Consider the wildest dream of a profit-maximising corporation: charging every customer exactly the maximum amount they are willing to pay at the moment they want to buy. Economists call this first-degree, or perfect, price discrimination. For most of history it was structurally impossible—a shopkeeper cannot look into your soul, estimate your net worth, and change the price tag on a litre of milk before you reach the till. But a technology platform intermediating millions of smartphone transactions does not need to read anyone's soul. It only needs to read their data.

Evidence from the United States

A recent investigation by Consumer Reports in the United States found that Uber and Lyft routinely charge different customers wildly different prices for the same ride, requested at the same time, from the same location. The median gap between the highest and lowest prices was about 50%. The platforms vehemently deny engaging in what regulators now call surveillance pricing—pricing based on individual demographic or behavioural data—and insist the differences simply reflect real-time marketplace conditions. The functional result is the same: the software sets a price based on what it believes it can extract from you at that exact second.

The term is not fringe rhetoric. In 2024 the US Federal Trade Commission opened a formal inquiry into surveillance pricing, sending information demands to major intermediaries to understand how browsing and personal data feed differentiated prices—an early sign that regulators are only beginning to build a vocabulary for the problem.

The code goes to Kenya and Nigeria

This dynamic is striking in highly regulated Western markets, but it is reshaping the foundational economics of Africa's emerging markets, where digital platforms are rapidly formalising historically informal economies using vast data networks.

Ride-hailing is a prime example. In Kenya and Nigeria, international giants such as Uber and Bolt have deployed dynamic pricing software that determines both the passenger's fare and the driver's cut. The software is ostensibly an impartial market-clearing mechanism—raising prices when demand is high to coax drivers onto the road, lowering them when it is quiet—but it optimises for platform liquidity and revenue rather than macroeconomic stability.

In Nigeria, petrol prices recently spiked by nearly 50% due to global supply disruptions and the removal of local subsidies. A human taxi driver would immediately raise base fares to cover fuel costs; software only cares about the demand curve. The automated pricing left Nigerian drivers with shrinking margins and effectively no control over their own unit economics, prompting the Amalgamated Union of App-Based Transporters of Nigeria to launch strikes. Workers are realising they are not independent contractors negotiating in a market—they are variables in an optimisation function. The system is so rigid that competitors like inDrive have secured 150 million downloads globally largely by treating price negotiation as a product feature rather than an automated output, letting drivers and riders haggle directly.

A bespoke megabyte just for you

The telecommunications sector is where automated price discrimination becomes philosophically strange. In Kenya, Safaricom has invested heavily in predictive data analytics to pivot toward what it calls customer obsession. Using a platform co-developed with Huawei called Idea-to-Cash, Safaricom dynamically matches subscribers with highly personalised data and voice offers. The system analyses a customer's usage behaviour, spending patterns, and real-time context to generate bespoke "Just for You" packages.

The approach has been a major corporate success. Safaricom doubled conversion rates on targeted packages, reduced time-to-market for new offerings by 90%, and drastically cut customer complaints about hourly data bundles. Yet a megabyte of data is a pure commodity: it costs the network the same to deliver to Customer A as to Customer B.

Safaricom is essentially isolating consumers into individualised micro-economies by using data analytics to price telecom services dynamically. If the system infers a high willingness to pay from past consumption, that customer receives a different deal than a highly price-sensitive student. The product remains standard, but the pricing becomes a behavioural trap. When everyone receives a personalised discount, the baseline price is essentially meaningless—it means the software has calculated exactly how much perceived discount is required to trigger a purchase from a specific psychological profile. All of this operates against the backdrop of Kenya's Data Protection Act of 2019, which restricts automated decision-making on personal data without consent—a framework whose enforcement capacity, rather than its text, will determine how far personalised pricing can go.

Your battery is low, so your interest rate is high

Nowhere is this digital alchemy more visible than in African digital lending. Over the past decade, platforms like M-Shwari, Tala and Branch pioneered uncollateralised lending to unbanked populations by replacing traditional credit bureaus with alternative mobile phone data.

The original thesis was simple and brilliant: without a formal credit history, a program can ingest thousands of data points from a phone—including mobile money transaction histories, the diversity of a contact list, texting patterns, and mobility patterns based on cell towers—to build a predictive credit scoring model. The benefits are undeniable. This technology has driven massive financial inclusion and expanded formal financial access in Kenya. A Harvard Business School study found that marginal borrowers randomly approved by Tala experienced significant improvements in financial well-being, especially when they used the capital for small enterprises.

But the cost of this automated access is astronomical, and the software is ruthlessly efficient at pricing risk. Because administrative costs are practically zero, digital lenders can profitably extend tiny loans to millions of people. Because default rates are high, the system offsets this by charging what is effectively a poverty premium—learning exactly what exorbitant interest rate a desperate borrower will tolerate to secure short-term liquidity. This creates a doom loop in which borrowers take loans from one app to pay off the massive interest on another. By 2025, borrowers had defaulted on 83.1% of digital loans below KES 1,000 ($8), illustrating the difficulty of recovering small loans from low-income households. The Central Bank of Kenya (CBK) eventually had to step in with the Business Laws Amendment Act to regulate the sector, mandate transparency, and stop lenders from extracting all the wealth from the bottom of the pyramid—part of a broader CBK licensing regime that brought digital credit providers under central bank oversight starting in 2022.

Recommending things you did not know you could not afford

It is not only ride-hailing, lending, and telecoms. Big tech platforms across the continent are quietly deploying predictive software to shape consumer reality. Jumia, the pan-African e-commerce giant, relies heavily on personalised recommendation systems that analyse browsing behaviour, purchase history, and user preferences to tailor product suggestions in markets like Lagos, Nigeria.

In a Western market, a program suggesting a pair of shoes is simply targeted marketing. In an emerging market characterised by severe socio-economic disparities, varying levels of digital literacy, and trust deficits, a program dictating what products a user sees effectively constructs a walled garden. The software learns what a user can afford and hides the rest of the economy, converting structural inequality into a perfectly optimised, frictionless digital experience.

The endgame of perfect extraction

Technology circles tend to view this software deployment as an unalloyed good. The benefits are real: standing in the rain in Nairobi, getting matched with a driver via a price surge is arguably better than being stranded; needing inventory for a fruit stall, a 150% annual percentage rate loan approved by a computer in three seconds beats dealing with local loan sharks. The efficiency gains are spectacular.

The problem is the endgame. Predictive models deployed by profit-driven platforms are designed not to distribute social goods but to extract economic surplus. Surveillance pricing and automated product placement thrive in environments marked by market failures and information asymmetries. In a perfectly competitive market, a telecom provider charging a power user more for data would simply prompt them to switch providers. But African digital markets are rarely perfectly competitive—they are dominated by tech oligopolies such as Safaricom in Kenyan telecoms, Uber and Bolt in mobility, and large banks in digital credit.

When these dominant firms deploy proprietary software, the program ceases to be a tool for market discovery and becomes a tool for maximum extraction. Consumers have zero visibility into market benchmarks, while platforms have perfect visibility into a buyer's wallet and desperation. Eventually, the concept of a fair, standard price vanishes entirely. Consumers are no longer participating in a broader economy but trapped in a bespoke, algorithmically generated simulation designed to empty their pockets as efficiently as mathematically possible. The wildest part, as the original report notes, is that the software will convince them they are getting a discount. How this plays out will depend heavily on whether regulators—drawing on tools like Kenya's Data Protection Act and Nigeria's Data Protection Act of 2023, and on precedents from the FTC's surveillance pricing inquiry—can force transparency into systems whose entire advantage depends on remaining opaque.

Source: TechCabal, by Kenn Abuya, senior reporter and lead of the Startups Desk.