Tech blogs in Kenya have some shortcomings, in short, they suck!
It's something I've said before on other platforms and fora. I knew it empirically but didn't have the hard data to back me up.
I do now.
Let's get into why I'm the right person to be writing about this:
- I've been reading tech news on an almost daily basis for more than eight years now.
- I've been following the local tech scene in Kenya almost as long.
- I have a degree in Computer Science.
- I'm a part of the ecosystem.
- I have the skills to do this analysis.
- I've been thinking about this for a while now, almost a year.
While it's entirely possible to analyse more than one blog I decided to do only one as a sample. The blog I choose for this is Kachwanya because:
- It's been around for a while (since at least 2008)
- It appears on various lists around the web
- The founder of the site, Kachwanya, is the head of the Bloggers Association of Kenya
- Kachwanya’s operations are based out of the Nailab
- It's as good an example as any (other Kenyan tech news blog)
- It apparently makes money
I worked with the assumption that the titles of the blog's posts would give me insight into the overall content of the site. So I used a service called Kimono to extract all the blog titles for the past two years, to about 2 months ago.
This came to approximately 2,300 entries. I also extracted other meta-data, such as the author of the posts and the tags used on them.
Next up, I did a simple frequency analysis to see which were the most frequently occurring words in the dataset. I had to remove stop words (commonly occurring words such as 'a', 'the', 'it') that would add no value to the analysis. I also made a word-cloud that visually represents these frequencies.
Finally, I wanted to look at the words associated with certain keywords that appear most frequently and visually represent what those words look like.
First, I did something called 'Association Mining'. This measures the correlation (how often certain words appear along side others) between two entities, in this case words. If we take two words like, for example, "Sidney" and "prince" and the association is something like 0.3, it means that every time the word "Sidney" appears, there's a 30% chance that the word "prince" is near it.
Second, using association mining and Network Analysis, not only can we determine which terms appear together frequently, we can visualize how keywords and phrases are connected as a network of terms. This way, we can resolve the number of connections keywords have with one another, and how many connections a specific keyword has with other keywords.
Using network terminology, our keywords are the nodes/dots in the network, which are called vertices. Connections, which denote relationships, are named edges. In general, nodes which are related are shown close together, whereas unrelated nodes are shown further apart. The colour of the node is determined by which cluster the node belongs to.
Nokia Word Relations
It's not surprising that Kenya by far dominates the frequently occurring words, probably something to do with search engine optimisation (SEO), but the next 15 most frequently words do offer more of a story.
In there, we have some interesting words:
These are the brands or words that aren't generic that appear in the top 15. Anyone who follows this blog and really most other Kenyan ones would not be shocked at this composition. What did surprise me was the fact Nokia still featured so highly given that there haven’t been many posts about them on Kachwanya of late. It's probably a combination of lack of interesting news from them and the fact they stopped doing heavy promotion The Interpretationin the country (free gadget reviews).
Samsung Word Relations
Looking at the network graph for Samsung, it's all about their phones. They talk about launches, pricing and units sales in millions. Something that I found deeply amusing is that iPhone features (to the left in red).
Kachwanya has gotten the reputation of comparing the high-end Samsung smartphones to iPhones, sometimes without even testing the latter device or the phone being released.
The graph for Airtel is also about their services, what I want to point out is the green cluster which groups together Internet-related terms. Right in the centre, there is the word unlimited. This is a recently launched service, which at the time I scraped this data was barely five months old. That it features here gives me more confidence my suspicions there were a number of sponsored posts on the topic.
I put the Safaricom graph together for comparison.
Samsung Word Relations
We now come to the meat of the post. What I was looking for when I scraped the data; how reporting on Kenyan startups is done. Below, you have the network graph of the posts tagged “Startup”, or mentioning it and its variants.
There were only 80 posts on startups, out of the 2,300 articles found on the blog.For this analysis, I not only relied on the techniques mentioned above, but also went through each and every title, making this analysis both quantitative and qualitative.
Startup Word Relations
This is what the data revealed:
- Most of the posts were about some competition or other
- M-Lab is frequently mentioned
- There is only one mention of a Kenyan startup exiting, Weza Tele
- One mention of a startup raising money, Brck
- There's mention of some incubators such as 88mph, but little follow up
- There's mention of just more than five individual startups other than the two I've already pointed out
If you were to use Kachwanya as a source of information for the Kenyan tech scene, you would likely come to the conclusion that a lot of people have about the ecosystem: there's a lot of fluff.
You would come out wondering if there's any money being made/raised in our (Kenya) ecosystem and that we're totally supported by various competitions and hackathons.
What we see with Kachwanya is a lot of lazy writing and no research.
It's easy to take a press release and post it as news. It's even okay to do so. But even with all the posts about startup competitions and incubators, I found no follow-up on the start-ups or even individual stories written about them. This is particularly egregious with Kachwanya because he's based out of the Nailab and every six months a new group of startups come in.
Nailab is currently accepting applications for sixth season of incubation, that means about 30 startups so far have gone through incubation yet from this blog, which I’d remind you is run from there, you'd not know which have folded, raised funding or even what they're current status is.
I've heard several excuses, but doing some sort of follow-up is not hard. From my position, admittedly privileged, working within the tech ecosystem, I've heard of a couple of acquisitions, valuations, funding rounds, user numbers and engagement, large partnerships and international expansion plans. For a lot of this all I had to do was ask, follow the right feeds and people, and pay attention at various events held around Nairobi. In fact, one startup not only had billboards around town announcing their huge new partnership they even had ads in the paper, I'm yet to see anyone write on it.
I did all this not to take a dig at Kachwanya but to point out something that needs to change and hopefully get a conversation going around it. It's fine to write posts that bring in views, I'm even fine with posts that are sponsored (it would be polite to point it out though), but I would like to have more about the actual ecosystem. You can't have in your title "Kenya Tech News" when there's no actually tech news about Kenya.
I should point out this is not just the problem with Kachwanya alone, but almost all local tech blogs. I'd go as far as saying there's no real exclusively Kenyan tech blog, but if you know one please, please mention it in the comments. I'd love to be proven wrong.
Meanwhile until this changes I will continue, as I told Kachwanya once on Twitter, to use his blog as an example of the failure of the local tech blogs.
All the data and visualisations in this post can be found here if you'd like to explore it and the code can be found on my GitHub repo here. Question, comments and responses are welcome in the comments.