By Chris Maclay

When I started my career in youth employment over a decade ago, a prevailing narrative pervaded the NGO space: “This is Jessica. She used to be unemployed. Now she has a chicken. And she is therefore employed.”

This narrative wasn’t true then and it isn’t true now. Jessica wasn’t all of a sudden employed because she gained one part-time income stream from her chicken. Moreover, she probably wasn’t unemployed before. Few young Africans really are. They simply can’t afford to be. In fact, Africa has a lower unemployment rate than the world average. 

The problem is huge underemployment – 80% of jobs on the continent are in the informal sector, with often low pay and insecurity. Adding new income streams (or better or more resilient income streams) is important, but does it necessarily count as ‘a job created’? What if it was just a dollar? What if this additional income stream caused the loss of another income stream? If any single additional income stream doesn’t necessarily truly constitute ‘a job created’, what does? These questions are all the starker in the jobtech space.

Job Counting, Not Just Chicken Counting

Jobmatching platforms (Brighter Monday, Jobberman, Fuzu, etc) should be relatively easy to count jobs from: an individual matched to a full-time role equals one job created (let’s leave the attribution question for another day). Most other jobtech platforms — from ride hailing (think: Uber), to gigmatching (think: Taskrabbit), to creator platforms (think: YouTube) — offer largely, though not exclusively, part-time jobs. Not because they don’t necessarily provide a good income, but because most people tend to combine these income streams with others in their portfolio of work. 

Even for ride hailing, which in Africa (as opposed to Uber’s original design in the USA) usually constitutes a full-time role, few drivers actually earn their full salaries from any one app. Rather, they multi-home, taking jobs from Uber, Bolt, Yego, and any other number of ride hailing apps on their phone. So at what point can any one of these platforms say that they have created a job? This blog will explore how to tackle the issue.

The Five Fallacies of Jobtech Job Counting

We’ve faced this counting challenge in the Jobtech Alliance, where we support a range of startups through Growth Support, to enable them to create: (1) More jobs, through market expansion or launch of new products; (2) Better quality jobs, through higher user incomes or utilization; (3) More inclusive jobs, reaching traditionally marginalized populations like women and refugees. We’ve built a model for counting jobs influenced by a range of phenomena that we call the five fallacies of jobtech job counting:

1. Jobs that are not jobs at all: We see a huge number of platforms claiming ‘empowerment’ through mere usage. This is very often perpetuated by donors seeking scale numbers. When the formal sector is so small in Africa, a traditional job-matching platform can feasibly have millions of users (jobseekers) registering accounts and (possibly though not necessarily) looking for jobs, but only hundreds of them actually getting jobs. This is not to understate the value of e-learning platforms, or the learning experience that can come from applying for jobs, but these are not jobs in themselves.

2. Jobs that are bad jobs: Some platforms provide work that is unsafe (from construction gigs without PPE to online work full of cyber-bullying), or pays extremely poorly. This is often a concern for the ride hailing platforms, which — on the face of it — often provide high gross incomes, but when one considers costs (fuel, car leasing) can result in take-home incomes that are far below minimum wage.

3. Jobs that don’t matter at all/are barely jobs: We’d argue that a platform that matches people to gigs, but only ever gives them a single KES 500 or NGN 2,500 or CFA 2,000 (<$5) per gig is not a job creator. If we counted any single payout as a job, AI kingpin Sam Altman’s crypto identity initiative Worldcoin would probably be Africa’s biggest employer in 2023, having given 350,000 young Kenyans $50 to seize their identity “empower” them through blockchain.

4. Jobs that don’t really matter for the context of the user: We could decide on some income cut-offs, but these are difficult to standardize. A reliable additional $30 a month for someone displaced in northern Nigeria could be an incredibly meaningful contribution to a livelihood, but that same $30 for an Egyptian user of a platform which matches lawyers to full time jobs would not. Context and user is everything.

5. Jobs that are part-time by design: While we would naturally want every user of a platform to be earning the equivalent of full time roles, the shifting nature of work (indeed, the ‘future of work’ at large as we’ve discussed previously) is likely going to involve more and more people finding their ‘employment’ by a portfolio of work rather than any single ‘job’. Part-time jobs aren’t inherently a bad thing. And many platforms’ business and operational models mean that they ultimately provide part-time work by design. Examples include:

• The Uber case cited  above, where drivers are on multiple platforms simultaneously.
• Many platforms for digitizing microenterprises can offer meaningful improvements to self-employment (improved supply chain, margins, efficiencies) but do not create new work, they simply augment or improve it.
• A Jobtech Alliance acceleration portfolio company, Wowzi, provides gigs to nano-influencers – i.e. ‘post about how much you love Coca-Cola on your Facebook page and receive $5’ – starting with people with just 250 friends or followers. While some of Wowzi’s power users can earn full-time equivalent incomes through high-value posts, most users only earn from lower-priced postings, and by design, only receive a finite quantity of such gigs per month, as the social value of nano-influencing has diminished returns if the same individual is posting about a different product every day. Nevertheless, $30 for 5 posts and a few hours of work per month is a pretty good flexible additional income for the young people who tend to use it, and can be particularly valuable for people like young mothers who seek to work on their own timelines.

A model for counting jobs in jobtech

So where does this leave us? At the Jobtech Alliance, we’re pretty strict about what we consider to be a job created or improved. Based on our experience in the space, and engagement with hundreds of platforms, we’ve developed a model that we’re currently using to count ‘what is a job’. And context is indeed everything. We currently have 10 platforms in our acceleration portfolio, and expect to bring on up to 30 more over the next few years. This means a lot of work with platforms to co-create what counts as a job, and a great amount of access to platform data to verify the jobs being created. We’re grateful to our platform partners for their deep collaboration.

For each platform, we develop a definition of ‘what is a job’ based on three things:

Inclusion criteria: the minimum standards for a platform to consider a job as  ‘created’ or ‘improved’. This is highly variable depending on the platform:

1. Income:
   1. Equal to or more than the minimum wage in the country of operation, with consideration for geography within country
   2. For part-time work, the hourly rate must be commensurate with the hourly rate of minimum wage in that country. We don’t consider anything below $30/month as sufficient to count as a job, regardless of the number of hours worked. Though any income is appreciated for many young people, we want at least this threshold to be met.
   3. We look at net income by accounting for the lost costs of delivering e.g. net profit for entrepreneurs, and considering cost of fuel/car in ride hailing.
2. Safety:
   1. We do not work with platforms that put their users at risk. Where we do see gaps, this might be part of our intervention — to help them improve security on their platforms.
3. Timeframe:
   1. We choose a suitable regularity of income before we count the job. Initially we considered this to be 3 months of consecutive income above our threshold, or three months in six months if the work was particularly “lumpy” (eg. someone on Fiverr may complete a big project every two months and only get paid then, rather than each month). However, so far we have learned that the original 3 consecutive months threshold might not have been the right one – we will discuss this more in our conclusion.
4. Contribution to livelihood:
   1. We consider contribution to overall livelihood portfolio to judge the right levels of income — for example, $30 part-time work could be valuable to someone otherwise unemployed or unskilled, but would not be a valuable contribution to someone already with a business generating $1,000 a month.

Exclusion criteria: While the above helps us to define what is ‘in’, we also need to consider what is ‘out’. We ringfence which jobs we believe we have contributed to e.g. if we are helping a platform expand to Uganda, we won’t count its jobs in Kenya; if we’re helping them to improve their female participation, we won’t (with exceptions) count new sign-ups of men.

Hidden hierarchies: In some cases, there may be hidden jobs created beyond platform data e.g. one carpentry workshop is listed on a platform, but 10 individuals could be working in that shop. When that is the case, we would identify a multiplier based on reasonable evidence, and the inclusion/exclusion criteria would need to match the hidden hierarchy — incomes would need to be suitable for all the workers.

Implications: moving from Average Income Per User, to Percentage Achieving A Job

We believe this work of defining ‘what is a job’ for different platforms has a secondary impact: it helps them to understand a suitable level of engagement for their users. We’ve noticed that most platforms are interested in the performance of their users, but they’re normally using the wrong metric. 

Platforms tend to count either total payout to users or average revenue/payout to users (commonly known as ARPU). The problem is that many jobtech platforms, particularly marketplaces, have a very uneven distribution of work, which we have explored previously; on creator platforms like YouTube, for example, the top 10% of users earn 73% of the income. So dividing total payout by number of users may be masking the fact that most users are earning significantly below the thresholds that would be deemed quality. This is bad for users, but also for platforms — unsatisfactory levels of income will result in user churn. 

There are only a few platforms that we know of that effectively build with this consideration. Some ride-hailing platforms balance their activation and de-activation of drivers based on floors (a minimum number of rides per hour) and ceilings (a maximum of surge pricing). When surge pricing goes too high, they add more riders. When riders do not have enough rides per hour, they reduce riders. Such decisions have upsides and downsides for riders themselves, but demonstrate a consideration for what good looks like. We’re therefore excited to work with platforms to help them introduce these thresholds of what counts into their operational strategy.

What’s next: adapting the model as we move

We’ve only been doing acceleration for nine months, and we collect platform data every six months, which means that we still have insufficient data to know if we’ve really got the above right. From the data that we have seen, for example, we think that our initial model of 3 months of consecutive income was probably wrong. Ultimately we should have known this. One of the positives of platform work is the flexibility — when a user earns a big contract from one of their other income streams, they can focus on that; when a user has family responsibilities up-country, they can do that. We’ve seen many platforms see users come-and-go in this way. 

We’ve also seen other reasons why engagement being less regular than we expected. For example, with one of our platforms for digitizing microenterprises, we’ve seen usage becomes less frequent as retailers grow as they may move from purchasing small amounts of inventory on a bi-weekly or monthly basis, to much bigger purchases on a bi-monthly basis. So we’ll likely need to update our ‘What Counts’ definitions to recognise this.

We’re not making any moves to change things yet, as we want to be guided by the data. Our next data collection will take place in January, so you can expect a follow-up blog to this one in early 2024 explaining what we’ve learned!

The author is the Program Director of the Jobtech Alliance at Mercy Corps