches were proposed, and while everyone was fielding their own/favorite, this did not have many logical grounds to oppose. And hence won, because other approaches got more opposition than support.

Consider a team that is desperate to find new ways to improve the metrics (listed in the problem space) — they would agree to the first logically sounding, easy to implement solution that works in their favor. And hence, this could have been their way out.

It’s not a good look, but hey, if it solves their internal problems and they can focus on bigger things, good for them — as a consumer, I’m OK if they spend some time managing a user that’s not me. In the long run, I’d want to see features that benefit me.

Theory 3 — Theory of Color/Marketing Benefit

The internet has theorized that this is a possibility: ‘marketing uses color theory so this might help’. I reject it based on a simple ‘not-enough-evidence’ reason.

While it is possible, the dots-connection done to reach this a bit far-fetched. It might help to disguise the ads (yellow themed video, yellow themed ad — they merge? I don’t know how it feels connected or improves any metrics YT might be tracking. Or the alternative, Yellow themed video, Red colored ad — stands out and maybe has a better recall? I don’t know, but feels really far fetched).

Conclusion

Product Overview

There is something to be said for having a holistic view of the product and the impact each function would have on others — and the utilization of the resources to achieve the bigger platform engagement mantra:
quality content → better viewership → better ad engagement → more revenue more revenue to share with content creators → who create better content.

But then again, this is my outside-in perspective. The truth might be something else altogether — maybe YTs focus is to get more content, getting them more views, and hence, by volume, more ad views, leading to more revenue.

Nonetheless, it is not the team's fault that they were tasked with a small portion of the product and incentivized to increase their throughput in the KPIs tracked.

I’ll throw in an analogy in this already long article — we’ve all seen the machine cogs at work. Without load, it is easy to spin a section of the cogs and see movement in another. When load is applied, each cog has to overcome the overall interia of the cog-dependancies it has in the cycle ahead.

This might be the case here — YT has these different teams with such intricate inter-dependencies (both mapped and unmapped) that the pressure on each cog of the YT machine to spin in a certain direction and gain momentum might end up damaging the machine. The only way it can work is if each cog is tempered for output and weighed on the overall system performance scale.

Who knows, they might be already doing that! And this (the color filters) is an output from their end… the outcome remains to be seen.

Internet Sensationalization

The jump from ‘Hey! This is a possibility’ to ‘Damn! It actually does work’ isn’t that hard. In fact, in a different environment, it would be called out-of-the-box thinking. But sensationalizing things is the mantra to getting views online, and hence the pundits have declared this feature a blasphemy.

I’d encourage readers to consider both sides — and try to find reasons why things are the way they are without assigning stupidity or malice to the intentions. These types of analysis help you think in the light of possibilities, rather than in the darkness of presumptions.

While it might make you a better Product manager, it might also make you a better, more grounded human being.