Digital lending software in Kenya earns its keep on the day a borrower stops paying. Everything before that day is onboarding, pricing and speed.
The hard part is what happens afterwards. A lender needs to know the balance, the promised date, the contact history and the provisioning impact without rebuilding the file from messages.
What a lending book looks like outside a spreadsheet
Most lenders start on a spreadsheet because it is free and flexible. A few hundred loans later, the same file has six versions and nobody trusts the totals.
The second problem is version control. Two officers update different copies, a director emails a third, and the board pack quotes whichever file arrived last. A single book inside digital lending software in Kenya removes that argument entirely.
Spreadsheets hide their own errors. A row sorted without the balance column, a formula dragged one cell short, and the ageing report quietly misstates the book.
Loan-level truth needs a ledger, not a grid. Every disbursement, accrual, repayment, fee and waiver should post against a loan account that can be reconstructed later.
Daily accrual is where spreadsheets fail most often. Skipping Sundays and public holidays understates income every month, and nobody notices until the audit. Interest in digital lending software in Kenya should accrue across weekends and holidays without a manual adjustment.
That is the minimum standard for digital lending software in Kenya. Anything less is a reporting tool pretending to be a core system.
CBK licensing and the Digital Credit Providers rules
Digital credit providers in Kenya sit under the Central Bank, following the 2022 regulations that brought the sector into the formal perimeter. Licensing changes what the software must evidence.
Licensing also fixes who may own the business and how customer money is handled. Those rules reach into the platform through user roles, approval limits and segregation of duties. A system that lets one person do everything is hard to defend in a supervisory visit.
Capital, governance and consumer protection requirements all leave traces. A lender must show how complaints were handled, how charges were disclosed and how data was protected.
Supervision is document-based. Examiners ask for policies, sample loan files, pricing disclosures and complaint logs, then check whether the system’s records match the narrative.
Annual returns, complaint statistics and pricing schedules are easier to produce from a system than from a folder. digital lending software in Kenya that stores them beside the loan book saves a week every quarter, and keeps the versions consistent.
A platform that cannot produce a complete loan file on request turns a routine inspection into a long week. Digital lending software in Kenya should make the sample easy to assemble.
Onboarding: ID, selfie and the customer register
Onboarding is where growth and control pull against each other. A borrower wants money in minutes, while the lender needs identity, contactability and a defensible consent record.
National ID verification is the starting point. Checking the number against the register, matching the name and capturing a selfie reduce the simplest form of impersonation.
Phone ownership matters as much as identity. The M-PESA number used for disbursement should belong to the applicant, and a recent SIM swap is a warning rather than a formality.
Location data is often collected without a clear need. If the lender cannot explain why a borrower’s GPS trace is retained, the safest answer is not to collect it. digital lending software in Kenya should make each location request plain to the customer.
Consent needs to be specific. Sharing data with a credit bureau, sending marketing messages and processing location data are different permissions.
A blanket tick box satisfies nobody. The platform should capture each purpose separately, with a timestamp and the wording the customer actually saw. That record is what digital lending software in Kenya should preserve for every applicant.
Duplicate borrowers are a quieter problem. The same person may apply under a slightly different name or a second phone, and the system should flag the overlap before a second limit is issued.
Agent-assisted onboarding needs its own controls. The person who fills the form should not approve the loan, and the handset used should be recorded against the application. Otherwise a single agent can manufacture a book that looks healthy for two months.
Credit scoring when the file is thin
Most applicants have no formal credit history. Scoring therefore leans on mobile money behaviour, utility payments, employment signals and the lender’s own repayment record.
Alternative data is useful but easy to misuse. Airtime purchases and betting transactions correlate with something, though not always with willingness to repay.
Thin-file lending rewards discipline over complexity. A simple scorecard with five well-understood variables often beats a model nobody can explain to a supervisor. digital lending software in Kenya should show which variables drove a decision, not just the final score.
Model design should start with a policy question. Who is the lender willing to serve, at what ticket size, and what loss rate is acceptable for that segment?
The score then ranks applicants within that policy. Digital lending software in Kenya that produces a number without a policy behind it gives false comfort.
Feature stability matters more than sophistication. A model trained on a promotional period, or on borrowers acquired through one agent, may not survive a normal month.
Monitoring catches that drift. Approval rates, score distributions and vintage loss curves should be reviewed monthly, not once a year. The review is easier when digital lending software in Kenya keeps the history intact.
Reject inference matters when only approved applicants have repayment histories. Ignoring it biases the model towards customers the lender already likes. That is why digital lending software in Kenya should log declined applications with enough detail to support the analysis.
Pricing: reducing balance, flat rate and disclosure
Interest presentation is a consumer protection issue as much as a commercial one. A flat monthly rate on the original principal is not the same as a reducing balance charge, even at the same headline number.
The difference compounds over a twelve-month loan. Borrowers compare the advertised rate, not the effective cost, and regulators increasingly expect the gap to be explained.
Seasonality distorts pricing comparisons. A thirty-day loan priced for a salary cycle behaves differently from a stock loan repaid after a market season. digital lending software in Kenya should let each product carry its own calendar and repayment assumptions.
Disclosure should show total cost, not just the rate. Fees, insurance, processing charges and excise duty all belong in the same illustration the customer sees before accepting.
That illustration must match what the ledger will actually charge. A mismatch discovered at collections is a dispute, a refund and sometimes a penalty.
Configurable products help here. A lender may run a salary advance at one price and a stock loan at another, each with its own schedule and grace period.
Early settlement deserves a rule. Charging the full schedule when a borrower clears early invites complaints, while waiving everything can be exploited. digital lending software in Kenya should apply the disclosed policy automatically and show the saving before payment.
The platform should compute the schedule once and use the same numbers for disclosure, invoicing and collections. Digital lending software in Kenya that recalculates in three places will eventually disagree with itself.
Disbursement over M-PESA and bank rails
Disbursement looks like the simplest step and fails more often than expected. A B2C request can time out, a recipient can be barred, or a name mismatch can hold a payment.
Timeouts are the dangerous case. The request may have succeeded at the provider while the lender’s call returned an error, and a retry then pays twice.
Provider status checks are not instant. A payout may sit pending for minutes, so the platform should hold the loan in a matching state. Marking it disbursed inside digital lending software in Kenya before confirmation creates a balance that never existed.
Idempotency is the defence. Each payout should carry a unique reference, and the platform should reconcile status before attempting a retry.
Bank rails bring their own delays. EFT batches cut off in the afternoon, and a payment sent on Friday may land on Monday while the borrower waits. A status message from digital lending software in Kenya prevents most of those calls.
The customer message should reflect reality. Promising instant money on a bank rail creates support tickets that no agent can resolve.
Cash disbursement through agents still exists in parts of the market. It needs vouchers, agent float tracking and a reconciliation routine, which is a different discipline from an API call.
Bank and mobile rails differ in cut-off times and reference formats. The reconciliation routine should normalise both into one settlement view, with unmatched items visible the same day. digital lending software in Kenya that hides the exceptions turns a small gap into a month-end project.
Repayment collection and the promise-to-pay
Collections begin before the due date. A reminder two days early costs less than a call two weeks late, and it preserves the relationship.
The promise-to-pay is the core object. A borrower says they will pay on Friday, an agent records the amount and date, and the system schedules the follow-up automatically.
Without that record, every call starts from scratch. The borrower repeats the story, the agent guesses at the history, and the promise quietly disappears.
Contact frequency needs a cap. Calling a borrower six times a day is harassment, and it wastes agent time that could go to accounts with a real chance of payment. digital lending software in Kenya can rank the queue by likelihood rather than by noise.
Digital lending software in Kenya should show the last contact, the promise, the kept or broken status and the next action on one screen. That single view is worth more than most analytics dashboards.
Payment plans need discipline. A restructure that simply moves the due date without changing affordability usually produces another missed promise. digital lending software in Kenya should test affordability before approving the change.
Channel choice matters too. A WhatsApp reminder reaches some borrowers; a call reaches others; a formal notice is required for a subset.
Language matters in collections. A message in English to a borrower who thinks in Kiswahili or a local language is less likely to be read, let alone answered. Templates inside digital lending software in Kenya should exist in the languages the book actually speaks.
CRB reporting and the dispute queue
Credit reference reporting is a regulated activity with real consequences for borrowers. Lenders submit data to licensed bureaus, and errors can block a customer from a job or a loan elsewhere.
Submission discipline starts with identifiers. A wrong ID number or a transposed phone can attach a default to the wrong person, and unwinding that takes weeks.
Bureau data should be submitted on a schedule, not when someone remembers. A weekly file with a control total is easier to defend than an ad hoc upload. digital lending software in Kenya should keep the submitted file and the response together.
Positive data matters as much as negative data. A borrower who clears a facility should see that reflected, not only the original default.
Disputes need a route and a clock. When a customer says the loan was settled, someone must check the ledger, respond and correct the bureau record.
Clearance certificates matter to borrowers who need a clean record for a job or a mortgage. The process should start automatically after settlement, with the date recorded. digital lending software in Kenya that leaves it to a manual request creates a queue nobody owns.
The audit trail should show who changed what. Digital lending software in Kenya that overwrites the original submission leaves the lender defending a decision it cannot explain.
Data Protection Act: consent, retention and subject access
Kenya’s Data Protection Act governs nearly everything a lender touches. Names, phone numbers, IDs, location traces and repayment behaviour are all personal data.
Registration as a controller or processor with the Office of the Data Protection Commissioner is the starting point. So is a lawful basis for each processing activity.
Consent must be informed and specific. A borrower should understand what is collected, who receives it and how long it is kept.
Retention is where many lenders drift. Application data for declined customers is often kept indefinitely because nobody decided otherwise.
Access control is part of the same duty. A collections agent does not need the applicant’s full employment history to make a call. digital lending software in Kenya should default to the minimum view and widen it only by role.
A written schedule, enforced in the system, is the practical answer. Digital lending software in Kenya should delete or anonymise records automatically when the retention period lapses.
Subject access requests need a workflow. A customer asking for their data should get a complete file, not a screenshot of the loan summary.
Correction requests are harder. A wrong phone number or a disputed default touches the ledger and the bureau record, so the fix must propagate.
Breach response also belongs in the design. If a database is exposed, the lender must know what was in it and who was affected, quickly.
Data processing agreements with bureaus, messaging providers and cloud hosts belong in the compliance file. So does a record of where the data physically sits. digital lending software in Kenya should document hosts and regions without a separate inventory.
Interest caps, fees and the in-duplum rule
The in-duplum rule limits what a lender can recover. In essence, a borrower should not pay more than the principal plus agreed interest because of compounding or penalties.
Kenyan courts have applied the principle for decades. Lenders who rely on automatic penalty accrual sometimes discover this at the enforcement stage.
Fee structures deserve the same scrutiny. A facility fee, a processing fee and a daily service charge can together exceed the headline interest.
Transparency about penalties is not optional. A late fee that appears only after the first missed instalment is a conduct issue, even when the contract allows it. digital lending software in Kenya should show the charge before the customer accepts the loan.
Regulators look at total cost, not line items. Disclosure that separates the charges in a way only a finance officer can follow is not disclosure.
The ledger should compute the effective cost for each product and compare it with the disclosed illustration. Digital lending software in Kenya can flag products where the two diverge before a customer complains.
Waivers and refunds need approval trails. An agent who can waive interest freely is running a parallel credit policy.
Product design should make the cheap option obvious. If a borrower can save money by paying on a different day, the app should say so. digital lending software in Kenya can show the saving before the payment is made.
Loan officers, commissions and field agents
Field agents drive volume in the salary-advance and stock-loan segments. They also introduce conduct risk that shows up long after the commission is paid.
Commission design shapes behaviour. Paying on disbursement alone encourages volume; paying partly on collection aligns the agent with the book.
The system must attribute each loan to the officer who originated it. Otherwise commission calculations become a monthly argument.
Territory allocation affects portfolio quality. An officer covering too wide an area visits less often, and monitoring weakens exactly where it is needed. digital lending software in Kenya should surface collection coverage by officer rather than by branch.
Visibility matters for supervision. A dashboard showing each officer’s approval rate, delinquency and arrears rate is more useful than a sales leaderboard.
Agent apps need offline capability. A visit to a market stall in Gikomba may happen where the signal is poor, and the report must survive the gap.
Cash collection by staff is the riskiest pattern. Where it is unavoidable, receipts, daily banking and a reconciliation against the ledger are essential.
Staff fraud often starts small. A waiver here, a backdated payment there, and the pattern becomes visible only when someone reads the exception log. digital lending software in Kenya should make that log readable without a query.
Provisioning, PAR30 and portfolio quality
Portfolio quality is measured in buckets, not adjectives. Current, one to thirty days past due, thirty-one to sixty, sixty-one to ninety, and beyond.
PAR30 is the share of the book more than thirty days late. It is the number most lenders watch, and the number most easily distorted by restructuring.
Provisioning turns those buckets into a cost. Expected credit loss models, regulatory minimums and management judgement all feed the figure.
The calculation should run from the ledger, not from a manual spreadsheet. A provision that lags the book by a month misleads the board.
Write-off policy and provisioning policy should agree. A book that writes off at ninety days but provisions at one hundred and eighty is telling two stories. digital lending software in Kenya should reconcile the two in one view for the auditor.
Vintage analysis adds the dimension that totals hide. Loans disbursed in March may behave differently from those disbursed in December.
Cohort curves show whether a product is improving or quietly deteriorating. Digital lending software in Kenya that reports only monthly totals cannot answer that question.
Concentration matters too. One employer, one county or one dealer account can dominate the book and turn a sector shock into a solvency event.
Stress testing is worth doing at small scale. A two-month delay in salary payments across one employer can move PAR30 sharply. digital lending software in Kenya should let an analyst apply that scenario without exporting the book.
Restructuring, recovery and write-off
Not every default ends in write-off, and not every restructure is kindness. Some borrowers need more time; others need a firm boundary.
Restructuring should change the schedule deliberately. Extending tenure, reducing instalments or granting a payment holiday each has a cost that must be recorded.
Re-ageing a loan without disclosure flatters the arrears report. Auditors and regulators treat that as misrepresentation, not customer service.
Communication during recovery should be documented. A borrower told one thing by an agent and another by a letter has grounds for a complaint. digital lending software in Kenya should keep the letters, calls and messages in one history.
Recovery has legal limits. Demand letters, guarantor notices and court processes follow a sequence, and each step should be logged.
Third-party collection agencies need scoping. Which accounts they receive, what they may say, how payments reach the lender and how complaints are handled.
Write-off is an accounting decision, not an erasure. The debt may be written off the books while remaining legally recoverable.
Recoveries after write-off need their own ledger. Cash received against a written-off account should reduce the loss, not disappear into miscellaneous income. digital lending software in Kenya should preserve that distinction through the full life of the account.
The system should keep both facts separate. Digital lending software in Kenya that deletes the account on write-off loses the recovery history and the audit trail.
Fraud: SIM swaps, mule accounts and device signals
Fraud follows the money, and digital lending moves money quickly. The common patterns are impersonation, collusion and account takeover.
SIM swap fraud targets the disbursement number. A borrower’s line is ported, the OTP goes to the fraudster, and the payout lands in a mule account.
OTP delivery needs a fallback that does not create a new hole. Sending the code to an email address the fraudster controls defeats the purpose. digital lending software in Kenya should restrict the fallback to channels verified at onboarding.
Device signals help. A phone that has never been used for applications suddenly installing the app and requesting a maximum limit is worth a pause.
Behavioural checks add context. An applicant who cannot answer questions about their own address, or who rushes the process, may be coached.
Mule accounts are harder. Several borrowers may direct funds to one number, which the system should detect as a pattern rather than a coincidence.
Staff collusion is the least discussed risk. An officer who approves loans to relatives, or who knows which checks are not enforced, needs segregation and review.
Velocity rules catch what individual checks miss. Twelve applications from one device in an hour is a pattern, not twelve coincidences. digital lending software in Kenya should score those patterns in real time rather than overnight.
Integrations: accounting, SMS, USSD and WhatsApp
A lending platform is rarely alone. It feeds an accounting system, a CRM, a bureau, a payment provider and a messaging gateway.
Integration quality decides operating cost. A CSV export that must be keyed into another system consumes a person every week.
Accounting integration should post at transaction level. Principal, interest, fees and write-offs need separate accounts, or the finance team cannot reconcile.
Reconciliation between the ledger and the accounting system should run daily. A month-end catch-up hides errors until they are expensive to unwind. digital lending software in Kenya should post the journal automatically and flag the difference.
Messaging has to be deliberate. SMS is reliable and costs money per message; WhatsApp is cheaper but depends on template approval.
USSD still reaches feature phones. A short menu for balance checks and repayment initiation covers customers who never install an app.
Template approval with WhatsApp takes time, so plan the message library before launch. A rushed template gets rejected and delays the whole channel. digital lending software in Kenya should track approval status per template.
Payment integrations need idempotency and status checks. Digital lending software in Kenya should treat a callback as a hint to verify, not as proof of payment.
Uptime, reconciliation and audit trails
Lending runs on trust, and trust depends on records. If the platform cannot explain a balance, every other control weakens. Explaining it requires digital lending software in Kenya that posts every movement to the ledger.
Uptime matters most at month end and on salary days. Those are the moments when volumes spike and downtime is most expensive.
Reconciliation should be daily. Provider statements, bank statements and the ledger should agree, with differences investigated rather than carried.
Floats and suspense accounts need owners. Money sitting in an unresolved state is a fraud risk and a reporting error at once.
Incident reporting should include customer impact, not just technical cause. A four-hour outage on salary day is a different event from one at midnight. digital lending software in Kenya should record both, with the volume lost.
Audit trails must be immutable enough to matter. A user who can edit a repayment date without trace undermines the whole control environment.
Backups belong in the same conversation. A restore test once a quarter is worth more than a policy document promising daily backups.
Access reviews are unglamorous and effective. A former employee with active credentials is a breach waiting to happen.
Change management protects uptime. Deploying a pricing change on the last working day of the month is a choice that will be remembered. digital lending software in Kenya should support staged releases and a quick rollback.
Buying, building or renting the platform
The build-versus-buy decision depends on where the lender expects to differentiate. Scoring models and customer experience may justify investment; ledger plumbing rarely does.
Buying a mature platform brings speed and a feature set shaped by other lenders. It also brings licence costs and a roadmap the lender cannot control.
Building gives control but demands a team that understands accounting, regulation and operations. Underestimating that scope is the usual failure.
Vendor due diligence should include a client list and a reference call with a lender of similar size. Demos are designed for optimism. Buyers of digital lending software in Kenya should ask for both before signing anything.
Renting through a hosted model suits small books. The trade-off is data portability and the cost of growth.
Whatever the route, insist on data ownership. Exporting the full loan book, with transactions and audit history, should be possible without a project.
Ask what happens at exit. A vendor who cannot describe an orderly migration is asking the lender to bet the business on the relationship.
Contract terms should cover service levels, data location and exit assistance. A cheap licence with no exit clause is an expensive relationship. digital lending software in Kenya should be priced against the cost of leaving, not only the cost of joining.
Test the reporting first. If the platform cannot produce a clean PAR30, a bureau submission file and a customer statement, the rest does not matter.
Implementation should be staged. Start with one product, reconcile it for a full cycle, then add the next.
Train the collections team early. They will find the gaps that testing missed, and their feedback is the cheapest quality assurance available.
Measure the cost per loan before and after implementation. If the platform does not reduce manual effort, the business case was a story. digital lending software in Kenya should earn its fee in hours saved within the first year.
Finally, revisit the model every year. Regulation, funding costs and borrower behaviour all move, and a platform that cannot adapt becomes a constraint.