Assumptions and exclusions are the part of a statement of work nobody reads until something has gone wrong, at which point they are the only part anybody reads.
We looked at what actually happened when professional services scope disputes reached court, and at how experienced drafting bodies structure risk. The pattern is not the one most firms follow.
The pattern that survives
Clarity helps, but clarity alone will not rescue a provision that fails a statutory reasonableness test or sits awkwardly against the rest of the bargain.
England and Wales: precision helps, the whole bargain still matters
Start with Persimmon Homes Ltd v Ove Arup & Partners Ltd. The consulting engineer's appointment contained a remarkably short sentence: "Liability for any claim in relation to asbestos is excluded." The Court of Appeal upheld that exclusion against claims that Arup had negligently failed to identify asbestos.
One sentence, upheld on appeal, against a sophisticated counterparty. That is the strongest available evidence that specificity beats volume.
It is not unlimited licence. In The Trustees of Ampleforth Abbey Trust v Turner & Townsend Project Management Ltd, a clearly incorporated professional services liability cap still failed the reasonableness test under the Unfair Contract Terms Act, in part because it sat uneasily with the contract's insurance allocation. A clause is assessed as part of the whole arrangement, not in isolation.
US law: the negotiation record can matter as much as the clause
There is no useful single "US rule" here. State statutes and public policy doctrine differ, and the practical lesson from California is about process rather than wording.
In Markborough California, Inc. v Superior Court, an engineering consultant used a standard-form limitation of liability. The parties had never actually discussed the provision. The court held that, for the statute at issue, what was required was a fair opportunity to accept, reject or modify the provision. Actual discussion was not necessary.
That cuts both ways for a drafter. You do not have to walk a client through every limitation. You do need to be able to show they had a genuine opportunity to negotiate it, which is an argument about how the document was presented rather than what it said.
Assumptions as change events, not footnotes
The best structural answer we found comes from the UK government's sourcing model. The Cabinet Office's Risk Allocation and Pricing Approaches guidance, updated 15 June 2026, describes allowable assumptions for use where full due diligence before signature is impractical.
Rather than forcing the supplier to price an unknown or bury it in a disclaimer, the assumption is recorded explicitly. If it proves inaccurate and has an agreed cost effect, the supplier can propose a contract change through the normal controls.
An assumption without a stated consequence is a disclaimer. An assumption with a stated consequence is a mechanism.
That is the single most transferable idea in this research, and it is free to adopt. Every assumption in your SOW should answer: if this turns out to be false, what changes, and how does that change get agreed.
Longer lists are not shown to reduce disputes
We found no credible empirical study measuring professional services dispute frequency against the number or length of assumptions and exclusions in the underlying SOW. That absence is worth knowing, because the industry behaves as though the relationship is established.
Court records tell you what happened after particular wording failed. Model contract bodies tell you how experienced drafters structure risk. Neither establishes that fifteen exclusions produce fewer escalations than five.
The honest counterweight is that courts do sometimes enforce very broad protective wording. The Fort Knox record is uncomfortable for any thesis that short and specific always beats broad and exhaustive: an extensive clause reached multiple theories of liability and survived appellate challenge. We would still draft for specificity, and we would not claim the evidence forces it.
What to actually write
Translating the case law into drafting produces four moves, and none of them requires lengthening the document.
State the deliverable as an outcome, not an activity. "Interviews, workshops and a playbook" describes what you will do. It does not say what will exist afterwards, which means it cannot be tested for completion, which is how scope arguments start. This is the same failure that turns scope sections into service lists.
List client inputs as dependencies with owners and dates. "Subject to timely client input" is not an assumption, it is a wish. "Access to the CRM extract by day five, owned by the RevOps lead" is a dependency somebody can miss, which is precisely what makes it useful when they do.
Write exclusions narrowly and specifically. The Persimmon sentence is eleven words and covers one named hazard. A reader can tell instantly whether a given claim falls inside it. Compare that with a paragraph excluding "any consequential, indirect or related losses howsoever arising", which invites an argument about every word.
Attach a consequence to every assumption. This is the move most documents skip. An assumption that says what changes if it proves false, and how that change gets agreed, converts a future dispute into a scheduled conversation.
Why the commercial reader matters as much as the lawyer
There is a real tension between language that survives litigation and a document a client will sign without escalating to counsel they had not budgeted for.
Broader contract scholarship is relevant here. Work by Choi, Gulati and Scott on contract boilerplate describes how standardised protective language accumulates over time, drifts from the specific bargain it was written for, and can end up understood by nobody who signs it. A clause nobody reads is a clause nobody negotiated, which is exactly the exposure the Markborough analysis identifies.
The practical resolution is that specificity serves both readers. A short, concrete exclusion is easier for a commercial buyer to accept because they can see its edges, and it performed well in the case that matters most. An exhaustive schedule is harder to sign, harder to explain, and on the available evidence not demonstrably better at preventing disputes.
Common questions
Are jurisdiction differences really that significant?
Yes, and this is the one place we would refuse to generalise. The England and Wales reasonableness analysis and the California opportunity-to-negotiate analysis are different tests producing different drafting advice. Anything presented as universal here should be treated with suspicion.
Will a client sign something with an explicit change mechanism?
In our experience it is easier to sign than an ambiguous scope, because it converts an unbounded risk into a defined process. The allowable-assumptions model exists precisely because sophisticated buyers prefer a known mechanism to a priced-in contingency.
Should exclusions live in the SOW or the master agreement?
Liability and IP terms belong in the master agreement. Scope-specific assumptions belong in the SOW, because they describe this engagement and should change when the engagement does.
How specific is specific enough?
Specific enough that a reader can tell whether a given future event falls inside or outside it without asking you. "Liability for any claim in relation to asbestos is excluded" passes that test in eleven words.