I have been thinking about Twitter for many years. One of the interesting trends that many of us who share an interest in social networks have been tracking is the decline in engagement on Twitter.

Indeed this decline is not only evident from Twitter’s own metrics and reporting, but also to anyone who has been an active user of Twitter since the early days of the service.

Response Per Message Has Declined

In my own experience, Twitter used to be very different than it is today. It used to be more like an online community. In the early days of Twitter, when I Tweeted, I would get a lot of retweets, replies, and new followers.

Today, however, when I Tweet, I notice that the number of retweets, replies and new followers per message is much less than in the early days. It’s not that my content has changed (it really hasn’t), it’s that the way people use Twitter has changed.

Twitter’s Social Design Doesn’t Scale

In thinking about why this change has occurred, I have concluded that it is mainly a symptom of Twitter’s success in acquiring users. The social network design that underlies Twitter does not scale to a large audience.

The user-interface and interaction design of the Twitter Web site and the Twitter mobile app are also, for the most part, the same as they were in the early days. But the way people use Twitter, and the volume of content on Twitter, have outgrown these paradigms.

As Twitter has scaled over the years, each user has gradually followed more people on average. This has lead to social graph saturation — there is a huge amount of social overlap in the graph, meaning that people are more likely to get the same Tweet or news item, many times, from multiple people they follow. This leads to a lot of noise and redundancy in the content stream.

Social Graph Overlap Makes Discovery Harder

In addition, more automated bots and content sites have begun to post more content to Twitter per unit of time. The frequency of posting has increased. There is more content per unit time than before, and this continues to grow. But this is not necessarily good. More content also means more information overload.

These trends have collided, leading to a situation where the average daily number of messages that a typical Twitter user receives in their home timeline has grown dramatically.

Because of this growth in timeline message volume Twitter has become virtually unusable for discovery. Nobody can possibly keep up with all the messages in their home timelines (even in an hour, let alone a day).

A Less Efficient Publishing Channel

Secondly, Twitter has become less and less effective for publishing — at least if you want attention to what you post on Twitter. The probability that anyone will see or engage with anything you publish on Twitter seems to have declined dramatically (and this probability falls off very fast to zero, in a matter under an hour it seems).

The problem is that the follower graph in Twitter has reached a saturation point where it is almost irrelevant — following people has no benefit over not following them — the information overload in either case is just overwhelming.

Filter Failure Leads to Social Overload

When you follow hundreds to thousands of people and outlets, you get too many Tweets. It’s too irrelevant. It moves too fast. It’s simply unmanageable. There is no filter on the firehose anymore. The only solution is just to ignore it all. And that is what most people seem to be doing.

The filter used to be who we chose to follow – but that is no longer effective. Because even within that set there is just too much content coming into home timelines. As a result, in my own case at least, I almost never look at my home timeline in Twitter anymore.

Solutions Waiting in the Wings

Of course there are various ways Twitter could try to solve this. They range from for example, automatically ranking and prioritizing Tweets based on popularity, or how relevant and interesting they might be to me, or on some other metric (like how much someone paid for my attention etc.).

Any or all of these solutions in combination could help improve timeline signal-to-noise in Twitter. But so far I haven’t seen anything that solves it come about.

“Fire and Forget” Behaviors

Meanwhile, because of the declining signal-to-noise ratio, most people are using Twitter in a new mode. Whereas in the early days it was truly a conversational medium where people really paid attention to people they followed and engaged in dialogue with them, today it is more of a “fire-and-forget” medium where people simply post things into the aether, hoping that someone will see and at least Retweet them (which happens less and less).

Why is this happening? In a crowded room where everyone is shouting, the only way to be heard, even when you are talking to just a few people, is to shout even louder. And that is what I see happening on Twitter. More people and more publishers, posting more stuff, more often, in order to hopefully get noticed.

This is a self-amplifying feedback loop that results in total information overload eventually. If Twitter doesn’t solve it, their engagement will continue to fall. And at least if Twitter relies on advertising dollars from eyeballs, this is a serious problem for their business model.

The Third Party App Gap

Unfortunately, the decision made years ago, to stop all third-party innovation around the Twitter public API and eliminate third-party Twitter client apps, has made this situation worse, not better.

Although closing down the third-party Twitter app ecosystem gave Twitter more control over the advertising dollars on their content, it eliminated many apps and services that were actually helping to filter and personalize Twitter content. Ironically those same apps that were eliminated, were actually helping to sustain and grow higher engagement on Twitter.

Twitter has yet to fill this gap with their own apps and services — none of which currently solve the engagement and signal-to-noise problem effectively. But the potential is there.

It’s a bit of mystery to me at least why Twitter has not made solving this their top priority. There has been relatively little innovation or improvement to their core apps and services for many years now. Meanwhile Twitter has been acquiring companies that have little relevance to solving this problem.

Twitter Pivots From The Inside Out

It appears to me that Twitter may have shifted their strategy — they may have given up on improving internal engagement, and begun to just accept that they are more of a fire-and-forget medium going forward. In that reality, monetization strategies shift from internal to external opportunities.

For example, even if people no longer engage at all with each other inside of Twitter, the fact that they post teaches Twitter a lot about their interests, and this can be used to sell re-targeting and advertising intelligence outside of Twitter. In short, Twitter is sitting on an incredibly valuable personalization graph that they can monetize outside of Twitter.

It’s Not a Social Network, It’s An Ad Network!

Another way to monetize the content in Twitter, without increasing engagement, is to sell ads on manually or automatically curated subsets of the content, outside of Twitter. And we see this happening with Twitter’s recent move to enable their ads to run on their content in third-party sites.

But We Still Need a Social Network…

I continue to hope that Twitter will solve this with their own apps — with a new consumer experience designed for the reality of their much larger audience and super-saturated follower graph. I truly believe the world really needs Twitter, or something like it.

However, as the startup economy continues to show us, if Twitter does not solve it, someone and something else surely will.