The end of MTN’s free Twitter promotion on September 25 could possibly have brought about a huge decrease in the volume of tweets because of the users of these service, nevertheless it doesn’t appear stood a major effect within the volume of activity over the social networking in Nigeria.
On September 22, MTN announced that, after a lot more than 4 years, it was actually ending its promotion that offered South African subscribers having access to Twitter with no width=”439″ height=”292″ />
But it is possible to actual impact? And also the users who were using free Twitter access definitely will leave prestashop simply because still cannot afford the data required? Are these claims short-sighted by MTN and sure to trigger churn looking at the network? What element of the South African Twitter community were using free Twitter? None of these questions are super easy to answer, because much of your data is actually difficult to read about.
Analysis of easy research discovered that the “freeTwitter” users’ activity declined by at the least 30%, and that around 18% to possess stopped posting on the network altogether. Markets ., the impact for the entire South African Twitter community have been minimal.
In fact, volumes of tweets in Nigeria actually increased following the end on the promotion period.
Called into question
Some on the response to your analysis also call into question MTN’s claims in connection with amount of Twitter users on its network. A back-of-an-envelope (and really imprecise) calculation, according to the amount of users that stopped activity altogether, hands the proportion of South African Twitter users on MTN’s network at approximately 25% – consistent with its mobile network share of the market in Africa. As outlined by MTN’s estimates, that will suggest that the Twitter community in South Africa is 52-million strong (18% on the entire global Twitter users list), which is extremely unlikely.
Let’s commence with the important numbers: the quantity of Twitter users are there in Africa in total? Nobody knows exactly, except maybe Twitter – and this doesn’t release that information. Global Worx’s SA Social websites Landscape 2018?report (released in September 2017) estimates the South African Twitter community at eight million users. This could be slightly conservative resulting from sampling limitations but seems reasonable because of the connected population in Nigeria (21 million, good research firm) additionally, the global user base of Twitter (approximately 300 million).
However, MTN claimed there were 13 million users of free Twitter on his or her network. That is the big discrepancy – MTN is arguing the fact that Twitter users on its network alone (MTN has around 30% business of mobile connections in South Africa) outnumber All over the world Worx’s whole country estimate by 60%. Thirteen million also looks suspect because MTN simply doesn’t have visibility on its network of actual Twitter user details. What MTN could see is data traffic, therefore it may identify that traffic determined by Ip address addresses, by Sim numbers and sure by device IMEI (International Mobile Equipment Identity) numbers. None of those will provide the actual Twitter user account information, so all MTN can measure is the info travelling collected from one of of that network IPs on the Twitter IP. A guess is always that the origin of the number that MTN has publicised is from 13 million separate devices who have accessed the Twitter IPs across the period that your promotion was active (4 years and nearly five months). It is quite difficult to link that into a specific range of Twitter users, however.
So, can you really study the impact with the end with the promotion with virtually no definitive data about the variety of users on MTN as well as size of the South African Twitter community?
We can’t state definitively what are the overall impact is, but it is a possibility to be diligent with a sample of Twitter users which are highly likely to end up consumers that use service, and measure their aggregated activity within the social media.
That should give some indication your house activity has reduced significantly. Likewise, we can source a random sample of South African Twitter users and then determine whether we have witnessed similar variations in activity.
To receive a sample of “freeTwitter” users, they’re certified using Twitter’s standard search API pulled most of the tweets that mentioned the hashtags #RIPfreeTwitter, #freeTwitter and (lower incidents) #MTNmustfall. This search query harvested tweets between 14 September and 26 September 2018. It resulted in a specimen of roughly 38 000 separate Twitter users which had used these hashtags. Unfortunately, considering that the #RIPfreeTwitter and #freeTwitter hashtags trended after MTN’s announcement within the promotion’s end, there was many spam accounts that utilised the hashtags in order to gain some exposure. These “hashtag-jacking” frequently occurs with trending topics on Twitter.
To reduce the impact of such spam accounts, these were filtered by using a benchmark proposed by Ben Nimmo on the Atlantic Council’s Digital Forensic Research Lab. He proposes than a Twitter account that posts 72 times every day if not more over a longer period might be automated or “spammy”. Seventy-two posts every day is the same as a tweet every 10 minutes for 12 hours – an unlikely frequency over the sustained period.
After a preliminary filter for spam, the overall Twitter timelines from the remaining is in charge of the time scale 1 to 27 September were harvested. The MTN promotion ended at nighttime on 25 September, so the sampled tweets includes 2 days right after the end within the promotion.
Once the timelines were pulled, another spam filter was set you back get rid of the accounts that posted with an unreasonably high frequency in the specified period, and also a final sample around 6.7 million individual tweets ended up being analysed.
To receive a control sample to be sure of that any significant modifications to activity in the “freeTwitter” sample are relevant (but not resulting from general posting patterns) a sample was sourced by conducting a seek out tweets geolocated in Africa on September 25. From that result, 700 random accounts were selected as well as their timelines for your studied period were pulled. After filtering the “control” sample for spam, 599 accounts remained during the sample. If Twitter’s geolocation is 90% correct, this sample would generate a 2.4% confidence interval on a 95% confidence level. In other words, the sample really should be connected the complete South African Twitter community with regards to activity.
What is instantly apparent on the results are that you’ve got a considerable drop-off within the activity in the “freeTwitter” sample following the take off on September 25. The mean daily aggregated tweets of the sample fell by 30%. It’s not visible within the control sample where this measure actually increased by 80%.
To experience a more granular have a look at the tweet activity, you can visualise both samples by decile (see charts below). Within both samples, the top 10% from the accounts are accountable for a substantially higher tweet frequency as opposed to other deciles – this is certainly likely due to some spam accounts which were operating underneath the threshold of 72 tweets/day.
Daily tweet volume by decile, September 2018: control community
n=599. Total volume of tweets: 292K
Daily tweet volume by decile, September 2018: freeTwitter community
n=31330. Total level of tweets: 6.7m
Another way of measuring the impact with the end on this promotion will be to look at the net quantity of accounts that are not anymore from the period after September 25. We could compare the proportion of the accounts in the “freeTwitter” and control samples. In the free Twitter sample, 18.7% of accounts that had been involved in the month prior to the cut-off were silent after September 25. On the other hand, the control group had only 4.4% of accounts that had been not active after the cut-off.
It is obvious that this end of free Twitter has already established an impact with a particular community in South Africa: the sample of users that aligned while using the free Twitter concept during the content in their tweets has a marked decline in activity, generally speaking by around 30%, but higher within the lower deciles (between 60 and 65% for deciles 1