Reverse network effects: Is twitter losing its mojo?

We’re all familiar with the concept of network effects. We hear it used all the time to explain the success of companies like Google, Facebook, Twitter etc. As these networks scale, the value of the network to its users increases. The bigger the network, the more successful these platforms are, right?

Maybe Not! As a network scales, network effects make the network more useful for users until a point after which further scale makes the network less useful to the users. This is reverse network effects with scale. There are several cases where there are reverse network effects WITHOUT scale. E.g. Users on Facebook do not like advertising so having advertisers on the network makes it a poor experience for users. However, this is not an effect that kicks in only after a certain scale is reached. Even a small number of advertisers on the network can be equally annoying.

When reverse network effects occur with scale, the value of the network keeps on increasing at first, till it reaches a point where new user addition decreases the value of the network.

Understanding Twitter as a network

The term ‘social network’ has become extremely generic and is used to describe a wide range of platforms which have little in common except that they’re all networks of users. Twitter, as a network, can be broken down into the following:

1. Communication Network: Users interact with other users, either one-to-one or one-to-many. Twitter, essentially is IM with persistence when seen in a communication paradigm.
2. Publishing Network: Users create content that is consumed by other users. Twitter is the new face of publishing and in lowering the barriers to creating content, it opened up a new market of publishers who had never published before. Publishing networks, typically, have a wide range of contributors with varying degrees of quality and need effective curation tools and algorithms to maximize the relevance of content that is surfaced for every user.
3. Influence Network: The one-way relationship that has grown to characterize Twitter allows users to gain followers without any obligation to reciprocate. The followers-to-following ratio has become the default vital statistic for Twitter. In influence networks, some or all active users are driven by a desire to maximize influence on followers or leadership on a leaderboard. Such networks tend to get inadvertently biased against news users after reaching a certain scale, as we shall soon demonstrate.

 

Communication: Size and Dunbar’s Number

There’s been a lot of discussion lately about how social networks are getting too big for us to feasibly maintain relationships with all our contacts. Dunbar’s number is often quoted as the evidence for the case. Even if this does apply to symmetrical relationship networks (though I feel, even in that case, it doesn’t), it’s a poor criticism for a network like Twitter for a couple of reasons

1) Users have a primary relationship with their feed, rather than with every other user they follow.

2) There is no obligation to maintain user-user relationships. Irrespective of the number of people in your network, users only maintain their relationship with the feed. This has other problems, though, which are mentioned below.

3) Dunbar’s number applies to contexts where relationships require significant investment, especially real world interactions. Fringe relationships can easily exist online without the continued need for investment or reciprocation.

In general, Twitter as a communication network doesn’t seem to struggle with reverse network effects.

Publishing: Signal-To-Noise Ratio

Publishing platforms, where users create persistent content, run the risk of generating noise. Successful platforms with high user engagement typically do a good job of curating user-generated content through algorithms and community-driven tools (voting, rating, reporting, etc.).

As a publishing network, Twitter has strong network effects as more users on the network translates to more options for feed sources.

However, what distinguishes and makes Twitter so user-friendly could also be its bane as a publishing network. Twitter gives inordinate importance to recency over relevance. Unlike Reddit, Twitter gives precedence to current over important. Consequently, a user’s realtime feed always displays the most current information irrespective of relevance.

The problem is that as a user follows more users, the reverse-chronological stream ensures that the user sees activity from only a fraction of users who are active at the time of her log in. However, this fractional stream may not be the most relevant. Attention is the single most important currency on online platforms and users are unable to appropriately allocate attention when there is too much noise.

This problem is further compounded when networks like Facebook and Twitter implement monetization models like Promoted Posts/Tweets which potentially decrease the signal to noise ratio on the network. These models make sense at scale when there is more granularity in the targeting and Twitter simply doesn’t have enough relevant real-time inventory of promotions to make this a user-friendly feature right now.

 

Influence: Diminishing Returns For New Influencers

In an influence network, users want to be heard. In the case of Twitter, the tweeters want to be followed and the perceived value that a tweeter derives from the system increases with the number of followers. Consequently, there is a positive network effect going on as users join Twitter because other users are already active on it, and there is a potential to develop a following.

The problem, though, is that Twitter possibly suffers from reverse network effects as an influence network beyond a certain point.

1. As mentioned above, following too many users can lead to a drop in the signal-to-noise ratio.
2. Since relevant content cannot be surfaced, the user resorts to pruning the list of people she follows to ensure that only the best tweeters remain. Twitter is uniquely suited to this because unfollowing requires very low investment.
3. Over time, as users do this, their propensity to follow new users decreases and their “following” list tends to stabilize at a certain number where earlier it had been growing linearly.
4. As existing users stop following others, new users joining the system get relatively lower returns. Effectively, the speed at which a new user can grow to a certain number of followers decreases as the overall network scales beyond a point.
5. The reduced attention and the lower number of followers that new users successively get decreases the value of the platform for the users.

The other problem with an influence network is a large number of dummy profiles that get created to increase the “perceived” influence of existing profiles. Twitter has had its fair share of spam with marketplaces like Twiends emerging for ‘trading’ influence. Increased noise again circles back to the loop described above.

So will users start quitting at some point?

Negative feedback loops of this kind do not kick off a death spiral instantly. This is because users join the network in cohorts. Users joined Twitter over time, and the reduced value of a noisy stream first becomes apparent to users that joined earlier. Hence, early adopters start moving on to start with. Late adopters stop engaging when they can’t reap enough from the system. Over time, the number of users entering the above cycles increases and value to users progressively decreases. Also, just like acceleration hits a tipping point, deceleration does too.

And finally, the new walls may make this more difficult

Third party apps have been at the core of what Twitter does. Users and developers have enjoyed some form of cross-side positive network effects for two reasons:

1. Relevance in a real-time feed makes it more user-friendly: Twitter’s focus on recency over relevancy gave us the real-time web. However, this focus also threatens to undermine its utility as a publishing and discovery platform. Till date, third party developers have effectively helped Twitter solve this problem by building relevancy focused apps on top of the stream. However, with Twitter closing the gates to its garden of late, there is a risk of Twitter not being as productive as it used to be (remember IFTTT triggers from Twitter? they’re not allowed anymore either).
2. Nothing beats cross-platform influence: Twitter’s growth as an influence network is, in part, attributable to the fact that it started off as a cross-platform open network, something that is very important for users on an influence network. However, its recent move to a more closed platform could impact the utility for influencers on Twitter.

With Twitter tightening the developer side, the utility of the service for users also gets impacted. In many ways, Twitter seems to be setting itself up for reverse network effects.