Why SAP Cloud ERP programmes fail
and most of the answers sit before the contract is signed.
Delivery controls execution: schedule, team, data quality, change management. What it inherits was decided before it existed, and cannot be revised from inside it. It inherits the route, and with it the volume of process redesign the organisation will have to absorb. It inherits the integrator, its seniority profile and its characteristic failure mode. It inherits the scope the signed order form froze. It inherits the vendor’s operating model and the obligations the contract distributed. A well-run programme on poor inheritances delivers exactly what those inheritances allow.
What delivery controls, and what it receives finished.
Execution
Cutover plan, data quality and reconciliation, user training, test rhythm, code remediation, process validation by the business owners.
This is where most executive attention concentrates, and where the programme is judged when it goes wrong.
Inheritance
Migration route, integrator and the failure mode of the profile chosen, scope frozen in the order form, provisioning milestones, the vendor’s operating model, the responsibility boundary between vendor and integrator.
Every item here had a legible signal before signature, and after it becomes a constraint, assessed with no alternative on the other side.
The deviation starts before delivery, in the phase where the route is chosen.
Published research measures the deviation of this class of programme. In Horváth's Business Transformation Unlocked study, covering 200 SAP-using companies with at least 200 million euros in annual revenue across the DACH region, selected Northern and Eastern European countries and the United States, with fieldwork in the first quarter of 2025, S/4HANA programmes run on average 30 per cent longer than planned, a quarter of organisations heavily exceeded budget and a further 40 per cent strongly exceeded it, and 65 per cent report strong to very strong quality deficits. The causes respondents themselves cite include expansion of project scope during the project and revision loops on concepts and processes, which is the form the redesign dispute takes once delivery is already under way.
The same study's authors place the origin before delivery. They write that the foundation for many of the problems is already laid in the programme setup, where inadequate planning and insufficient consideration of the transformation approach appropriate to the current situation leave the organisation overwhelmed, with complexity underestimated and organisational competence overestimated. From the customer side, ASUG's Pulse of the SAP Customer 2025 research, with 787 members surveyed in November 2024, places the move to S/4HANA as the most-cited number-one challenge at 26 per cent, against 10 per cent for budget. Neither study covers Latin America, and neither measures the outcome conditioned on the mismatch between route and organisation. What they establish is the class and the attribution.
What makes the cost avoidable is the order. The question about appetite for process redesign is answerable in a few executive conversations, and the answer is an input to the route decision. Run in reverse, the route is chosen first and the organisation discovers its own appetite during the workshops, with the contract already signed.
For organisations starting from ECC, the calendar pressure on this phase comes from dates SAP itself publishes. For the core applications of SAP Business Suite 7 (which includes SAP ERP 6.0, the ECC, and is the on-premise generation before S/4HANA), mainstream maintenance runs to the end of 2027, with optional extended maintenance at a premium of two percentage points on the maintenance basis through the end of 2030, both limited to the three most recent enhancement packages (EHP) of each application. The SAP ERP, private edition, transition option provides a continuity subscription between 2031 and 2033 and requires migration to SAP ERP, private edition on SAP HANA to be completed before the end of 2030. The effect of those dates on the programme is indirect: they shorten the phase in which the route can still be chosen.
Nine classes of signal, and what each one produces later.
The screen Alkemia runs at qualification covers nine categories. Below are the categories and the consequence each one predicts. The questions that produce the diagnosis, the weight of each finding and the escalation criteria are part of the method and stay off this page.
- A
The route, against the culture the organisation actually has
The migration approach is chosen in one conversation. Whether the organisation can execute it is a different question, and it is rarely asked.
An organisation used to preserving its processes, placed inside a delivery that requires redesigning them, enters dispute in the first workshops. The dispute shows up as calendar that slips and scope that returns to the table.
What it produces later: Published research attributes this deviation to the phase where the route is chosen. In the Horváth study Business Transformation Unlocked, covering 200 SAP-using companies with at least 200 million euros in annual revenue across the DACH region, selected Northern and Eastern European countries and the United States, with fieldwork in the first quarter of 2025, S/4HANA programmes run on average 30 per cent longer than planned, and the causes respondents themselves cite include expansion of project scope during the project and revision loops on concepts and processes.
- B
The integrator, and the basis on which it was chosen
After signature the integrator is what executes, and that is where cost and delivery risk concentrate. The choice is made on claims that arrive in a proposal.
Each integrator profile carries a strength and a failure mode of its own: the global integrator brings scale and then swaps the senior discovery team for a junior delivery team; the regional firm brings proximity and depends on one senior who can be pulled to a parallel deal; the boutique brings depth and becomes the bottleneck when the rollout grows. None is better, and choosing without naming the failure mode is choosing blind.
What it produces later: Integrator governance with no owner. When the answer to who decides to go live with open findings is the integrator that executes, the go-live decision has been outsourced to the party with an interest in it.
- C
The clock that starts before the environment exists
Signature date, subscription start date and provisioning date are three different dates, and the contract treats the first two as if they were one.
The invoice lands on the contractual date. The environment usually does not exist on that date. On complex deals provisioning stretches to 30 or 60 days or beyond, and the real clock only starts when an onboarding form is filed, which happens after signature.
What it produces later: The business case timeline slips by the duration of the delay, and that slip propagates to every deliverable anchored to it. The customer’s finance function books cost against a service it cannot yet use.
- D
The scope that entered as a placeholder
Legal entities declared as "all entities of the group", integration endpoints not yet inventoried, document volume estimated and not yet measured.
None of this looks like a problem at proposal stage. A placeholder draws revision, and the revision arrives later, once commitment already exists.
What it produces later: The bill of materials inflates between the preliminary version and the signed order form, and the difference surfaces as cost nobody approved, inside a budget that has already been defended internally.
- E
The operating model the customer assumes
The expectation of the vendor-side managed service, against the model that service actually runs.
An organisation expecting a daily call with the vendor’s delivery leadership, and receiving a ticket-based model with a published escalation path, discovers the difference at the first incident. And an IT function that does not separate the vendor’s run-phase obligations from the integrator’s functional-configuration obligations will chase the wrong party.
What it produces later: Relationship strain in the first ninety days after signature usually starts here, exactly the window in which the programme needs cooperation to move.
- F
The legal and procurement posture
When procurement enters the conversation, and whether counsel has seen a cloud contract before or only perpetual licence agreements.
They are two different instruments, with different logics of obligation, measurement and exit. A legal team fluent in perpetual licensing reviews a subscription contract looking for the wrong clauses.
What it produces later: The negotiation stalls near the end, under calendar pressure, or worse: it closes on time and produces post-signature obligations nobody can enforce when the moment comes.
- G
The licence metric, and what decides which one applies
The contract meters consumption on more than one metric, and what decides which applies to an integration is a technical detail the proposal does not discuss.
The rules are published by the vendor and are almost never read against the architecture the customer actually runs. The metric is decided by the identity that reaches the core and by the path through which the document is created, so two integrations delivering the same business outcome land on different metrics when those two elements differ. The proposal describes the outcome. The path sits in the architecture, which nobody attaches.
What it produces later: The bill arrives later, with the integration already in production, when changing the path costs more than choosing it would have. The estimation report from the vendor does not settle it: it counts records that exist today, while the licence meters documents created within a period, so the two error sources push in opposite directions.
- H
The technical exposure already sitting in the reports the customer owns
The reports that size the conversion come out of the customer system and exist before signature. Usually only the summary is read.
An aggregate loses information exactly where the specific exposure lives: a count of objects cannot carry which process, which table or which regulatory obligation is at risk. Reading only the summary inherits the blindness of whatever produced it, and presents that as coverage.
What it produces later: Sizing enters the contract calibrated on a number nobody opened. When the aggregate and the report text disagree, that discrepancy was the finding, and it surfaces with the scope already fixed.
- I
The boundary of who does what, and how much of it sits in the fee
Every operational task answers two independent questions: is it inside the fee, and who has the right to perform it.
The two answers do not travel together. There are tasks included and performed by the vendor, tasks only the customer may perform, extra-cost tasks only the vendor performs, and extra-cost tasks where the customer chooses who does them. Some exist only if written into the commercial document, and what is not written was not contracted.
What it produces later: The discovery happens at the first incident nobody is obliged to answer, or on the first invoice for a service the customer assumed was included. In both cases the answer was already in the signed contract.
A single signal rarely sinks a programme. What decides is their accumulation and their distribution across categories.
Three claims that do not survive the public record.
- 01
Certified, staffed and assigned are three different populations
The people the vendor certifies are one set. The people a proposal counts as consultants are a larger set, including uncertified and adjacent profiles. The people named in the customer’s statement of work are a much smaller set drawn from one of the two. The proposal quotes the largest of the three, and the customer hears the smallest.
- 02
Global competency, with zero engagements in the customer’s country
An integrator can hold genuine competency at the top of the vendor’s scale, with a substantial engagement count, and have zero engagements recorded in the customer’s own country. The capability exists and can be brought in, which turns the claim into four distinct questions: staffing, language, rate and continuity.
- 03
The signing entity can be younger than the track record presented
The entity that executes the contract may not be the entity whose capability was presented. Where it is younger than the history the proposal claims, that history belongs to a different legal person and does not travel with the contract. It is a check of minutes that changes a decision of years.
The proposal is the only record in the process written for the customer to read. Others exist, written for other purposes, and the divergence between them is the finding. It is systematic: it runs in one direction, and it runs largest on exactly the claims a customer cannot check without help.
A well-run programme on poor inheritances delivers exactly what those inheritances allow.
Alexandre Pereira
Read the signals while they are still reversible.
For the executive sponsor who answers for the programme, the CFO who answers for the multi-year commitment and the CIO who answers for the architecture, the operational question is: of the nine categories, how many already show a signal in the material on the table today. A 30-minute conversation, with no obligation, is enough to identify which categories deserve verification and which depend on evidence that has not yet been requested. If the conclusion is that the programme is well positioned, that is a useful answer too. The best time to do this is before signature. The second best is before go-live. The third is today.
Discuss your decisioncontact@alkemia.io
When the work needs to go beyond a conversation, the entry point is the pre-signature diagnostic.
The contract itself, and what in it fixes cost across the term, is in six decisions before signing.
The dates cited here are published vendor positions.
Programme failure, in decision terms.
- Why do ERP programmes fail?
- Failure surfaces in delivery and is usually explained by it: dirty data, insufficient training, weak change management. A large share of those causes already had a signal before signature. The migration route chosen against the culture the organisation has, the integrator selected on claims nobody verified, the scope that entered as a placeholder, the contractual clock that starts before the environment exists. Those signals are legible in the qualification phase, and that is the phase the schedule cuts first, because it is the only one that produces no visible deliverable.
- What does choosing the wrong route cost?
- No published study measures the outcome conditioned on this specific mismatch. What exists is the deviation profile of the whole class. In the Horváth study Business Transformation Unlocked, covering 200 SAP-using companies with at least 200 million euros in annual revenue across the DACH region, selected Northern and Eastern European countries and the United States, with fieldwork in the first quarter of 2025, S/4HANA programmes run on average 30 per cent longer than planned, a quarter of organisations heavily exceeded budget and a further 40 per cent strongly exceeded it, and 65 per cent report strong to very strong quality deficits. The study authors place the origin in the programme setup and name, among the causes, insufficient consideration of the transformation approach appropriate to the current situation of the organisation. Neither of these studies covers Latin America.
- How do you actually verify an implementation partner?
- The proposal is the only record in the process written for the customer to read. Other public records exist that were not written for that purpose, and the divergence between them is the finding. That divergence is systematic: it runs in one direction, and it runs largest on exactly the claims a customer cannot check unaided, such as certified-consultant counts per solution, in-country presence, and the identity of the entity that signs. Verification is a qualification activity, run against the shortlist, and the result changes both the choice and the protections the customer carries into the negotiation.
- Is this not the integrator’s or the vendor’s job?
- Both hold a legitimate interest in closing the deal, and neither is a neutral party in the choice of the other. The integrator does not assess its own selection, and the vendor does not arbitrate the delivery capability of the integrator that resells its product. The independent read exists because of that asymmetry, which remains even when everyone acts in good faith.
- We have already signed. Is it still worth looking at this?
- It is, and the question changes. Before signature the objective is to choose and to protect; afterwards it is to find what is already exposed and what still carries leverage. What remains splits in two. The operational part does not require reopening the contract: assigning an owner to the obligations the contract created and nobody took, installing integrator governance, tracking the provisioning milestones. Anything touching term, metric or price is corrected only by reopening the contract, and reopening has a price. Making that split early is what keeps more of the list on the cheap side.
- Why is the pre-signature phase always the one that gets cut?
- Because it produces no visible deliverable and competes for calendar with a date that has already been communicated. The time pressure has a real origin: mainstream maintenance for the core applications of SAP Business Suite 7 (which includes SAP ERP 6.0, the ECC, and is the on-premise generation before S/4HANA) runs to the end of 2027, with optional extended maintenance at a premium of two percentage points on the maintenance basis through the end of 2030, both limited to the three most recent enhancement packages of each application. The SAP ERP, private edition, transition option provides a continuity subscription between 2031 and 2033 and requires migration to SAP ERP, private edition on SAP HANA to be completed before the end of 2030. The effect of those dates on the programme is indirect: they shorten the phase in which the route can still be chosen.
The patterns described here come from advisory observation within a declared scope, and hold as hypotheses calibrated to that scope. Before acting, it is worth validating each category against the reality of your own programme, sector and region.
SAP and SAP product names are trademarks of SAP SE or its affiliates. Alkemia is an independent consultancy with no affiliation to, endorsement from, or representation of SAP, and is not an implementation partner: it does not bid for, sell, or deliver the implementation project. This material is informational, describes patterns in general terms, and constitutes neither legal advice nor an assessment of any specific vendor or integrator. The maintenance dates reflect positions published by SAP and consulted in August 2026.