The Tier Trap: Why Oversized ERP Fails Smaller Organisations

01The question, and the short answer

When an ERP programme fails, the post-mortem usually names a vendor, an integrator or a budget. The evidence reviewed in this paper points to something more basic: a mismatch between the organisation the system was designed for and the organisation that bought it. A system built for a large, well-staffed enterprise assumes trained users, standard processes, clean data and a budget for continuous change. A smaller organisation, or one working in a lower-capacity environment, may have none of these. This paper calls the result the tier trap.

It tests the idea against five questions that boards and funders ask. Does the trap exist? Is it caused by influence or by poor baselining? Is money lost, and who pays? Is there a lifeline for change? And is the cause a particular vendor? The paper draws on global surveys, reviews by the IMF and the World Bank, academic case studies, auditors' reports and press accounts, and it says how strong each answer is. Nothing here is an argument against Oracle, Microsoft, SAP or any other vendor. Sources are numbered and listed at the end.

The short answer is in Exhibit 2. The trap exists, and it is the mismatch rather than the brand that carries the risk. The answers also say where the evidence is thin. No dataset of ERP failure rates for African small and medium organisations exists, and the African cases on the public record are mostly national financial systems and a few large companies.

02How ERP projects fail: base rates and mechanism

ERP failure is not an African problem, and the first task is to say so. A senior director of ERP strategy at Gartner told the firm's Symposium in November 2025 that 70% of ERP initiatives fail to fully meet their original business case goals and that 25% fail catastrophically.¹ Gartner's published research predicts that by 2027 more than 70% of recently implemented ERP initiatives will not fully achieve their goals.⁵ Across IT projects generally, McKinsey and Oxford University's analysis of more than 5,400 projects found that large ones, with budgets above US$15m, ran on average 45% over budget and 7% over schedule while delivering 56% less value than predicted.⁶ In Panorama Consulting's 2010 survey, 54% of ERP implementations exceeded their budget.⁷

The question is therefore not whether ERP is risky but whether the risk rises when the system is large relative to the organisation. Three lines of research say it does.

Design-reality gaps. Hawari and Heeks traced ERP failure in developing countries to "sizeable gaps between the assumptions and requirements built into the ERP system design" and the realities of the organisation using it.⁸

Complexity and capacity. The IMF's review of government financial systems in developing countries began from what it called "almost universal failure to implement and sustain" them. It warned that overly sophisticated systems clash with users' limited experience, and it favoured systems with modest initial goals that can grow as skills develop.⁹

Tier. In Panorama's 2010 survey of about 1,600 organisations, 30% of Tier I projects, the largest vendors, which the survey lists as including SAP, Oracle and Microsoft, overran their schedule. The figure was 18% for Tier II and 5% for Tier III (Exhibit 3).²

Two cautions keep the reading honest. The Panorama survey is old, global and measures one kind of failure, and projects at the smaller tiers are also simpler for reasons unrelated to the vendor. And large systems do get built. The World Bank's review of 87 financial-management projects in 51 countries since 1984 found 49 of 55 completed systems operational and 67% of outcome ratings satisfactory or better. It also found that comprehensive projects take at least six to seven years, and that no single package supplies all the functionality a country needs.¹⁰ Large systems can work. They work when the organisation can carry them.

03What Africa's own cases show

The public record holds national financial systems, which auditors and parliaments examine, and a handful of large-company cases. Exhibit 4 sets six of them side by side. Three patterns stand out: systems chosen without a baseline, money spent without results, and systems that are live but not used.

Selection without a baseline. Malawi's policy designers visited Tanzania in March 2005 to see an Epicor system that the donor community was praising. A contract followed in July and the system was formally introduced in November, eight months from study tour to go-live.¹⁴ Tanzania's own success, as the IMF assessed it in the same year, rested on conditions beyond the software: a prior review of expenditure-management processes, capacity building in the Accountant General's office, enabling legislation, a mid-range commercial package with a strong local partner, adequate donor support and strong political backing.⁹ Copying the package without the conditions is the clearest illustration in the record of influence choosing the system and the absence of a baseline letting it through (Exhibit 5). That reading is an inference from two sources: the Harvard case records what Malawi did and when, not what it chose not to study. A review ordered in 2009 found that access was not authenticated beyond user ID and password and that the system lacked an alert to detect fraud.¹⁴ The Cashgate fraud that followed, in which US$30m was lost from government between April and December 2013, had many causes beyond software, but it happened in a system whose controls had been questioned four years earlier.¹⁴ Even Tanzania's system, evaluated with 2012 data, had not realised some important objectives, with interoperability problems and manual and automated systems running in parallel.¹⁸

Money spent without results. South Africa's National Treasury told Parliament in May 2023 that about R2.4bn had been spent on its integrated financial management system, which had "yielded no concrete results". That sum comprised R779m on a first phase that Cabinet scrapped in 2013 and a further R1.6bn since. Oracle licences bought in 2016 for R448m had not been used.⁴ The Auditor-General classed the spending as fruitless and wasteful, which Treasury disputed. The contract is the subject of legal proceedings, and in November 2025 Treasury was reassessing the contract with Oracle while the programme had still not got off the ground.^16,19

The private sector supplies the counter-example that keeps the argument honest. SPAR, a large South African retailer with a large budget, lost about R2bn in sales, roughly US$107m, in the year to September 2024 after problems with an SAP rollout that began in 2023. Reporting on the case noted that manual pricing and margin processes hurt its margins.¹⁷ Tier mismatch was not SPAR's problem. Size did not protect it either. The tier trap is one route to failure among several; weak change management and a rushed cutover are others, and they are covered below.

Who pays. In the public cases the bill falls on taxpayers and donors. The World Bank alone has financed 87 such projects since 1984,¹⁰ and Malawi's replacement software in 2016 was funded through a World Bank project.¹⁴ In the private cases it falls on shareholders and customers, as at SPAR (Exhibit 6).

Live but not used. In Kenya, the Auditor-General's report on national government for 2023/24 states that "the benefits expected from the deployment of IFMIS have not been realised due to inadequacies of the system."¹¹ The Auditor-General also disclosed that the Oracle-based system was down for three weeks in October 2023, hindering the initiation and finalisation of procurement and payment processes.¹² It was not the first disruption: in April 2017 a breakdown disrupted payments in all 47 counties.¹³ In Malawi's local governments, where Serenic Navigator ERP was introduced from 2009, a study using 2015 data found the system, about eight years on, still used mainly as a basic accounting package, with too many administrators, untrained users, unused modules and shared passwords.¹⁵

04Oracle or Microsoft? The wrong question

A common reading of these cases blames Oracle. The record does not support it. The failures above span Oracle (Kenya, South Africa), Epicor (Malawi), Serenic (Malawi's local governments) and SAP (SPAR), and the best-documented African success, Tanzania, ran on a mid-range commercial package.^9,14 Nor does Microsoft escape the pattern: Panorama counted it among the Tier I vendors.² What differs between products is tier, and tier shows up first in price.

Microsoft publishes its prices. Business Central, its product for small and mid-sized businesses, lists at US$80 (Essentials) and US$110 (Premium) per user per month, and Dynamics 365 Finance lists at US$210.^20,21 NetSuite and Oracle Fusion do not publish list prices. Third-party estimates put NetSuite at US$99 to US$199 per user per month plus a US$999 monthly platform fee, and Oracle Fusion at US$175 to US$625.^22,23 These estimates conflict across sources and should be used to rank options, not to set budgets. For a 20-user team the annual licence bill runs from about US$19,000 to US$26,000 for Business Central to US$50,000 for Dynamics 365 Finance (Exhibit 7).

Implementation costs are harder to compare. The same third-party publisher estimates a typical mid-market NetSuite project at US$75,000 to US$250,000 over four to nine months, and an Oracle Fusion core-finance project at US$400,000 to US$800,000, rising to US$1.5m to US$3m for a full suite.^22,23 No comparable independent range for Microsoft's products was located, which is itself a finding: buyers should ask several partners for fixed-scope quotes before shortlisting (Exhibit 8).

Licences are charged per seat, so low adoption raises the real price of every active user. At Business Central Essentials' list price, an organisation in which only half of licensed users are active pays US$160 per active user per month. At Dynamics 365 Finance's list price it pays US$420 (Exhibit 9). The higher tier makes sense when its capacity is used. When it is not, adoption has to climb a steeper hill to earn the cost back.

None of this is an argument for the cheaper vendor. A higher tier is the right answer for an organisation with several legal entities, high transaction volumes or complex consolidation. The argument is that tier should follow the organisation's measured needs, whichever vendor supplies it.

05The lifeline: skills, partners and an upgrade path

Every system needs a lifeline for change: people who can configure and support it, a partner network that can fill gaps, and a vendor path for upgrades. The evidence suggests these are the scarce items. In an SAP Africa survey in the fourth quarter of 2022, 60% of surveyed organisations in Kenya, 78% in South Africa and all of those in Nigeria said tech-skills shortages had hurt them, and 53%, 73% and 80% expected a gap in the coming year (Exhibit 10).²⁴ The survey covered technology skills in general, not ERP, it did not state its sample sizes, and it was run by a software vendor. It shows, nonetheless, what a buyer is competing for.

Partners matter as much as people. The IMF credited part of Tanzania's success to a strong local partner for a mid-range package.⁹ The World Bank's review concluded that developing counterpart capacity early is crucial.¹⁰ A higher-tier product, which needs scarcer skills, raises the bar on both.

The upgrade path matters because systems change under their users. In Kenya, press reporting on the Auditor-General's report describes updates to the procurement module made without prior notice to users, and frequent downtime at the start and end of the financial year.¹² Buyers should require vendor-managed upgrades with notice periods, open interfaces and a right to export their data, so that the system can be changed, or left, without a rebuild.

06Seven minimums for any system

The minimum standard is not a longer evaluation. It is seven gates, each with a test a board can ask to see evidence for and a failure that should stop the purchase (Exhibit 11). They are the author's proposals, anchored to the evidence above.

Three gates carry most of the weight.

Baseline (gate 5). Measure the current state before shortlisting: processes, volumes, users, spreadsheet workarounds and skills. Insist on a demonstration with the organisation's own data, not the vendor's. The IMF's guidance for public systems says the same in other words: identify the critical requirements for an initial version and defer the rest.⁹

Fit (gate 2). Gartner lists customisation tied too closely to existing processes among the causes of ERP failure.¹ The IMF warned that establishing a system "should not be viewed as merely computerizing existing procedures".⁹ The default should be the vendor's standard processes, with every customisation carrying a named sponsor and an upgrade-cost line.

Adoption (gate 7). Define success as use, not go-live. A system that is live but little used, as in Malawi's local governments, has passed every technical test and failed the only one that matters.¹⁵ One failed gate means stop.

07Change management decides uptake

Of all the evidence reviewed, change management has the strongest quantified link to outcomes. Prosci, a change-management firm, surveyed more than 2,600 practitioners. Of projects with excellent change management, 88% met or exceeded their objectives, against 73% for good, 39% for fair and 13% for poor (Exhibit 12). Excellent change management made a project about seven times as likely to meet its objectives as poor change management.³ The source has an interest in the answer, and the figures are practitioners' ratings across all project types, not ERP alone. They nonetheless point the same way as independent findings.

Gartner names limited organisational change management among the causes of ERP failure.¹ The IMF makes change management a condition of success and cites Malawi's earlier project, whose team was disbanded before completion, as a case of neglecting it.⁹ Malawi's local-government system, eight years on, is the end state: modules unused, users untrained, passwords shared.¹⁵

The practical consequence is a budget rule. Training, usability testing and change support are easy to cut when a project overruns. They should be protected, and funded before features.

08Six indicators that show whether it is working

Boards need leading indicators, not a go-live date. Exhibit 13 proposes six, each tied to a failure pattern in this paper. Active users against licensed seats catches the Malawi outcome early. The share of transactions processed in the system rather than in spreadsheets shows whether the system has replaced the old way of working. Time to make a change tests the lifeline. Cost per active user tests the economics in Exhibit 9. Month-end close time and benefits realised against plan show whether any of it is paying off. The thresholds are proposals, not findings from a study, and should be calibrated to the organisation's own baseline.

09What boards and funders should do

1. Ask for the baseline before the shortlist. No shortlist without a measured account of current processes, volumes, users and skills.
2. Ask which tier, and why. The sponsor should show that the tier matches today's headcount plus a three-year plan, with a path up and down.
3. Require five-year cost per active user. Include support and partner fees, and secure an exit right.
4. Fund adoption before features. Protect training, usability testing and change support in the budget.
5. Agree stop criteria and indicators before signing. One failed gate means stop.

Tier, not brand, is the variable. Buy the smallest system that passes seven gates, and fund adoption before features.

—References

1. L. Clark, "ERP carnage continues as orgs jump in unprepared", The Register, 13 November 2025, reporting Dixie John, senior director of ERP strategy, Gartner, at Gartner Symposium, Barcelona.
2. Computerworld, "A new study plays a familiar tune about delays and cost overruns" (also published as "Widespread discontent persists with ERP projects"), 17 March 2010, reporting Panorama Consulting Group's survey of about 1,600 organisations.
3. Prosci, "The Correlation Between Change Management and Project Success" (summarising the Best Practices in Change Management study, more than 2,600 practitioners), prosci.com, accessed 27 October 2026.
4. Business Day (South Africa), "Treasury splurged R2.4bn on inactive financial system, MPs told", 10 May 2023.
5. D. Torii, "What IT Leaders Must Do to Avoid Disappointing ERP Initiatives", Gartner research, gartner.com, accessed 27 October 2026. Prediction: by 2027, more than 70% of recently implemented ERP initiatives will not fully achieve their original business goals.
6. M. Bloch, S. Blumberg and J. Laartz, "Delivering large-scale IT projects on time, on budget, and on value", McKinsey & Company, with the BT Centre for Major Programme Management, University of Oxford, 1 October 2012.
7. Network World, "ERP costs: 3 signs companies are wasting less money", reporting Panorama Consulting Group's 2010 ERP Report: average cost US$6.2m, 54% of implementations over budget, average duration 18.4 months.
8. A. Hawari and R. Heeks, "Explaining ERP Failure in Developing Countries: A Jordanian Case Study", University of Manchester, Centre for Development Informatics, Working Paper 45, 2010.
9. J. Diamond and P. Khemani, "Introducing Financial Management Information Systems in Developing Countries", IMF Working Paper WP/05/196, International Monetary Fund, October 2005.
10. C. Dener, J. Watkins and W. Dorotinsky, "Financial Management Information Systems: 25 Years of World Bank Experience on What Works and What Doesn't", World Bank, 2011 (openknowledge.worldbank.org/handle/10986/2297).
11. Eastleigh Voice, "Auditor General Nancy Gathungu exposes IFMIS failures, Sh10.2 billion unapproved government spending", reporting the Auditor-General's 2023/24 report on national government, eastleighvoice.co.ke.
12. Kenyans.co.ke, "Auditor General Discloses IFMIS' Three-Week Breakdown", 5 April 2024.
13. Business Daily (Kenya), "Counties at a standstill as IFMIS disrupts operations", April 2017, reporting a statement by the Council of Governors.
14. M. Andrews and N. Cole, "Malawi's Cashgate Scandal", Case Study C-012P, Harvard Kennedy School (case note written for the 2017 executive programme "PFM in a Changing World"; online edition August 2025).
15. E. G. Laizer and R. Suomi, "Evaluation of Integrated Financial Management Information System (IFMIS) in Malawi", University of Turku repository record (data collected in 2015 at 15 local-government sites).
16. Sunday Times (South Africa), "Treasury will respect court decisions on IFMS tender", 5 November 2025.
17. L. Clark, "Stumbling ERP project racks up sales losses greater than budget", The Register, 3 January 2025.
18. E. G. Laizer and R. Suomi, "Evaluation of Integrated Financial Management Information System (IFMIS) in Tanzania", University of Turku repository record (data collected in 2012).
19. Parliamentary Monitoring Group, "IFMS: SIU investigations & National Treasury progress", Standing Committee on Public Accounts, 13 September 2023.
20. Microsoft, Dynamics 365 Business Central pricing page, microsoft.com, accessed 27 October 2026.
21. Microsoft, Dynamics 365 Finance pricing page, microsoft.com, accessed 27 October 2026.
22. ERP Research, "NetSuite pricing", erpresearch.com, accessed 27 October 2026. Third-party estimate; NetSuite does not publish a price list.
23. ERP Research, "Oracle Fusion Cloud ERP Pricing 2026: Module Price List", erpresearch.com, accessed 27 October 2026. Third-party estimate; Oracle does not publish list prices.
24. SAP News Africa, "SAP research reveals top tech skills challenges for African organisations", March 2023 (survey in the fourth quarter of 2022: Kenya, Nigeria, South Africa).