Hello Greg. Firstly, kudos to you and the rest of the Medium team for your work.

Medium’s original vision fills a huge need in our modern thought exchange.

The aspiration to be a platform where “it is not about who you are or who you know, but about what you have to say” is a necessity for all the users that have gravitated towards this platform.

In choosing Medium in the midst of other blogging, posting, tweeting, snapping, friending platforms, the users of Medium, including myself, are on it for one overarching reason:

The vision of a platform built on quality words and quality thoughts resonates deeply with us.

Medium is an oasis of fresh thought for those dying of thirst in a sea of saltwater clickbait.

However, I believe this vision is at risk and I urge you and your team to think deeply about the implications.

This is not a result of insufficient effort or ill will, far from it. You and the team have built a phenomenal platform that is intuitive, modern, clean, powerful, collaborative and responsive.

However, to truly achieve its goal, I believe Medium must overcome a dangerous flaw in its design.

This flaw is core to the issue that you touched on in this comment.

Discovery.

Central to Medium’s mission is the concept of quality.

What makes a quality article? How do we then bubble up a quality article to a user who would find it most relevant? How do we connect a thoughtful writer with a thoughtful reader more often than not?

To date, as you described, this discovery mechanism relies heavily on the inputs of Recommendations and Follows.

If one were to scan the breadth of Internet paradigms. The “follow” and “like” paradigms, as embodied by Twitter and Facebook, respectively, have, for all intents and purposes, been wildly successful.

In this sense, the concepts of Recs and Follows would seem almost to be no-brainers. In the Darwinian evolution of internet paradigms, these two have emerged as kings of the jungle.

All this is true, except for one thing:

Medium was not founded to be a Twitter or a Facebook.

I put myself at risk to say that it was instead founded to be intentionally and dramatically distinct from either of those vehicles in almost all facets.

For design, Medium is sparse where others are dense.
For content, Medium promotes depth where others promote frequency.
For intent, Medium pursues quality where others pursue growth.

On all these points, Medium has been successful in its efforts which is no small feat.

So why then, when it comes to the life blood of the platform itself, its Discovery mechanism, does Medium choose to rely on principles that are so similar to those it wishes to distance itself from?

So what are the specific issues at hand with Follows and Recs? Let’s take them one at a time.

Follows (The “Twitter” Paradigm)

Follows inherently are not flawed. It makes sense to be able to select other writers that you find interesting and relevant. After all, this is the raison d’etre of Medium, to connect readers and writers and vice versa.

However, in execution, there are two areas of improvement in my opinion.

First, it seems that the weighting in terms of which articles are shown by followed writers is seemingly random. Rather than showing the articles written by those you follow you largely see articles that are recommended by those you follow. This may be relevant in some cases but is not nearly as relevant to the reader as something written directly by someone you have selected.

Suggestion: There should be an option or ability to toggle the weighting of priority given to articles written by followed authors rather than articles recommended by followed authors. Perhaps there needs to be a dedicated channel or publication for the user that strictly shows content created by followed authors. I believe this would prove hugely relevant and useful for your users (both readers and writers).

Second, Follows as currently defined are extremely one-dimensional. It clearly follows the Twitter paradigm where you effectively subscribe to a user and that is that.

In my opinion, this behavior does not serve the purpose of Medium’s mission because it does not reflect the depth of relationships that can form between readers and writers. There are varying degrees of engagement between readers and writers whereas the “follow” only offers a binary option. On or off.

Suggestion: Clearly, putting in a “4 star rating” system for writers is a complete non-starter. However, perhaps there are more subtle ways to measure the level of engagement between readers and writers.

For example, once you follow someone, why can’t the algorithm also take into account the frequency with which the reader and writer (or two writers ideally) open a dialogue and respond to each other? Clearly, with the “every comment is an article” backbone of Medium already set up, this is something that can be tracked and quantified. If this measure were then laid on top of followed users as an additional metric of relevance, you may not only be able to separate “passive” follows from “engaged” follows but, given the distribution of exchanges between various pairs of users, you can also begin to measure on a curve the more subtle gradients of engagement.

For example, content from a user that I have exchanged dozens of responses with may very well be more relevant to me than content from a writer that I have never responded to or has never responded to me.

What I am getting at here is that the Follow as currently set up does not do what I believe you want it to, which is to log a relationship between two users.

Instead, it is currently set up as a measure of “subscriptions” rather than relationships. Within this paradigm, it is then no surprise that your users are beginning to sacrifice quality of thought and writing in order to “hack the system” and amass the largest numbers of subscriptions possible.

While on the surface this has the appearance of engagement, I would in fact argue it is detrimental to your primary goal because it is a clear incentive for your user base to sacrifice quality (thoughtful and personal writing) for quantity (clickbait to draw eyeballs).

Once this phenomenon hits a critical mass, I fear Medium will begin to hit a vicious spiral downward where the original vision of a platform based on “what you have to say” no longer rings true.

Recommends (The “Facebook” Paradigm)

Although recommends are the much more troublesome metric in my opinion, I don’t want to belabor it too much as this is already running long (and I wonder whether anyone on the Medium team will actually want to sit through this thing) and also because I touch upon it in my other article.

Suffice it to say, that recommends are hugely insufficient to measure either quality or engagement.

Several reasons for this: First, I would guess there is an extremely high correlation between recommends and the number of followers a user has (you guys can probably run the numbers on this if you haven’t already). As I mentioned in the other article, like a game of poker, this rewards those who have already amassed large warchests of followers vs. those who have just “bought in” to the game and sat down with their measly stack of chips.

This phenomenon is hugely disadvantageous to new users and will, over time, create an increasing level of dissonance between what you present as quality (articles showing up in a person’s feed with high recs) vs what is actually quality (thoughtful article that is uniquely relevant to that user).

This dissonance, once it reaches a certain threshold will, hate to say it, either kill Medium or, at a minimum, change what it stands for in the eyes of its user base. Not sure there is a difference between the two.

That said, I posed a solution for this in my other article which basically qualifies the recommends based on the number of followers a user has. Bit of a handicap in golfing terms if you will.

Second, recommends are only a snapshot and, similar to my comment on Follows, are extremely one-dimensional as a result. Binary. On or Off. Like or not. Again, this does not pay due respect to the varying degrees of engagement that a user may have to a particular piece.

Again, I understand the realities of product development that you need to serve the lowest common denominator. Introduction of likes, dislikes, sorta likes, kinda likes would wreak havoc with your UX and with your data.

However, again, perhaps there are more subtle ways to measure engagement. You may already be doing this, but why not include in your algorithm the number of highlights that an article receives as well as simple recommends. Or, to steal a page (no pun intended) from google and track the number of links that others include in their articles to the one in question.

I believe these paradigms would prove to be much more relevant to the user and over time, in combination with a more gradated measure of relationships, would really begin to bolster the relevance of a user’s feed.

Finally, I am of the (perhaps extreme) opinion of getting rid of recommends altogether. Or, to steal a playbook from Reddit, to obfuscate it behind a largely meaningless and nonlinear quotient. It could give some indication of relevance to the user but couldn’t be directly tied or compared between articles.

This would avoid the final issue with recommendations that it is human nature for people to interpret these numbers as a “score”. High recs = high score = relevant.

I believe Medium needs to break out of this behavior on both sides of the fence, to address it within its algorithms as well as address it in terms of its users’ behaviors and tendencies. Save us from ourselves!

Man, I gotta wrap this up.

This post has dragged on a lot longer than I had anticipated. Sorry for that. But please take this as encouragement rather than a rant.

I wouldn’t bother to write this or to think about this stuff if I didn’t think Medium was tantalizingly close to having a breakthrough.

If Medium is truly able to transcend prior paradigms that have come to define our Internet lives, it would not only be a positive influence on the conscious development of the Medium community but could also redefine the very concepts of “quality” that are so inadequate and yet so pervasive in our society.

In my personal opinion, this misunderstanding and misrepresentation of what quality is is, above all else, the imminent crisis that faces not only Medium, but all of us.