# Digital fraud in Kenya: 77% of surveyed consumers say they were targeted in three months

The number is striking, but it needs to be read carefully. In a study published on 14 October 2026, 77% of 462 Kenyan adults surveyed by TransUnion said they had been targeted by an email, online, phone-call or text-message fraud attempt in the previous three months. Across the full sample, 68% said they had been targeted but did not fall for the scam, while 9% said they became victims. The latter figure was 10% a year earlier.

This distinction matters. The share of people targeted measures self-reported exposure to attempted fraud. It is not a transaction fraud rate, a total loss figure or a measure of payment-system performance. The victim rate is also self-reported. It nevertheless shows that digital fraud has become a routine experience for a large share of connected Kenyan consumers.

For African banks, fintechs, mobile-money operators, PSPs and merchants, the key lesson is not simply that there is “more fraud”. It is that controls must cover the entire customer lifecycle: account creation, login, beneficiary addition, payment, refund and the movement of funds out of the account.

Three metrics, three different questions

Fraud discussions often mix indicators that answer different questions.

The first is reported exposure: a person says they received a message, call or solicitation they identified as fraudulent. This is what the Kenyan 77% figure mainly measures.

The second is reported victimisation: a person says the scheme succeeded. Nine percent of the Kenyan sample falls into that category. The figure does not show whether the loss was borne by the customer, reimbursed by a provider or later recovered.

The third is the suspected transaction rate observed by a provider. In its Africa-wide report published in June 2026 and based on 2025 data, TransUnion estimated that 2.6% of the digital transactions it analysed in Africa were suspected of fraud. This category is broader than confirmed fraud: it includes transactions denied in real time because of fraud indicators or internal policy rules, as well as cases classified after investigation.

Directly comparing 77%, 9% and 2.6% would therefore be a mistake. The populations, time periods, denominators and definitions differ. Taken together, however, the indicators describe a consistent pattern: many consumers are solicited, a smaller share becomes victims, and technical systems must filter a narrower but potentially costly volume of risky activity.

Risk is shifting towards identity and account access

TransUnion’s H1 2026 report describes African fraud as more organised, more targeted and increasingly identity-driven. In Kenya, 2025 data placed the highest risk at account creation. South Africa showed a different pattern: 3.0% of login attempts were suspected of fraud, compared with 2.4% at account creation and 0.7% at the financial-transaction stage.

The differences between markets rule out a one-size-fits-all model. More importantly, they show that checking only the final payment is no longer sufficient.

A synthetic or stolen identity can be used to open an account. Compromised credentials can enable an account takeover. A manipulated victim may personally authorise a payment that appears technically genuine. A mule account can then receive and disperse the funds before the alert reaches the relevant institution.

INTERPOL described the same change in scale in August 2026. Its African Cyberthreat Assessment said artificial intelligence was linked to 55% of reported cybercrime across the continent, including automated social engineering, credential harvesting and the creation of synthetic profiles. INTERPOL also highlighted a lack of real-time data sharing among banks, telecom operators and law-enforcement agencies.

What payment providers should measure

A lower headline suspected-fraud rate can coexist with more concentrated losses or attacks that are harder to detect. A useful dashboard must therefore go beyond a single percentage.

At onboarding, providers should track the alert rate, confirmed-case rate, review time, false positives and customer abandonment caused by verification. Controls that are too weak admit risky identities; controls that are too heavy block legitimate customers and undermine inclusion.

At login, teams should separate ordinary failures from account-takeover signals: a new device, SIM change, unusual password recovery, location inconsistency, rapid contact-detail changes or behaviour far outside the account’s history.

At payment, measurement should not stop at declines. Confirmed fraud, chargebacks, refunds, net losses, average value, channel, payment method and time to detection need to be reconciled. Without that view, a company can reduce approval rates without knowing whether it has actually reduced fraud.

At beneficiary level, detecting mule accounts requires analysis of incoming and outgoing flows, velocity, counterparty concentration and links between accounts. Rwanda illustrates the importance of this layer: 29% of surveyed consumers who said they lost money to fraud cited mule schemes as the leading cause.

Security should not turn every payment into an ordeal

The Kenyan consumers surveyed were not passive. Among those notified that identity or account information had been stolen, 57% changed the affected password, 54% updated passwords on other accounts and 46% checked accounts for unauthorised activity. During the preceding 60 days, 50% said they had enabled passwordless login or multi-factor authentication in response to cybersecurity concerns.

These behaviours suggest users will accept some security steps when they are understandable and proportionate. The objective is not to add the same friction everywhere, but to trigger it according to risk.

Step-up authentication can be reserved for a new device, a sensitive profile change or an unusual payment. Alerts should clearly explain the required action. Account-recovery flows deserve the same protection as the initial login. Support teams also need short, documented procedures to freeze an account, verify a report and trace a dispute.

The lesson for Francophone Africa — without copying the statistics

The consumer surveys cited cover Kenya, Namibia, Rwanda, South Africa and Zambia. They do not establish that 77% of consumers in WAEMU or CEMAC are targeted, nor that the same fraud schemes dominate there.

Their value for Francophone markets is operational rather than statistical. They provide a framework to test locally: where does risk concentrate in the journey — creation, login, transaction or cash-out? How much loss comes from customer manipulation, credential theft, account takeover or mule accounts? Which signals can issuers, acquirers, telecom operators, merchants and authorities share while complying with data-protection rules?

Before investing in another tool, every provider should be able to answer those questions with its own data. That is the best defence against two opposite mistakes: underestimating fraud as it moves towards identity, or adding controls without measuring whether they work.

The Kenyan signal does not mean that three consumers in four lost money. It means fraud attempts have become part of everyday digital life for a large majority of the people surveyed. The response can no longer rest only on blocking the final transaction. It must protect identity, account access and the movement of funds while keeping the journey workable for legitimate customers.

Sources and methodology

The consumer exposure figures come from TransUnion Africa’s study published on 14 October 2026. Data on suspected digital fraud and the stages of the customer journey are detailed in TransUnion’s H1 2026 report and its market releases for Kenya, Rwanda and South Africa.

The continental context is supplemented by INTERPOL’s African Cyberthreat Assessment released on 3 August 2026.

Cover photograph: Fiona Graham, Nairobi, 13 October 2016, via Wikimedia Commons, licensed under CC BY-SA 2.0, cropped to a horizontal format. The photograph illustrates the mobile-payment context in Kenya; it does not depict a fraud victim or incident.