Before You Migrate to Cloud ERP: What Kenyan and Tanzanian Firms Wish They Had Known

Short Answer

Cloud ERP is rarely the wrong answer — but migrations fail when the decision is made on the monthly price instead of the five-year total cost, on the assumption of constant connectivity, on an afternoon's data upload that really takes weeks, and on a reseller chosen as a procurement detail. Settle the cost, connectivity, data and partner questions before you sign, and the cloud stops being the risk.

Cloud ERP adoption is accelerating across East Africa — and so is the distance between the brochure and the bill. The Communications Authority of Kenya counted 58.5 million data subscriptions by mid-2025, up 27.3% in a single year; Microsoft and G42 are putting a billion dollars into a geothermal-powered data centre and an Azure region for the region. The rush is real. That is precisely why rushing is the risk.

I have spent fifteen years building and rescuing enterprise systems across East and West Africa, and cloud ERP migrations fail in a depressingly familiar way. Not because the cloud is bad — it is often the right answer — but because the decision gets made on the wrong number, with the wrong assumptions about connectivity, the wrong estimate of how long the data takes to move, and the wrong kind of vendor in the room.

What follows is the conversation I wish more buyers had before they signed.

The Monthly Price Is the Easy Number — and the Wrong One

A cloud ERP is sold on its monthly headline: a clean per-user figure that fits neatly into a slide. It is the easiest thing to compare and the least useful, because it is a fraction of what you will actually spend.

The number that decides the outcome is the five-year total cost of ownership: licences across every module and every user, yes, but also implementation and configuration, data migration, integrations to the systems you are keeping, training, the bandwidth to run it, the internal owner's time, and the annual price rises that arrive quietly in year two and year three. A Strathmore University study of SAP implementations in Kenya by Charles Kirimi Muthamia (2009) put it plainly: ERP is "huge in terms of financial and human resource requirements." The monthly figure hides most of that.

Entrepreneur conducting financial planning and goal setting — modelling the five-year total cost of ownership before a cloud ERP migration — The decision is made on the five-year total cost of ownership — not the per-user figure that fits on a slide

This is also why "we'll just move to the cloud to save money" so often disappoints. The saving is real on hardware and on the server you no longer babysit. It is frequently eaten by per-user licensing at scale, by the bandwidth bill, and by the integration work nobody costed. Model five years, not one month, before you believe the saving.

Bandwidth and the Day the Internet Drops

A cloud ERP assumes a steady, affordable connection. Vendors demo it on fibre in a boardroom. Your branch in a secondary town runs it on a contended mobile link that falters at month-end — exactly when the system matters most.

This is not a hypothetical here. Research by Hannah Orwa Bula and her colleague at Kenyan public universities found that ERP rollouts had been abandoned or stalled altogether, with the two main causes being "challenges associated with institutional connectivity and limited skilled expertise to drive the implementation." Connectivity was not a footnote; it was a primary cause of failure.

Businessman working late in an office under pressure — testing what happens to a cloud ERP when the internet connection drops — A system that assumes constant connectivity behaves, in practice, like a system that is constantly at risk

Before you migrate, ask the unglamorous questions: what happens to the shop floor when the link drops for an hour? Is there a genuine offline mode, or does the screen simply freeze? What does the data plan cost when every transaction is a round-trip to a server in another country? A system that assumes constant connectivity behaves, in practice, like a system that is constantly at risk.

Data Migration Is Measured in Weeks, Not the Afternoon You Were Promised

The single most under-estimated line in any migration is moving the data. The slide says "we'll import your records." The reality is that your records live across years of spreadsheets, a legacy system nobody fully understands, and three different spellings of every customer's name.

Office worker overwhelmed by messy records and emails — why ERP data migration is measured in weeks, not an afternoon — Years of spreadsheets and three spellings of every customer's name — cleaning the data is weeks of work, not an afternoon's upload

Cleaning that — standardising codes, reconciling duplicates, agreeing one source of truth, mapping old fields to new — is weeks of disciplined work, not an afternoon's upload. Skip it and you do not save time; you simply move the discovery to launch week, when it is most expensive. A University of Nairobi study by Kimani J. Mburu (2008) of ERP at Chevron Kenya and Kenya Pipeline caught the underlying error exactly: many implementations fail "because organisations which bought the systems thought that buying them was all they require to do." Buying is the start. The data is the work. Start it before the build, not during it.

A Partner, Not a Reseller — and How to Tell the Difference

The most consequential choice you make is who implements the system, and it is the one buyers rush. In the largest Kenyan study of its kind, John Simiyu Masika (2014, Kenyatta University) ranked the critical success factors for ERP across 56 public-sector organisations — and supplier selection came first, ahead of training, process redesign, and change management. The same study found 92% of projects ran behind schedule and 43% over budget. The vendor you pick is not a procurement detail; it is the largest single lever on whether the project works.

A reseller sells you a licence and a logo. A partner takes responsibility for the outcome. You tell them apart with questions, asked before you sign:

Who, by name, has done this in my sector and my country? Ask to speak to that client directly — not a reference page.
Show me your data-migration method. A partner has a documented process and a realistic week-count. A reseller waves it away.
What does year three actually cost? Licence rises, support tiers, the price of adding users. Get it in writing.
What happens when the connection is poor? A partner will have an honest answer about offline behaviour. A reseller will change the subject.
Will you localise it? A USIU-Africa study by Michael Wagema Githiga (2018) of ERP at the Communications Authority of Kenya recommended systems be localised "to fit the user's requirements… country, language, and cultural codes." Tax rules, mobile-money reconciliation, multi-currency, the local financial year — a partner builds for them; a reseller leaves you to discover the gaps.

What "Right" Looks Like — and What "Rushed" Looks Like

It can go well. New Kenya Co-operative Creameries moved its finance, materials, sales and payroll onto SAP across eight factories and a network serving roughly 100,000 smallholder farmers; farmers who once waited up to a month for payment are now paid in near real-time. Bidco Africa ran SAP on-premise from 2010 and only moved to the cloud in 2025 — fifteen years of a stable platform before a deliberate migration, not a panicked leap. The pattern in the successes is patience and fit.

Overwhelmed manager surrounded by deadlines and reports — what a rushed cloud ERP migration looks like when the boring steps are skipped — The failures rhyme across borders: cost, connectivity and capability, in that order — usually because the boring steps were skipped

The failures rhyme across borders. A 2023 case study of a Tanzanian mining operation found that the biggest barriers to its ERP were the high cost of implementation (cited by 87.5% of respondents), IT infrastructure that could not accommodate the new system (84.4%), and a shortage of qualified people to run it (71.9%) — cost, connectivity, capability, in that order. It is why Tanzanian researchers keep reaching the same conclusion: a 2016 study by Wilfred Mwakyusa and a colleague of the Tanzania Institute of Accountancy's Epicor rollout argued a phased adoption is more suitable than a big-bang switch, and a 2023 paper from the Institute of Finance Management by J.R.A. Mhina and Daniel Lashayo showed that staff performance after go-live depends on how well the system genuinely fits the task — something you can only know by analysing the work before you buy, not after.

The Pre-Migration Checklist

Before anyone commits budget:

Model five years, not one month. Licences, implementation, migration, integration, training, bandwidth, year-on-year rises. If the saving survives that, it is real.
Pressure-test connectivity. Insist on a demonstrated offline mode and a bandwidth cost estimate for your worst-connected site.
Schedule the data clean-up now. One source of truth per function, reconciled before the build. Budget it in weeks.
Choose a partner by interrogation, not by brochure. Sector references you can phone, a written migration method, year-three pricing, an honest offline answer, and localisation.
Tie payments to working deliverables, not dates. A go-live that does not work is not a go-live.

The cloud is not the risk. Migrating on enthusiasm instead of evidence is. If your operations have outgrown the systems running them, the right first step is not a request for a quote — it is a clear-eyed look at the five-year cost, the connectivity, and the data. That is the conversation worth having first. You can explore our consulting services, read more on what works and fails in ERP across East Africa, or get in touch for a conversation.

Frequently Asked Questions

What is the real cost of a cloud ERP — and why is the monthly price misleading?

The monthly per-user price is a fraction of what you will spend. The number that decides the outcome is the five-year total cost of ownership: licences across every module and user, implementation and configuration, data migration, integrations to systems you are keeping, training, the bandwidth to run it, the internal owner's time, and the annual price rises that arrive in years two and three. Model five years, not one month, before you believe any saving.

Why does connectivity cause cloud ERP projects to fail in East Africa?

A cloud ERP assumes a steady, affordable connection, but a branch in a secondary town often runs on a contended mobile link that falters at month-end — exactly when the system matters most. Research at Kenyan public universities found ERP rollouts abandoned or stalled, with institutional connectivity a major cause. Before migrating, demand a genuine offline mode and a bandwidth cost estimate for your worst-connected site.

How long does ERP data migration really take?

It is measured in weeks, not the afternoon vendors promise. Your records live across years of spreadsheets, a legacy system nobody fully understands, and three different spellings of every customer's name. Standardising codes, reconciling duplicates, agreeing one source of truth, and mapping old fields to new is disciplined work. Skip it and you simply move the discovery to launch week, when it is most expensive.

How do I tell a true ERP implementation partner from a reseller?

Ask, before you sign: who by name has done this in my sector and country (and can I phone them)? Show me your documented data-migration method and a realistic week-count. What does year three actually cost in writing? What happens when the connection is poor? Will you localise for tax rules, mobile-money reconciliation, multi-currency and the local financial year? A partner answers each; a reseller changes the subject.

What does the research say is the single biggest factor in ERP success?

In the largest Kenyan study of its kind, John Simiyu Masika (2014, Kenyatta University) ranked supplier selection first among critical success factors across 56 public-sector organisations — ahead of training, process redesign and change management. The same study found 92% of projects ran behind schedule and 43% over budget. The vendor you pick is the largest single lever on whether the project works.

Should I migrate to cloud ERP all at once or in phases?

The evidence favours a phased adoption over a big-bang switch. Tanzanian research on the Tanzania Institute of Accountancy's Epicor rollout argued phased adoption is more suitable, and the successful cases — such as New KCC and Bidco Africa, which ran SAP on-premise for fifteen years before a deliberate cloud move — share a pattern of patience and fit rather than a panicked leap.

Sources & the Researchers Worth Crediting

Masika, J.S. (2014), Critical success factors for effective implementation of ERP in selected public service sector organisations in Kenya, MSc thesis, Kenyatta University. · Githiga, M.W. (2018), Implication of ERP systems implementation in public institutions in Kenya: a case of the Communications Authority of Kenya, MBA, USIU-Africa. · Bula, H.O. and Wang'ombe (Ndung'u), J. (2021), ERP systems and implementation experiences for selected public universities in Kenya, The University Journal. · Muthamia, C.K. (2009), Critical success factors for SAP implementation in Kenya, MSc, Strathmore University. · Mburu, K.J. (2008), Critical success factors in implementing enterprise information systems in Kenya: Chevron Kenya and Kenya Pipeline, MBA, University of Nairobi. · Mwakyusa, W.P. and Kavuta, K. (2016), Enhancing ERP adoption in government training institutions: a case of the Tanzania Institute of Accountancy, Journal of Resources Development and Management. · Mhina, J.R.A. and Lashayo, D.M. (2023), Implementing ERP in Tanzanian higher education: the influence of task-technology fit on staff performance, IJITCS, Institute of Finance Management. · Market data: Communications Authority of Kenya (Q4 FY2024/25); Tanzania Communications Regulatory Authority (2025). Vendor case studies (SAP — New KCC, Bidco) are company-published; the Tanzanian mining case statistics come from a single case study and are reported here as indicative, not definitive.