Portland, a strategic communications consultancy firm recently released a report on a study conducted in the last three months titled “How Africa Tweets“. The summary of the report is as follows:
- Johannesburg is the most active city in Africa, with 344,215 geo-located tweets during Q4 2013, followed by Ekurhuleni (264,172) and Cairo (227,509). Durban (163,019) and Alexandria (159,534) make up the remainder of the top five most active cities.
- Nairobi is the most active city in East Africa and the sixth most active on the continent, with 123,078 geo-located tweets in Q4 2013
- Accra is the most active city in West Africa and the eighth most active on the continent, with 78,575 geo-located tweets in Q4 2013
The research by Portland cannot be said to provide any relevant information on how Africa tweets. This is because Portland decided to do a research based geotagged tweets, which are an insignificant fraction of tweets overall. The above information is not important unless one wants to use them to report on the characteristics of geotagged tweets by city in Africa. But such a report won’t be complete as it will lack information from non-geotagged tweets over the same period of time in the same cities for comparison purposes.
According to anecdotal report found here, less than 1% Twitter users geotag majority of the tweets leading to only 1% of tweets being geotagged in some way. A research by Kalev Leetaru et al titled Mapping the global Twitter heartbeat: The geography of Twitter reported that “2.02 percent of all tweets included geographic metadata, with 1.8 percent having a Place indicator, 1.6 percent having Exact Location, and 1.4 percent having both”.
The fact that only geotagged tweets was used in the study has elicited debate on Twitter involving @Kachwanya, @NairobiWP and @wiselar among many others. Here is how the conversation has developed:
People turn on twitter location while travelling & not when in Kenya. Part of why you shouldn’t quote that Africa Tweeting cities research
— Kachwanya.com (@kachwanya) March 13, 2014
@kachwanya Exactly. Flawed study.
— Latiff Cherono (@NairobiWP) March 13, 2014
@nairobiwp@kachwanya actually statistically speaking the people who geo-locate their tweets could very well be a good sample size.
— /wslr (@wiselar) March 13, 2014
The above tweeps hold to the view that the research is flawed, which I agree with, but @wiselar is of the opinion that the size of geotagged tweets provide enough sample size for studying tweeter trends in the said Africa cities. When looked at sample size alone, one would say the geotag tweets are sufficient, but one needs to analyze the geotagged tweets a bit deeper.
When taken at surface value, a marketer would think that Jo’burg is the most tweeting city given the above statistics but those in Lagos hardly tweet as Lagos, despite being the most populous city in Africa, is not in the list of top five tweeting cities. So what are the unique characteristics of geotagged tweets that will make them unsuitable for analyzing tweeter trends in African cities or elsewhere?
First, by default, the geotag feature is turned off. However, one can choose to turn on it either from the setting or ‘Add Location’ for each individual tweet when composing. It has been found that most of those (especially Kenyans) who travel to foreign lands are the ones who find it necessary to turn on the location feature. With this in mind, it is likely that the tweets analyzed across the cities had significant contribution by tweeps from other cities; e.g. Naoribians in Jo’burg could have contributed to Jo’burg being considered the top tweeting city in Africa.
Secondly, different populations could be having different levels of tech savviness. For one to change some settings whether on TV, radio, phones or even social media and email accounts, a certain level of tech appreciation is required. The cities with high number of tech savvy guys is expected to have a higher number of tweeps changing their Twitter setting to allow for geotagged tweets as opposed to a city with a population that is not as tech savvy.
Lastly, geotagged tweets raises privacy issues and there are a number of articles online advising people against turning on the geotag. A population with high level of privacy awareness and at the same time are worried about their security would not tag their tweets with geographic meta data.
These reasons – traveling vs turning on geotag, tech savviness, and privacy concerns makes geotag tweets be biased. Generalizing tweets based on information gathered from geotagged tweets will lead to a flawed report – and this makes the report by Portland a flawed report hence irrelevant. In conclusion, geotagged tweets cannot be a useful sample in analyzing the trend of tweets in African cities as over 98% of tweets with totally different characteristics were left out from the study. I won’t advise anyone to use the results of the study for any marketing or otherwise campaign as reality is far from the figures in that flawed report.
IMHO, the actual reason why using geotagged tweets is a bit misguided is because most people are not on high end Twitter clients that enable geolocation, not because turning on is such a hassle really. The privacy one is definitely legit.
I agree with you on the high end Twitter clients argument but in my experience the more sophisticated someone is, the more likely he/she can change default settings of any system.
The absence of the most populous City in Africa alone is reason enough for the study to be suspect.
The information in the research is highly relevant, but you need to acknowledge the limitations. That is the way it works in research.
Unless Twitter opens up its data for full analysis, this is the best data that is available. Let us use the information to dig deeper.