The ERP Clause Nobody Reads Decides the Ten-Year Cost
An ERP system's ten-year cost is set by five contract clauses, not the software you pick. The Diageo case, the EU Data Act and the pre-signature checklist.
In 2017, a beverage group lost a court ruling over a single word in a contract signed in 2004. The word was “access”. Diageo had connected roughly 5,800 of its own staff to SAP through a Salesforce front end, assuming those people were Salesforce users. SAP read the old licence differently: anyone whose work touched the SAP system was a named SAP user, billable as such. The High Court in London agreed with SAP’s reading of the wording. The claim came to 54 million pounds (SAP UK Ltd v Diageo Great Britain Ltd, [2017] EWHC 189 (TCC), 16 February 2017).
Nobody at Diageo had treated that clause as a financial decision. It sat in the contract for thirteen years, unread, until a judge interpreted it. That is the part most ERP buyers get wrong. They argue for weeks over feature lists and licence list prices, then hand the actual contract to procurement to “tidy up the terms”. The terms are where the ten-year cost lives.
The selection is the cheap decision
A mid-market buyer in the German-speaking market typically runs a structured ERP selection: requirements workshops, vendor demos, a scoring matrix, a shortlist. All of that effort goes into picking the software. Very little goes into reading the contract that governs how that software may be used, maintained, exited and audited for the next decade.
That is backwards. Two systems with near-identical feature sets can carry wildly different total cost over ten years, and the difference rarely sits in the licence price. It sits in five clauses: how “use” is defined, whether indirect use becomes billable, how the annual maintenance fee may rise, what you get back when you leave, and on what terms the vendor may audit you.
I have read contracts where the “cheaper” of two offers, eight percent lower on the headline price, was the more expensive one over the contract life, because it left the definition of “use” open and capped nothing on the maintenance increase. The eight percent saving was real on day one and gone by year three.
Clause one: how “use” is defined
The Diageo ruling is the textbook case of indirect use, and it is worth understanding precisely because it is not exotic. Indirect use happens whenever a third system (a CRM, a web shop, a reporting tool, increasingly an AI agent) reads or writes ERP data without its end users logging into the ERP directly. If the contract defines “use” loosely, the vendor can later treat every one of those touchpoints as a licensable user.
SAP’s later Digital Access model was a direct response to the Diageo problem: it tried to price indirect use by document rather than by user. That helps, but it is a vendor’s framing of a vendor’s risk. The buyer’s job is to define, in the contract, exactly what counts as use and what does not, separating named human users from machine and integration access, before signing rather than after a back-bill arrives. German copyright law gives a second floor here: section 69d of the Urheberrechtsgesetz protects use that is necessary for the software’s intended purpose, even where a licence clause says otherwise.
Clause two: how maintenance is allowed to rise
The Diageo number is dramatic, but the quieter clause is the one that drains money every single year: the maintenance and support fee, and the rule for how it escalates.
The DSAG Investment Report 2026 (26 February 2026), from the German-speaking SAP user group whose annual report tracks how its members actually spend, found that around 37 percent of SAP ECC users plan to move to S/4HANA by the end of 2027, with roughly half deferring until 2030. Those who defer pay extended-maintenance surcharges. The delay itself is a defensible business choice. The problem is the contract clause that lets the surcharge climb at the vendor’s discretion, signed years earlier by someone who never modelled what an open-ended escalation does over a decade.
Trovarit, an independent Aachen-based analyst house whose “ERP in Practice 2024/25” study surveys more than 1,700 users across vendors, found the same weakness from the user side. Software itself scored 1.80 on the German school grade scale, while vendor service and support scored only 1.96, the study’s clear pain point. Maintenance is consistently the weakest part of the relationship in users’ own judgement, and it is the part the contract regulates least.
Clause three: getting your data out
For a long time, the standard answer to “what happens when we leave?” was a shrug. Exit and data-handover terms were treated as non-negotiable boilerplate, especially in cloud contracts.
That changed on 12 September 2025, when the operative provisions of the EU Data Act (Regulation (EU) 2023/2854, Articles 23 to 31) became applicable. The Data Act forces cloud providers to support switching and to hand data back in a usable form. From 12 January 2027, switching charges are prohibited in principle. The European Commission published a draft of standard contractual clauses for switching, exit and termination on 19 November 2025, which gives buyers concrete model wording to negotiate against.
The law sets a floor; it does not write your contract for you. The format of the data export, the timeline, the obligations during a transition — those still have to be drafted, and the gap between “we will return your data” and “we will return your data in this format, within this period, at no charge” is exactly where the long-term cost hides. There is a related lesson in why mid-market companies are bringing their ERP back in-house: the exit you never specified is the one that traps you.
The legal ground has shifted
There is a reason this matters now rather than five years ago. Three things moved at once. The Data Act became applicable in September 2025, putting switching rights on a statutory footing. The SAP maintenance and transformation pressure toward 2027 and 2030 is pushing a wave of contract renewals and migrations, documented in the DSAG Investment Report 2026. And the Trovarit service criticism from 2024/25 confirms that the maintenance clause is where users already feel the pain.
There is also a longer line of case law that buyers underestimate. The European Court of Justice in UsedSoft v Oracle (Case C-128/11, 3 July 2012) held that the right to distribute a software copy is exhausted once it is sold, even for downloaded software, and the German Federal Court of Justice confirmed this nationally in UsedSoft II (I ZR 129/08, 17 July 2013) and again in the Adobe ruling (I ZR 8/13, 11 December 2014). Licence terms are not boundless. They sit inside a legal frame, and in Germany that frame includes the fairness review of standard terms under section 307 of the Civil Code (BGB) — under which even a clause printed as “non-negotiable” can be void.
Key Takeaways
The reason this list exists: the ten-year cost of an ERP system is set by five contract clauses, not by which software you picked. Read these five before anyone signs.
| # | Clause to check | What to demand |
|---|---|---|
| 1 | Definition of “use” | Spell out named human users versus indirect and machine access. The Diageo gap was an undefined “access” term. |
| 2 | Maintenance escalation | Replace open vendor discretion with an index link or a fixed annual cap. |
| 3 | Exit and data handover | Specify export format, handover deadline and cost; benchmark against the Data Act minimum. |
| 4 | Audit rights | Limit frequency, notice period and scope; agree who bears the cost when an audit finds nothing. |
| 5 | Standard-terms review | Know that clauses can be void under section 307 BGB — a lever even against “non-negotiable” wording. |
Before and after, in one case: Diageo’s 2004 contract left “use” undefined and the vendor’s reading produced a 54-million-pound claim. A contract that had explicitly excluded indirect integration access from the named-user count would have removed the claim’s basis entirely. Same software, same integration, opposite outcome — decided by one sentence.
Where this does not apply: pure hyperscaler IaaS terms below the negotiation threshold, where you genuinely cannot move the wording. There the lever is not clause negotiation but the Data Act’s statutory switching and export rights — you invoke the law instead of redrafting the contract.
What the case law does not say
The strongest objection here does not come from the vendors. It comes from serious procurement people: “With standard SaaS, the terms aren’t negotiable anyway. Clause optimisation is a fantasy against hyperscaler boilerplate.”
That is true for the preamble and false for the substance. The Data Act now compels the switching and exit rights that used to be presented as fixed, and the UsedSoft line plus section 307 BGB show that licence terms, even pre-printed ones, are subject to review and can be struck down. So the picture is not “negotiate everything” versus “negotiate nothing”. It is a sorting job: which clauses are genuinely negotiable, which are now backed by statute, and which standard terms are legally vulnerable.
The honest limit of the argument: clause work does not save a bad system, and below a certain contract size the marginal hour of legal review is not worth it. For a small standard subscription, read clauses three and five and move on. But for a core ERP that will run a mid-market business for a decade, the contract is the larger commitment, and treating it as procurement’s afterthought is the expensive mistake. The point is not that clauses beat software. It is that the clause decides the cost the software cannot.
Frequently asked questions
Which contract clause costs us the most over ten years?
Usually the maintenance escalation and the definition of “use” — not the licence list price. SAP UK v Diageo (16 February 2017) shows how a single undefined “access” clause produced a 54-million-pound claim. The headline price was never the problem.
Are standard SaaS contracts negotiable at all, or do we just have to accept the terms?
Partly negotiable. The EU Data Act (applicable since 12 September 2025) compels switching and exit rights even in standard contracts, and from 12 January 2027 switching fees are prohibited in principle. On top of that, standard terms are subject to the fairness review under section 307 BGB, so even “non-negotiable” clauses can be void.
What is indirect use and why is it a contractual risk?
Indirect use occurs when a third system (a CRM, a shop, an AI agent) reads ERP data without its end users logging into the ERP directly. If “use” is defined loosely in the contract, the vendor can later make those touchpoints licensable. That is precisely the Diageo situation, and AI agents make it more common, not less.
How do we check before signing whether a contract holds up long term?
Have five places read before anyone signs: the definition of use, indirect use, maintenance escalation (is there a cap?), the exit and data-handover clause (format, deadline, benchmarked against the Data Act), and audit rights (frequency, scope, cost consequence). Picking the software is the second step, not the first.
We have already signed — is it too late?
Not necessarily. The Data Act partly reaches into running cloud contracts (the position on existing contracts is not yet fully settled), and section 307 BGB can void unfair clauses after the fact. The natural leverage point is the next renewal or migration, which is also when you should re-read all five clauses.
Next step
Are you about to sign an ERP contract you have only read for features?
Send me the draft and I will read it against the five clauses above — what “use” means, how maintenance escalates, what your exit actually returns, how the audit right is bounded, and where standard-terms review gives you a lever. No pitch, just a second pair of eyes before the signature.
→ Or read more first: IT strategy and system selection
Sources and links: SAP UK v Diageo, [2017] EWHC 189 (TCC) · EU Data Act, Regulation (EU) 2023/2854 · DSAG Investment Report 2026 · Trovarit ERP in Practice 2024/25 · ECJ UsedSoft v Oracle, C-128/11 · Section 307 BGB
More on pfisterer.xyz: Cloud exit: bringing ERP back in-house · DSAG Investment Report 2026 · ERP migration: 5 mistakes that sink projects · DATEV ERP: never change cost centers mid-year