Research,  Social Computing

Challenging big social media, falling to partisanship: The Koo Conundrum

We examined the use of social media by politicians on Koo to understand the spread of political preference and engagement for the social media channel that has sought to challenge Twitter online. We find that the broad trend is not partisan, but almost entirely dominated by a single political party, the BJP. Moreover, we find that there is a clear pattern of repeat use, where politicians are not switching away from other platforms, but rather using Koo in addition to, and more importantly, as a secondary option to their existing accounts on Twitter. Both of these have consequences for the longer-term adoption of Koo, and the content experience of new users based on their political preferences.

We consider some of these findings in greater detail.

Party spread on Koo

First, we annotated a set of 1,408 known politicians from the General Elections 2019 dataset to seek out their Koo handles. We use Twitter as the key comparison point since a number of studies have shown that Indian politicians, particularly from mainstream political parties, tend to use Twitter as a major, and often primary means of outreach. After doing a match of politicians, we find that an overwhelming majority of politicians on Koo are from the BJP. While the BJP in general also dominates the political landscape on Twitter, its dominance on Koo is much more pronounced. In figure 1, we show the exact distribution of politicians from the General Elections dataset. BJP politicians constitute 79.1% of the politicians found on Koo in our dataset, yet they account for 93.41% of the posts across all politicians. Clearly, in both the presence and frequency of messages, the BJP significantly outnumbers other parties. 

Figure 1: Party-wise count of politicians on Koo and Twitter in the annotated dataset. While BJP and INC politicians are dominant, AAP is the only other party with a noticeable presence on Koo.
Figure 2: Party-wise average posts per politician on Koo. Politicians from the BJP tend to be more active than other parties. The error bars indicate the standard error (standard deviation divided by the number of observations).

Content Overlap

The second pattern we see is that of content overlap. We collected timelines of politicians who are on both Koo and Twitter and mapped the similarity of their content on the two platforms using Jaro-Wrinkler similarity (threshold of > 0.9 match). We also restricted the matching timeline to fall within 1 week, in order to eliminate spurious matches. The results confirm that 78% of politician’s posts on Koo are a near-perfect match of their posts on TwitterThat the two platforms mirror each others’ content may not be particularly interesting, since politicians typically have the same sets of messages across various platforms, which indeed allows them to be consistent as well. What is interesting though is that the origin space of most content is Twitter, not Koo. 

A vast majority of politicians post on Koo closely after they post on Twitter. Calculating the average time difference, we find that politicians post at an average difference of 42.22 minutes with a standard deviation of 95.16 minutes, thus confirming that a small number of politicians do in fact use Koo first.

Some politicians, including major leaders like Piyush Goyal, KP Maurya, H Raja, Giriraj Singh, and Nitin Gadkari, all tend to have almost exactly the same message on both platforms within an hour. Of the major leaders, only two use Koo before Twitter in 10 percent or more instances Piyush Goyal (16.9% instances) and Giriraj Singh (11.6% instances), suggesting that even the major supporters of the platform still overwhelmingly prefer Twitter as their primary output mechanism.

In Figure 3, we also see that the higher up the bubble, the more likely the politician does an exact or near-exact repeat on Koo of what they do on Twitter. This is important, since it implies that there is very little additional value on Koo for viewers who are already on Twitter, especially when the  politician they follow/care to listen to is a major figure in the party.

Figure 3: Bubble plot consisting of percentage of matching koos (y-axis) and median time differences between matched posts on Koo and Twitter (x-axis) on a log scale. Bubble sizes are scaled to the number of total posts on Koo and colour-coded by party affiliation. We find a large number of politicians post nearly identical koos and tweets, with a short time difference between these posts. We find that 72.5% of the politicians in our dataset have at least 75% of their posts on Koo almost perfectly matching with their tweets. Moreover, 78% of all posts on Koo made by all politicians from our dataset are nearly identical to their tweets.

In a nutshell, our findings here suggest that Koo has significant challenges in catching up with Twitter, not so much because of the growing pains that platforms face when up against established players, but because the differentiation is much weaker. If we think about recent cases of social media platforms upending or eating into existing well established players – such as TikTok or Clubhouse, they were able to provide something fundamentally different on one or another of key facets such as the production or output of content, the algorithms that made content available, the people producing the content, or the format of the content itself. Koo’s two major distinctions were language, and the size of the content itself, neither of which are particularly important differentiators – since Twitter already provides significant language support, and the increase in the character length does not appear to be an important factor, at least for the political accounts. We find that in the sample of politicians who are active on both platforms, non-English languages are used more than English. On Koo, non-English text is used by politicians 73.5% of the time, while on Twitter, the corresponding number is 58.2%. This suggests that although Koo is in fact more used in Indian languages, the number is high enough on Twitter that it does not make for a replacement factor.

Similarly, if Koo had been able to tap into a unique population of users who were not actively online or listened to, they may have had a better shot at creating that differentiator. This is a factor we do not study here, and it may be that Koo is creating its own primary and independent user base that is not active elsewhere. For instance, bundling into a mobile platform could enable this, and give the platform a lead with populations who are entirely new to social media use. 

Figure 4: Post length on Twitter vs Koo: we see that the higher character limit of 400 is used sparingly, most (politicians’) tweets are within the Twitter threshold of 280 characters, even on Koo.

What we see here is a duplication, rather than a move of existing players from one platform to another. This is nonetheless important, because it suggests that a sizable number of key politicians are willing to invest their time into Koo, and that they see the potential for an audience there. However, this is undercut by the fact that the majority of politicians are from a single party, and one that is known to have a relatively centralized and well-organized social media strategy. One means of expanding a new social media channel is to pick up those who are de-platforming from another channel, such as disgruntled users. But for this to happen, either the first platform needs to become dramatically unwelcoming for a large segment of users (eg. 8chan/8kun) or needs to find itself legally attacked in a significant way that makes the competitor unviable. 

Figure 5: Share of similar content on Twitter and Koo by politicians.

For now, without these in place Koo will find difficulties being accepted, even more so because it risks finding itself becoming slotted as an ideologically driven social media, especially given one party’s predominance on political conversations, which can be a very significant deterrent to new users. Indeed, some platforms like Quora have survived despite being slotted into an ideological corner, but they offer something fundamentally different that other platforms do not. For now, our research suggests that Koo is not on the path to seriously challenging its alternatives.

This research was conducted by Asmit Kumar Singh1†, Jivitesh Jain2†, Shradha Sehgal2†, Professor Joyojeet Pal3, and Professor Ponnurangam Kumaraguru2. This piece can also be accessed on Professor Joyojeet Pal’s blog.

1Indraprastha Institute of Information Technology, Delhi
2International Institute of Information Technology, Hyderabad
3Microsoft Research, India
Authors contributed equally to these findings.

Comments Off on Challenging big social media, falling to partisanship: The Koo Conundrum