In the previous article we looked at follower gain and viewership during the course of 24 hour streams. In this article we are going to compare follower gain and viewership during the stream with streams that occurred before and after. The aim being to determine if the gains during the stream are higher or lower than before, and if the gains remain at this level post the stream or if there is a negative effect to doing the stream.
Method is the same as stated in chapter one, with the exception being that we are including data for the 48 hours of streamed time before and after each 24 hour stream.
The following bar chart shows follower gains per hour before, during and after the stream. Please note, the figure for during is the first 24 hours only, so if a stream is say 36 hours in length only the first 24 hours will be included. The before and after figures are the 48 hours of streamed time before and after the 24 hour stream.
It's clear that on average there is a increased rate of followers during the course of a 24 hour stream, this rate however does not persist after the stream. If I was to speculate as to the reason for this, for channels of moderate size the increased visibility effect we discussed in part 1 could play a part. I would expect however that the event like nature of these streams coupled with increased advertising will have a greater effect, again this is purely speculation. Comparing upcoming stream mentions and comments on Twitter/forums etc. would be very interesting to see the effect they have both on 24 hour and shorter streams, this is well beyond the scope of this article however.
Viewership sees an increase during the course of the stream, but this increase is not as large as the increase in follower gain (roughly 25% vs 50%). We do however see a slight gain in viewership post the stream, compared with a slight decrease in follower gain.
As discussed before, global comparisons such at those above are not sufficient as the range of channels on Twitch is so great. The following chart compares the rate of follower gain during the stream with the rate before the stream, broken down by channel size. The before data is taken from the 48 hours of stream time before the 24 hour stream. A stream is "Same as before" if the follower gain during the stream is within 15% of the value seen before the stream, more than before is a 15% to 50% increase and much more is greater than 50%.
The majority of small channels do not see any change in follower gain, this is due to them not gaining any followers before or during the stream. Once a channel researches the 100 follower mark there is generally a 30% chance they will have a lower rate of follower gain, 20% chance they will see no change or a small change and a 45% chance of a large increase with about 5% variance between the groups.
There isn’t a large variation between the channel group sizes (apart form the smallest channels) when it comes to the distribution of those that lose/gain followers. Obviously there will be a large difference in the actual rates (larger channels will have a larger follower gain rate) but in percentage terms there is not a wide variation. Group 17 shows unusual losses when compared with the others, but I suspect this is down to sample size.
The above chart compares the rates of follower gain before and after the 24 hour stream. Again there is not a radical difference between the channel size groups (apart from the smallest channels). Most groups see at least 45% of channels having a lower rate of follower gain after doing a 24 hour stream when compared to before (the 48 hours of time streamed before and after the 24 hour stream). Around 20% of channels see no difference or a small increase and around 30% of channels see a significant increase in follower gain.
This data tends to suggest that a portion of channels do very well out of a 24 hour stream, others on the other hand may end up doing worse (at least in terms of follower gain) than they would have if they had streamed normally.
How does viewership compare when broken down into channel sizes? As before, the raw viewer numbers will be significantly different group to group, but in this comparison we are comparing each channel with itself, grouping the channel by size and then averaging the results.
Again the smallest channels see little difference because the majority of them will have little or no viewership before and during the course of the stream. Around half of medium and large channels will see an increase in viewership during the course of the stream, while around 20% will have less viewers. It’s interesting to note that channels in the 80-5120 follower range appear to have a slightly elevated chance of seeing much higher viewership during the course of the stream. This could be further evidence of increase visibility for channels of this size during the course of the stream discussed in the first article.
So how does viewership compare after the stream?
The majority of the smallest channels again see little difference, for moderate and large channels, roughly a third of channels have decreased viewshiper, a third no large difference and a third see increased viewship.
Doing a 24 stream does not appear to have a massive impact on retained viewership numbers for most channels, with a relatively even spread of losses/gains/no change. I would admit the number of channels seeing a decreased viewership is more than I was expecting and it would be interesting to examine the magnitude of this loss, whether or not it persists long term and how it compares with shorter streams.
Interestingly the 80-5120 follower range channels have a comparatively high viewership retention probability but also a higher probability of loss.