Who's the most influential on SA tech Twitter?
It started with a tweet:
Philip Joubert is the co-founder of Offerzen, a tech recruitment platform very popular in South Africa and more recently the Netherlands. He posted the question on Twitter, and got enough replies to be able to populate a list of 70 interesting people in the South African tech ecosystem.
VoxCroft builds OSINT (open-source intelligence) products for customers in the public and private industry, for example, the Arrow platform that combines insights from human analysts and machine learning to provide decision-makers with insights that are relevant to them. On our team, we have some of the foremost OSINT analysts, who eat, sleep, and breathe analytics; and they were up for the challenge.
When we start gathering insights on a new territory or topic, the first step is to determine where to get great data; in this case Twitter. A customer may know one or two influential people in a network and we can use this to find related users and score their prominence within the specific network. I asked our analysts to quickly use Philip's list as a seed and find other relevant accounts.
Here's the Top 10 SA Tech Twitter accounts according to our methodology:
How it Works
The analyst takes the Twitter list as a baseline or seed list. They extract the last 500 tweets from each of them and map the conversations which were happening in them, including mentions, replies, and retweets. This gives us an indication of who these individuals are interacting with, with which we build an expanded list.
We then use the PageRank algorithm to score this expanded list. PageRank is typically used to score the relative importance of websites by tracking quality links to and from a page – in our case, we do the same but with Twitter accounts as pages, and mentions, retweets, and replies as the links. This provides a fairly good estimate of interest.
A core truth of VoxCroft is that automation technologies can only get you so far, and you still need humans to do things like making judgments and understanding context. So our analyst went through the top 200 accounts on the list by hand and eliminated non-South African entities and other irrelevant entities. For example, @YouTube and @elonmusk were quite popular, but would not be relevant to this list since they are not active in any local conversations. We did include some companies there as well (including Offerzen!).
The end result is a list of those who are engaged most broadly and actively in tech Twitter in South Africa.
You can access the full list of 133 accounts here: link
You'll see there is a clear bias in the data, which is caused by the demographics of the local industry, but more relevant here the bias in the initial sampling. In networks like these, you'll expect bubbles of conversation, and part of what we'd typically do next is to try to identify the edges of these bubbles, and figure out how to expand more broadly.
Shoutout to the analyst who helped me with this!
If you want to read more about the problems that we are solving with OSINT, you can read our blog on how it can impact humanitarian action here.