The real cost of custom fintech infrastructure in Kenya

Why this article exists

Founders ask what software should cost in Kenya and get either a vague “it depends” or a single number with no idea what sits behind it. Procurement teams compare three quotes and still cannot tell whether they are buying the same thing. I am publishing ranges we actually see in 2026 for serious delivery — not weekend WordPress installs — and what Raven prices against when we scope work from Westlands.

This is not a rate card. It is a map: what moves a project to the top of a band, where agencies hide risk, and how to read a proposal so you know what you are funding.

A word on who this is for. If you need a five-page brochure site and a contact form, you do not need a product team — you need a competent web shop and a clear content owner. If you are wiring money, storing national IDs, or running payroll with statutory lines, you need engineering discipline, security review, and someone who has shipped under pressure before. The bands below assume the second category when numbers run high.

The three pricing models in Kenya

Fixed price. Scope, milestones, and acceptance criteria are written down; you pay against delivery checkpoints. Works when the problem is bounded — a marketing site with known content, a defined integration with documented APIs. It breaks when “small changes” pile up because discovery skipped edge cases. Good vendors bake contingency into margin; cheap vendors either change-order you to death or quietly cut quality.

Time and materials (T&M). You pay for capacity — day rates or hourly — with transparent burn. Best when discovery must stay honest: regulated workflows, multi-department sign-off, or integrations that need trial and error with sandbox APIs. You need weekly demos, a single product owner on your side, and discipline on backlog hygiene — otherwise T&M becomes an open tap.

Retainer. A fixed monthly fee for a defined slice of engineering and advisory — product iteration, reliability work, security patches. Makes sense when you are past first launch and need predictable throughput without renegotiating every sprint. Not a substitute for a first build; you still fund the initial delivery separately unless scope is tiny.

Most serious shops blend models. You might pay fixed price for a defined MVP, then move to retainer for six months of hardening — or T&M for discovery, then fixed price for build once scope is frozen. Anyone who quotes a single giant number without saying which model applies is asking you to sign a blank cheque on interpretation.

Typical ranges for common project types

These are indicative KES bands for vendors who staff projects properly — design, engineering, QA, deployment — not for a single freelancer quoting from a coffee shop.

• Corporate website: KES 250,000 – 800,000
• E-commerce (Shopify): KES 180,000 – 450,000
• Custom web app: KES 1.2M – 6M
• SACCO platform: KES 3.5M – 12M
• HRMS-class platform (multi-module): KES 2.8M – 8M
• Mobile app (iOS + Android): KES 2M – 7M

What pushes you up within a band: M-Pesa and bank integrations, KRA-facing reporting, role-based access across many departments, audit trails, uptime expectations, and data migration from messy legacy spreadsheets. What keeps you lower: clean requirements, one environment, a single admin persona, and acceptance that v1 is narrow.

Currency and staffing matter too. Senior engineers in Nairobi are not cheap relative to five years ago; if a quote assumes junior rates for senior work, someone is lying — either about seniority or about what will ship.

Compare apples to apples on support. A low build quote that omits a warranty period, SLA, or handover window is not cheaper — it is incomplete. Ask what happens in the thirty days after launch: who fixes bugs, who answers the phone when payments fail on Friday evening, and whether that time is included or billed extra.

Offshore teams can win on hourly rates but lose on communication lag and context. Local teams cost more per hour but may finish in fewer hours when the problem is “make Daraja callbacks idempotent under Kenyan mobile network behaviour,” not “integrate generic payment API.” Price per sprint is the wrong unit; cost per shipped, audited feature is closer to the truth.

What goes into the price — really

Discovery and specification. Workshops, process maps, API inventory, risk review. Skip this and you pay twice in rework. Serious firms charge for discovery as its own phase or fold it into fixed price with a clear exit artefact — signed scope, wireframes, integration list.

Design. Not just “make it pretty” — information architecture, form flows, error states, mobile-first layouts. For member-facing financial flows, design time is risk reduction.

Engineering. The bulk of cost. Count front-end, back-end, integrations, and data migration separately in your head even if the proposal bundles them.

QA and staging. Automated tests where they earn their keep, manual QA for regulated paths, a staging environment that mirrors production. If the quote has no line for QA, it is not serious.

Deployment and hosting. Vercel, AWS, managed databases — someone pays for SSL, backups, and monitoring. Clarify whether hosting is in your name or bundled.

Documentation and handover. Runbooks, environment variables, admin training. You cannot operate a system you cannot hand to a second vendor.

Fixed price hides risk inside margin; T&M exposes burn week by week. Pick based on how frozen your scope really is — not on which label sounds safer.

There is also an agency pyramid: partners and leads sell; seniors architect; mid-level engineers build; juniors fix CSS. A day rate should say who sits in your stand-ups. If you pay senior rates and get only junior output, your burn is mislabelled. Ask for team composition — not to micromanage, but to align risk: integrations and data migration should not be someone’s first production database.

Finally, budget for content and operations. The prettiest HR portal still fails if nobody owns employee data entry, or if finance has not signed off on payroll rules. Software quotes rarely include your internal time — but that time shows up in whether the project lands or drifts.

Where Kenyan agencies cut corners

No automated tests. Manual click-through before launch, then hope. First serious refactor breaks production.

No staging. Developers push to the same URL members use. You become the test environment.

No monitoring or alerting. Downtime is discovered by customers on WhatsApp.

Thin documentation. Knowledge lives in one contractor’s head. When they disappear, you re-buy the same discovery.

Integration theatre. “M-Pesa compatible” means nothing until you see STK push, callback handling, idempotency, and reconciliation against orders in a test harness.

You pay less on the first invoice until the first failure — then you pay again to unwind bad data, rebuild trust with customers, and hire someone to grep logs at midnight. The total cost rarely stays low.

Security is another corner: shared admin passwords, secrets in Git history, no dependency updates for a year. None of that shows up in a demo — it shows up when something breaks or leaks. Ask how dependencies are updated and who has access to production. Silence there is expensive later.

Fixed price vs time and materials — honest tradeoffs

Fixed price works when both sides can describe “done” without ambiguity: pages, roles, integrations with documented endpoints, acceptance tests. You should still expect change requests — the question is whether they are managed through a formal process with pricing.

T&M works when discovery must stay adaptive: you are replacing a workflow nobody has fully documented, or you need parallel experiments with payment providers. Your protection is cadence: weekly demos, burn visibility, and the right to pause scope when assumptions fail.

Hybrids are common: fixed price for a thin first release, then T&M for iteration — or T&M for a two-week discovery that produces a fixed bid for build. We use the model that matches how clear the problem is, not what sounds easier to finance.

From the buyer side, fixed price rewards decisive product ownership — someone who can say “no” to scope creep without reopening the contract weekly. T&M rewards teams that can absorb uncertainty and still keep a backlog ordered. If your organisation argues internally for six weeks on every decision, you will burn either way; fix governance before you blame the pricing model.

Red flags when someone quotes you

Impossible totals. A full SACCO-facing build at freelancer rates — the math does not close unless corners are cut you have not seen yet.

No discovery line item. Especially for anything touching money or personal data — they are guessing.

“We will sort M-Pesa later.” Payment rails are not a plugin; they are part of architecture.

No named environments. If they cannot describe dev, staging, and production, walk away.

Vague ownership of code and hosting. You should receive repos, deployment rights, and credentials — not a black box on someone else’s account.

Ask for a written assumptions list: user volumes, peak concurrency, browsers supported, third-party fees, who pays SSL and domains, and what happens after go-live. Silence on those is a quote, not a plan.

Another smell: the proposal is ten pages of methodology and one paragraph on your problem. You are not buying a methodology; you are buying a working system. Push for references in your sector or with similar integrations — not logos on a slide, but people you can call.

Lastly, watch for “unlimited revisions.” Unlimited usually means unmanaged feedback and a team that cannot say no — which delays launch and trains your organisation to treat software as free to change. Cap revisions in writing; you will get sharper feedback and faster decisions.

How we price at Raven Tech Group

We price from scope, integration risk, compliance depth, and what happens after launch — not from what we think your budget headline should say. Small, well-bounded projects get short proposals with clear deliverables. Larger regulated builds get phased delivery: exit criteria per phase, documented handover, and optional retainer for iteration.

We bias to honest ranges early. If your idea needs a number we cannot defend, we say so — and we tell you what would have to change (scope, timeline, or quality bar) to make the math work.

We also separate what we control from what you control. We can quote engineering and integration; we cannot quote how fast your finance team signs off on requirements — but we can structure phases so you are not paying for idle developers while approvals drift.

Maintenance is explicit: either a retainer line with hours, or a handover to your team with documentation and training priced in. “We will support you” means nothing until the channel, response time, and billing rate are named.

If you want specific numbers for your project, book a discovery call. We will give you a real range, in writing, within 48 hours of the call — not a brochure estimate copied from last year’s deck.