Archive for the ‘Digital Economy’ Category

Analysing the Perceptions of Digital Gig Workers: Mining Emotions from Job Reviews

In a previous post, we provided a discussion of how the analysis of user-generated content (e.g. comments/posts on social media and/or job review sites) can help in understanding perceptions of digital gig workers. The prevailing assumption is that generally, digital gig workers contend with non-standard working conditions, e.g. the lack of social security coverage, long working hours, lower salaries, and the lack of benefits. Nevertheless, it is believed that digital gig workers in the Global South in particular perceive their jobs as being better than local benchmarks (i.e. office-based work).

To test the above assumptions, we developed and employed automatic text analysis methods to answer the following research questions:

  • How do digital gig workers feel about their jobs?
  • Which topics pertaining to decent work standards do they frequently talk about?
  • Are there any differences—in terms of sentiments and topics—across different geographic locations, or across genders?

We hereby present the results of analytics in the way of answering the questions above.

Firstly, we collected online posts published by digital gig workers from Glassdoor, a web-based platform for sharing reviews of companies and their management. Focussing on reviews of the digital gig companies Upwork, Fiverr and Freelancer, we retrieved a total of 567 reviews, 297 of which include geographic metadata (i.e. the geographic location associated with the account/profile of the user posting a review). For our text analysis, we made use of the Pro and Con fields that each review came with.

Based on the NRC Emotion Lexicon, a dictionary-based emotion detection method (implemented in the R statistical package) was applied on the reviews, classifying them according to Robert Plutchik’s eight basic emotions: Joy, Trust, Fear, Surprise, Sadness, Anticipation, Anger, and Disgust. We then grouped the reviews as either coming from the Global North or the Global South based on the geographic metadata attached to them. Shown in the figure below are the 15 most frequent emotion-bearing words found within reviews, represented according to the emotions they express. Bars in amber correspond to words prevalent in reviews from the Global North (GN) while those in blue pertain to those in reviews from the Global South (GS). 

Riza GNGSemotion

It can be observed that there are more words within GS reviews containing emotions that are clearly positive. All of the 15 words associated with Trust were found more often in GS reviews. Furthermore, 10 and 8 words associated with Joy and Anticipation, respectively, were more frequent in GS reviews. These results support the belief that digital gig workers in the Global South (GS) do express positive feelings towards their jobs.

Meanwhile, our results show that digital gig workers from both GN and GS express negative emotions. On the one hand, GS reviews were the source of 11 and 10 words associated with Anger and Fear, respectively. On the other hand, 15 and 11 words associated with Sadness and Disgust, respectively, were contained in GN reviews. This suggests that generally speaking, digital gig workers do have to contend with less than ideal working conditions, which in turn trigger such negative emotions.

Finally, 10 words associated with Surprise came from GN, 5 from GS. It is worth noting though that this particular emotion can either be negative or positive depending on context.

These results are but “teasers” to the full results of our automated analysis. Further details including the topics/themes towards which such emotions are targeted, as well as answers to the second and third research questions stated above, will be presented by Dr Victoria Ikoro in the upcoming 3rd Annual ICT4D North Workshop to be held in the Management School of the University of Liverpool on the 6th June 2019.



What can we learn about e-commerce in Africa from Jumia’s IPO filing?

30 April 2019 1 comment


There has been growing discussion about the potential of e-commerce in developing countries. This discussion intensified recently when pan-African e-commerce firm Jumia went public in the US, becoming the “first African unicorn”.

The IPO prospectus, a 270-page outline of the firm released as part of this filing, has sparked much debate. Elsewhere, TechCrunch has dug into the financial intricacies of Jumia, and online debates have raged linked to the question “Is Jumia really an African firm?” (see here and here).

I won’t detail those two issues here. Instead, below I will discuss the insights that the prospectus provides us about e-commerce platforms operating in Africa. This is especially useful as we have been struggling with a lack of detail on e-commerce, with firms reluctant to share commercially sensitive information.




1) The market for e-commerce in Africa is already large, and growing rapidly
Jumia announces that it had 4 million active customers in the year 2018, growing rapidly from 2.7 million in 2017. This translates into “Gross Merchandise Volumes”  (GMV) of €828.3m in 2018, from €507m in 2017. As these figures suggest, e-commerce is bringing in a significant number of customers on the continent and is growing at a rate of over 50% in the last year.

Regionally, the largest countries for Jumia’s business are Nigeria (29% of GMV) and Egypt (20.5%). Interestingly, in the largest 5 markets, only 50% of goods are delivered in primary cities with the rest evenly split between secondary cities and rural areas. Most customers are mobile users, which comprise 81% of all traffic in 2018.


2) …but making profits is a challenge. Jumia is a loss-making firm
The growing size of e-commerce has not yet translated into profits. Jumia made a €170m loss in 2018.

This is attributed to a number of factors. As a platform, sales are mainly made by third-party sellers (90%). So the €828m GMV translates to just €130.6m revenue for Jumia.  Jumia then faces high costs for operating including warehousing, delivery, sales and advertising and the platform. Once these are considered, the €130.6m turns into a €170m loss.

How does Jumia plan to become profitable in the future? By becoming a market leader in Africa, Jumia will rely on expanding markets to increase sales volumes over the coming years. Alongside this, they describe the potential of reduced costs in fulfilment (storing and delivering goods) and growing use of mobile payments that should make each transaction more efficient over time.


3) The role and prospects for Jumia platform sellers
Jumia has 81,000 sellers and the prospectus hints that quite a number of these are African, although there is no data provided. Interestingly, commissions from sellers for sales only contributed moderately to revenues. Jumia makes similar amounts of revenue from services such as fulfilment (delivery costs to buyers) and “value-added services” (premium services for sellers).

As Jumia grows, its aim is to attract ever more sellers to the platform. “Competition between sellers is…essential to our monetization, as it increases the appetite for sellers to use our services”. This statement suggests that the platform will move low margins and high competition, with sellers relying on value-added services as a source of expansion. For local sellers, these future directions suggest concerns about this platform supporting widespread upgrading of local SMEs in Africa.


4) Data is an important component of the relationship with sellers
We know that data is vital to many platforms, and Jumia is no different. A seller score “..a data-driven scoring of a seller’s performance” is a key aspect of Jumia’s relationship with sellers. The score is important to the success of sellers, where platform algorithms use this score to determine the order of products in searches. Jumia also gamifies the “seller score”, associating it with a range of other advantages for sellers.

In order to support sellers, Jumia supports third-party financial services for its sellers, using this score as a way to demonstrate creditworthiness. This mirrors alternative credit-scoring schemes we have seen elsewhere in Africa.


5) African e-commerce platforms face significant and often unexpected risks
Many of the challenges for Jumia are similar to those challenges of e-commerce elsewhere [1]. One key challenge repeatedly mentioned is the “…failed deliveries, excessive returns, late collections, unrecoverable receivables and voucher abuse”, with 14% of all sales (by revenue) being failed deliveries or returns. These risks particularly emerge from the use of cash payments on delivery (COD), a common approach to payments for online goods in developing countries.

Beyond this challenge, an eye-opening aspect of the prospectus is the unexpected challenges encountered by Jumia:

  • In Kenya, a Jumia warehouse was robbed and €500K of merchandise was stolen
  • In Egypt, a €5000 fine was imposed by authorities due to the platform offering “unlisted drugs”
  • In Kenya, a scam involving electronic payments suppliers led to €550K losses
  • Jumia has undertaken investigations around allegations of internal fraud and bribes in countries such as Nigeria and Morocco connected to relationships with platform sellers


6) Policy challenges for e-commerce in Africa are significant
In line with other platforms, Jumia operates a relatively “asset light” way across multiple countries in Africa. Even with close to a billion euro in GMV, and with offices in 18 countries, it is a tiny direct employer with around 4,800 staff in Africa.

This “asset light” approach, however, comes into collision with African governments’ desire to regulate e-commerce. Jumia discusses the challenges faced in a number of countries including taxes, particularly VAT on imports; uneven data protection rules, and restriction of cross-border personal data transfer; and regulation on financial and mobile payments.

There has been pressure for harmonising such rules globally in order to support cross-border business models of digital firms such as Jumia. However, African governments are keen to ensure that they are able to operate their economies appropriately, including collecting taxes and nurturing local firms, in the face of e-commerce imports [2]. It is therefore understandable that policy might vary according to the political goals of different nations in Africa. These challenges are likely to intensify coming years given current disagreements on harmonising e-commerce at the WTO, and African CFTA discussions on e-commerce rules still at an early stage [3].

Platform firms, therefore, require careful mapping of national rules and regulations, and in Africa the “asset light” model may only be viable to well-funded platforms. For smaller platforms, it may be better to focus on a smaller subset of countries.


In summary, the Jumia prospectus indicates that digital economies are expanding and we can expect to see a growing set of firms operating across the continent. However, given the challenges encountered by Jumia, easy profit and growth are not a given. For platform sellers, e-commerce provides new potential market opportunities, but careful consideration of how platforms best support them, will be vital to success.


[1] UNCTAD (2015) Information Economy Report 2015 – Unlocking the Potential of E-Commerce for Developing Countries, UNCTAD, Geneva, Switzerland.
[2] Azmeh, S. & Foster, C. (2018) Bridging the Digital Divide and Supporting Increased Digital Trade: Scoping Study, Discussion Paper, GEG Africa, Pretoria, South Africa.
[3] Foster, C. & Azmeh, S. (2018) E-Commerce and the African Continental Free Trade Agreement (AfCFTA), Discussion Paper, GEG Africa, Pretoria, South Africa.


How Many Platform Workers Are There in the Global South?

29 January 2019 Leave a comment

In developing countries, there has been a rapid increase in the gig economy and in the presence of digital labour platforms: defined as “a set of digital resources – including services and content – that enable value-creating interactions between consumers and individual service-providing workers”[1].

But how many workers actually work for such platforms?

I am not going to provide a reliable answer to that question but I will give some kind of ballpark figure.

We start by dividing out two types of platform work: digital gig work that involves digitisable tasks like data entry, writing copy, web design, accounting, etc; and physical gig work that involves a physical task like taxi driving, food delivery, domestic work, etc.  A previous estimate[2], updated to account for growth, would be that there were something like 10 million active digital gig workers in the global South at the start of 2019 (and around ten times that number registered on digital labour platforms but with 90% of them inactive).

So how many physical gig workers are there?  I’m going to break this down by continent since the extent of physical gig work seems to vary significantly between the three main continents of the global South.


Calculations here are based on extrapolations from just two economies, and seek to take account of wealth and population[3].  Current research for the Fairwork project estimates around 30,000 physical gig workers in South Africa; about half in taxi-driving and the rest mainly in delivery and domestic work.  Estimates for Nigeria[4] plus re-use of some of the same ratios found in South Africa, suggest 20,000 such workers.  Accounting for GDP per capita and population suggests around 60 workers per US$1,000 GDP/capita and per 1 million population; i.e. per US$1bn GDP.  Multiplying up to the overall GDP of Africa produces an estimate of c.130,000 physical gig workers in Africa.  However, given there are at least 100,000 in Egypt alone, we can at least double that to 250,000.


Similar calculations can be undertaken in Asia, based on numbers associated with platforms in India and Indonesia.  Extrapolating from estimates for taxi-driving and food delivery platforms in India[5], I estimate around 2 million physical gig workers in India.  For Indonesia[6], the figure is closer to 1 million.  Accounting for GDP suggests around 800 workers per US$1bn of GDP.  Multiplying up to the overall GDP of Asia (excluding Japan) produces an estimate of roughly 18 million physical gig workers in developing Asia.

However, there is an alternative approach, which is to exclude China in this calculation, which produces a figure of 9 million, and then take at face value claims that Didi Chuxing employs 21 million physical gig workers in China[7].  This would lead to an estimate of 30 million physical gig workers in developing Asia.

Latin America

Here, I’ve taken a simpler approach based on some national and continent-wide estimates of taxi driving[8] and then re-using ratios from the South Africa work.  This produces an estimate of something like 2 million physical gig workers in Latin America.


The basis for these estimates is flimsy, and the extrapolations are worse, so please attach a strong health warning to this material.  Better still, come up with some improved statistics.  But my ballpark figure is that there are at least 30 million platform-based gig workers in the global South; 10 million digital and just over 20 million physical.  And that the figure could be more than 40 million, which would be around 1.5% of the global South workforce.

A proportion of these workers are not relying on this as their primary source of income.  For digital gig workers, this number is anything from two-thirds to a half[9].  It may be somewhat less for the physical gig economy, so another ballpark would be that around 15-20 million workers in developing countries are relying on digital platforms for their primary source of income.

(Annual turnover is an issue for another day but, globally and summing figures for the digital gig economy[10] and main physical gig platforms Uber[11] and Didi Chuxing[12], it must be at least US$50bn.)


[1] Adapted from Constantinides, P., Henfridsson, O., & Parker, G. G. (2018). Introduction—Platforms and Infrastructures in the Digital Age, Information Systems Research, 29(2), 381-400

[2] Heeks, R. (2017) Decent Work and the Digital Gig Economy, GDI Development Informatics Working Paper no.71, University of Manchester, UK

[3] An alternative approach would seek to extrapolate in terms of numbers of Internet users but that is correlated with GDP, and the figures still point to a strong under-representation of Africa in platform labour and strong over-representation of China.  Put another way, factors other than wealth and Internet access are needed to explain national differences in the proportions working in the platform economy.

[4] E.g. and

[5] E.g. and and

[6] e.g. and

[7] E.g. and

[8] E.g. and and and

[9] Heeks, R. (2017) Decent Work and the Digital Gig Economy, GDI Development Informatics Working Paper no.71, University of Manchester, UK

[10] Heeks, R. (2017) Decent Work and the Digital Gig Economy, GDI Development Informatics Working Paper no.71, University of Manchester, UK

[11] E.g.

[12] E.g.

%d bloggers like this: