The algorithmic reward table

This article lists and discusses the different types of signals user platforms measure to show the right content to the right users.

When are really just two hands full of platforms to build an audience. Interestingly, these platforms reward different types of signals. Most social networks, for example, look at a combination of engagement, proximity (the distance between two users on the user graph), and relevance. But not every platform measures the same signals. And, not every platform is open about the way they measure them.

User platform algorithms have the job of making the most relevant content discoverable. That's directly related to the user experience: the better the content, the better the experience, the more money the platform makes. As the platform grows, the mass of content becomes too large for users to sort through it. It needs to be sorted. Enter algorithms.

Search engines have "ranking" signals because they rank results for everyone who looks for the same keyword. The differences from user to user are small, though Google does personalize search results to an extend. But on all other platforms, the idea of a ranking signal relates more to reach or visibility, i.e. how many people see a post.

When we understand the signals platforms use to score behavior and content, we can create better content and reach more users. The algorithmic reward table is a comparison of how these signals are used by different platforms (or not).

The algorithmic reward table

PlatformRecencyRelevanceEngagementProximityPopularityContent formatQualityMeta-data
Google Searchxxxxxx
Tik Tokxxxxxxx

Signals definition

Let's start by defining the signals different algorithms measure:

  • Recency: freshness, a.k.a. how recently content was published
  • Relevance: how relevant the content is based on the user's intent or interest
  • Engagement: How much engagement a post or piece of content needs to have a high visibility
  • Proximity: How close the users need to be to the user or brand that's posting the content
  • Popularity: How popular the user or brand is within the network
  • Content format: text, video, audio, or live-streaming
  • Quality: how good the quality of the post is (measured in different ways)
  • Meta-data: hashtags, titles, descriptions, captions

I decided not to list what signals algorithms punish because they're all the same:

  • Spam
  • Scams and fraud
  • Pornographic and violent content
  • Low quality content
  • Misleading content

Note that there might be more signals that these algorithms take into account to a small degree. Security (SSL encryption) is a small ranking factor for Google Search, for example. But it doesn't make sense to incorporate them here because they exist at the margins.

Also, I considered the core product for each platform. Stories, as you can find them on Instagram, Youtube, or Twitter are driven by different algorithms. So is Google Discover. For simplicity's sake, I decided not to take those into account.

Lastly, there is something to be said about negative signals, such as links that take users outside their platform. It's not proven yet, but these "walled gardens" want to keep users, not send them away. So, while I can't say outgoing links are "bad" or punished, I have a suspicion they might be decreased in reach. This is different from "punishment", though, which can lead to complete disqualification of content.


There are some counter-intuitive takeaways from comparing platform algorithms.

First, it's interesting that not all algorithms reward the same signals, even when comparing the same type of platform. Facebook, Twitter, and Linkedin give more reach to live-streams > videos > images > text. Youtube or Tik Tok, of course, are video-based and therefore only provide one format. But, while Twitter doesn't measure the quality of a post, Linkedin does. I can post anything on Twitter, and my reach will depend on engagement. For Linkedin, that's not enough. You could also say that they use different means to determine quality, but there are differences at the end of the day.

Second, most platforms measure engagement in some way: watch time, likes, shares, or comments. Even Linkedin measures dwell time. Google Search is the only platform that doesn't, at least to the extend we know (some folks will argue with long vs. short clicks and CTR as ranking signals, but let's stay out of these waters for this one).

Third, gaming algorithms is much harder when engagement is measure in several ways. CTR + watch time + likes is hard to fake. You can try to catch people's attention with bold statements, but you won't succeed if you can't keep it. In Google Search, however, you can still optimize many factors outside the actual content to rank higher.

Fourth, different platforms vary in their degrees of openness about their algorithms. Linkedin, for example, describes their ranking signals and algorithms in-depth on their engineering blog. Google, on the other hand, guards the core Search algorithm like the Coca Cola recipe.