I have been thinking about data in Africa for a while now, and I believe it represents one of the most consequential opportunities of the next decade. Not data in the abstract, sanitised way Silicon Valley talks about it, but data as it actually exists here. Fragmented, undervalued, and waiting for someone to connect the dots. What follows is a thesis I have been developing, partly through observation, partly through intuition, and partly through watching how similar patterns played out elsewhere. I am putting it out publicly because I have had these kinds of predictions before and never documented them. This time, I want a record.
The opportunity breaks down into three distinct plays, each with its own logic, risks, and potential. They span from consumer to enterprise, from near-term to long-term. And while I am reasonably confident in the direction, I am less certain about the specifics of execution. That tension between conviction in the destination and uncertainty about the path is where I think the real thinking needs to happen.
1: The Aggregation Play
Africa is a segmented market. Within any given country, businesses are popping up constantly. Restaurants, lounges, bakeries, beauty studios. But there is no central infrastructure gathering information about them. If I want to find a good pastry shop in Accra, I have to navigate Instagram accounts, ask friends, scroll through fragmented posts. There is no single place I can go for discovery. The same is true for lifestyle broadly. Where to eat, where to go out, where to get my hair done. This category is going to be massive as discretionary spending grows, yet the discovery layer does not exist.
The opportunity seems obvious. Build a platform that aggregates these businesses and enables discovery. But the execution is brutally hard. These businesses do not see immediate value in getting listed on a platform because the platform does not yet have users. And the platform cannot attract users without businesses. This is the classic cold start problem, and it kills most attempts before they gain traction.
The unlock, I believe, is onboarding. If the process of getting a business onto the platform is too long, too complex, or requires too much effort, they will not bother. The four forces framework applies here. You need the pull of potential discovery to outweigh the push of switching costs, and the pull of an easy onboarding to outweigh the inertia of staying with what they know. Make listing a business take three minutes, not thirty. Make it feel effortless. That is the linchpin.
But here is where I start debating with myself. Let us say you crack the cold start problem and build a platform with real traffic. What then? The business model could evolve in several directions. Advertising, bookings and reservations, or vertically integrated tools that take you from discovery to in-store experience. Each of these is fundamentally a different business. Aggregation is a media play. You are selling eyeballs. Bookings is a marketplace. You are taking a cut of transactions. Vertical tools is SaaS. You are charging for software. The skills, capital requirements, and competitive moats are quite different for each.
My instinct is that in Africa, you probably need to build all three eventually. The market is not yet specialised enough, and discretionary spend is not yet concentrated enough, for a pure-play model to win. The company that dominates this space will likely need to own eyeballs, transactions, and tools. But, and this is important, you cannot build all three simultaneously. You have to sequence. My bet is that you lead with aggregation because it is the lowest friction entry point for businesses, it generates behavioural data even before transactions happen, and it builds the demand layer that makes the other plays possible.
One idea that keeps coming back to me is the price index angle. In markets with high price variability and low transparency, a reliable tracker of what things cost is genuinely valuable. Imagine a weekly "cost of going out" index for Accra. What a meal costs at different tiers, what drinks cost, how prices have shifted. That is content people would share. It is utility they would return for. And crucially, it gives you a reason to constantly update and a dataset that compounds over time. It might be a tighter wedge than generic "discovery" because it provides immediate, repeatable value.
2: The Behavioural Data Play
Most data initiatives in Africa are built on reported data. Surveys, questionnaires, self-assessments. The problem is that people lie. Not maliciously, but inevitably. If you ask me how often I eat out or how much I spend on entertainment, my answer will be a rough approximation at best, a self-flattering fiction at worst. Reported data is weak data.
The best data is collected as a byproduct of behaviour. This is why Meta, Google, and X are so powerful. They engineer behaviour and capture data as exhaust. You do not fill out a form telling Google what you are interested in. Google infers it from what you search, click, and linger on. The data is honest because it is observational, not declarative.
African startups, by and large, do not play this game. And for good reason. It requires patient capital and market scale that most do not have access to. You cannot spend years building a product that generates behavioural data as a secondary benefit when you need to make payroll next month. We are a cash-flow-driven environment, and these long-term data plays feel like luxuries we cannot afford.
But here is the thing. When Western capital finally comes looking for African consumer data, and they will, whoever has been quietly building behavioural datasets wins. I know Google has been running initiatives in places like Kenya to onboard local businesses onto Maps because the organic incentive to list there is not strong enough yet. They are trying to build the data layer themselves. But they are doing it from a distance, with limited context, and without the local understanding that would make the data truly rich.
The question I keep circling is this. What is the utility layer that captures behavioural data? Fintech has done it. M-Pesa and its successors generate enormous datasets as a byproduct of facilitating transactions. Social has done it. WhatsApp knows who talks to whom, when, and how often. What is the next utility layer that has not been captured? I suspect lifestyle is part of the answer. If you can build something people use regularly to navigate their leisure and consumption choices, you are sitting on behavioural gold.
And here is where AI changes the calculus entirely. Behavioural data is no longer just valuable for targeting ads or understanding markets. It is training data. Foundation models, the large language models and AI systems that are reshaping every industry, are trained predominantly on Western data. When companies want AI that actually works in African contexts, that understands local languages, consumer patterns, cultural nuances, and business dynamics, they will need African data. The startup that has been quietly collecting behavioural signals for seven years is not just sitting on business intelligence anymore. They are sitting on the raw material for localised AI.
3: The Enterprise Data Brokerage Play
The first two plays are consumer-facing. This one is different. A couple of years ago, I was judging a startup pitch competition, and I noticed something interesting. Almost every agri-tech company mentioned that they were gathering data. They spoke about it as if it were a strategic asset, a moat they were building. But I knew, and I suspect they knew too, that they were not going to do anything meaningful with that data. Their core business was not data. It was whatever agricultural service they were providing. The data was a byproduct, and byproducts get neglected.
This is true across sectors. Logistics companies have data on movement patterns. Healthcare startups have data on patient flows. Retail tech has data on purchasing behaviour. Each of these companies is sitting on potentially valuable datasets, but none of them has the incentive, capability, or business model to actually monetise it. The data just sits there, unstandardised, uncleaned, and unused.
The opportunity is a business that works with these startups to standardise how they collect data, clean it up, and then broker it to buyers who can actually use it. You provide value to the startups by helping them realise revenue from an asset they are currently ignoring. You build a library of datasets across industries. And you make money by connecting data supply to data demand.
The challenge, and I am honest that I do not have a complete answer here, is trust. Startups are protective of their data because they believe it is strategic, even when they have no strategy for using it. There is a kind of data hoarding instinct that kicks in. "This might be valuable someday, so I should not share it." To overcome that, the broker needs to demonstrate value before asking for data. Maybe you start with analytics-as-a-service. Show startups what their data could tell them, help them see patterns they are missing, and let the brokerage relationship emerge from there.
I think this is a viable business precisely because it is hard. Getting the right partners, building trust, figuring out the legal and ethical frameworks. None of this is easy. But defensibility often comes from difficulty. If it were simple, someone would have already done it.
The AI angle makes this play even more compelling. It is not just about selling data to researchers or consultants anymore. It is about packaging cross-industry African datasets that can train vertical AI models. An AI that understands African agriculture needs exactly the kind of data those agri-tech startups are sitting on. An AI that can navigate African logistics needs movement and delivery data. The broker who standardises and cleans this data is not just a middleman. They are building the foundation layer for African AI.
The AI Multiplier
I have been describing these three plays as if data is the end product. But AI changes the frame. Data is not the product. It is the input for the next wave of value creation. And this is where Africa's current position becomes strategically interesting.
Right now, African data is undervalued because the market is small. Global AI labs are not prioritising African training data because African markets do not move their revenue needle. But AI inverts this logic. Scarce data from underrepresented contexts becomes more valuable, not less, because it fills gaps in models that are saturated with Western training data. There is an arbitrage here. Collect data now when no one is paying attention, sell it later when AI companies realise their models do not work in Lagos or Nairobi or Accra.
Think about what happens when a global company wants to deploy AI in African markets. Their models, trained on American and European data, will stumble. They will not understand local languages properly. They will not grasp consumer behaviour patterns. They will not navigate the informal economy. These companies will need African data to fine-tune their models, and they will pay a premium for it because the alternative is building from scratch.
The question I do not have a complete answer to is this. Who exactly is the buyer? Is it global AI labs like OpenAI, Anthropic, or Google who want to improve their models' performance in African markets? Is it African governments building sovereign AI capabilities? Is it multinationals entering African markets who need AI that understands local context? Each buyer has different requirements, different willingness to pay, different timelines. The thesis is stronger with AI in the picture, but the go-to-market becomes more complex.
What I am confident about is the direction of travel. AI is going to be everywhere, and AI needs data to work. The companies and countries that control relevant, high-quality data will have leverage. Africa has a window to build that data infrastructure before the AI wave fully arrives. Whoever builds the pipes, for discovery, for behavioural capture, for enterprise brokerage, is not just building a data business. They are positioning themselves at the foundation of African AI.
The Defensibility Question
A reasonable objection to all three plays is this. What stops Google, Meta, or some other well-capitalised global player from coming in and eating your lunch? They have more money, more talent, and more patience than any African startup. If the opportunity is as big as I am suggesting, why would they not just build it themselves?
My answer is timing and context. Africa is not yet consequential enough for these companies to invest seriously in building data infrastructure here. Look at the Paystack acquisition by Stripe. Stripe did not build an African payments company from scratch. They bought one. They had the foresight to see Africa as part of their global ambition, but they preferred to acquire someone who had already done the hard work of building in context. That is the pattern I expect to repeat.
Africa is big enough that if you build something properly, global players will care. But it is not big enough, yet, for them to prioritise building it themselves. That creates a window. The startup that spends seven years quietly building the data pipes, understanding the local context, solving the trust problems. That is the startup that gets acquired for a couple hundred million when the big players finally decide Africa matters.
This is not defensibility in the traditional moat sense. It is positioning for acquisition. And that is a legitimate strategy. You are not trying to build a fortress that keeps the giants out forever. You are trying to become the obvious target when they want in.
What I Do Not Know
I want to be clear about the limits of this thesis. I have not done deep research on who is already building in these spaces. This is intuition and observation, not rigorous analysis. I am sure there are people already working on versions of these ideas, and I would love to learn from what they have discovered.
I also do not have complete answers on execution. The aggregation play requires cracking onboarding, but what specifically makes onboarding frictionless enough? The behavioural data play requires a utility layer, but what is the product that people use daily and generates rich signal? The enterprise brokerage play requires building trust, but what is the first transaction that proves value? These are the questions that separate thesis from company.
What I am confident in is the direction. Over the next five to ten years, data is going to become a crucial part of Africa's digital economy. The continent is on a ten-year delay from Western trends, which means the patterns we have seen play out elsewhere are coming here. Whoever builds the infrastructure to aggregate, capture, and broker data will be positioned at the centre of that transformation.