Bill Newns CPEng · NovoConsult · 2026
The views expressed in this paper are the personal views of the author. They are intended to contribute to industry discussion and should be read as such.
The informed client in underground construction
Procurement practice is built around the assumption that the client understands what they are buying. In most infrastructure sectors that assumption is reasonable. In underground construction, the gap between procurement complexity and client experience may be wider than in most other sectors — and experience is that the consequences of that gap can be unforgiving.
This is not a criticism of clients. It is a description of a condition. Tunnels, caverns, shafts and underground structures are procured infrequently, often by organisations for whom they represent a once-in-a-generation commitment. The industry that builds them has decades of accumulated experience with the ground, the risk profile, and the failure modes. The client organisation often does not — or has it in pockets, unevenly distributed, and not always present at the moment the most consequential decisions are made. The procurement machinery begins — and those decisions get made — before that gap has been closed.
This paper is addressed to the client: the funder, the project sponsor, the infrastructure agency, the organisation accountable for the outcome. It describes what procurement practice expects of you, what it will not volunteer, and what questions you need to be able to answer before the procurement begins — not after.
Three types of client
The risk is not the same for all of them
The institutional client
The rail authority, water utility, road agency. Has done projects before. Has some internal capability. May have lost institutional memory through restructuring or the passage of time between major underground programmes.
Risk: misplaced confidence. The market, the contract forms, and the technology have not stayed still.
The occasional client
The council, the port authority, the energy company. Underground construction is not their core business. Procuring something outside their core experience, with money they’re accountable for, into a risk environment where they may have limited independent technical capacity.
Risk: dependence on advisors whose interests may not be fully aligned with theirs.
The first-time client
The developer, the government agency standing up a new infrastructure programme. Limited or no institutional memory. The project team and its consultants are the primary internal source of knowledge about what is normal and what is not — and the client may not yet know what independent advice looks like or where to find it.
Risk: not knowing what they don’t know, and no internal reference point to test advice against.
Each of these clients faces a different version of the same problem: accountability for an outcome in which the technical complexity, the project uncertainty, and the commercial risk sit primarily with parties whose interests are not identical to theirs. Some try to bridge that gap. Some do what the last team did, or what the company culture requires.
The prior question
Decisions that precede procurement determine what procurement can achieve
Among the most consequential risk management decisions in an underground construction project are those made before the tender documents are issued. The alignment sets the risk profile. The procurement model determines who bears it and how. The investigation budget determines how well it is understood. The programme determines the quality of decision-making across the project — on scope, on impacts, on interfaces, on the business case itself. By the time a contract is awarded all of those decisions are fixed — and the exposure that will sit on the balance sheets of the client, the contractor, and their insurers has already been determined. What determines the quality of that exposure is due diligence and resourcing: whether the project was adequately investigated, whether the right people were in the room, whether there was sufficient time and budget to do the work properly. A variation claim before a shovel hits the ground, or two years into construction, is not the moment to discover that those questions were not asked.
Projects do not go wrong at construction. They start wrong.
The business case deserves particular scrutiny here. Flyvbjerg’s research across major infrastructure projects documents cost estimates for underground construction as among the most consistently unreliable in the sector. The reasons are not uniform. Some bias is unconscious: estimators anchor on comparable projects, underweight project-specific risks, and are subject to the same optimism that afflicts complex forecasting in any field. Some is structural: an estimate that is too high does not get approved. And some is deliberate — what Flyvbjerg calls strategic misrepresentation, where the number is shaped to secure approval rather than to reflect reality. The result is a number that carries an engineering letterhead but reflects a political decision about what the project needs to cost to proceed. The optimism is structural, not incidental. And sometimes it is not optimism at all — it is a number that was always wrong, submitted by people who knew it was wrong, because that is how projects get approved.
Underground construction’s most consistent cost and programme failures are not in the excavation. They are in what comes after it: the systems, the fit-out, the interfaces — track, power, signalling, controls, all of it required to function reliably together in an environment that did not previously exist. That complexity is rarely adequately scoped at business case stage. The estimate reflects the tunnel. It does not reflect what has to work inside it. This is examined further in Question seven below.
What the client should do
Before the procurement begins
The ITA/IMIA Code of Practice for Risk Management of Tunnel Works is not a regulatory instrument. It describes best practice — and in doing so, it describes what effective project risk management looks like. The Code requires that clients have sufficient technical and management competence available at all project stages, that they provide for design assurance and independent construction supervision, and that they foster a risk-aware culture evident in the collaborative behaviours of all parties. These are not health and safety obligations in the regulatory sense. They are the conditions under which projects succeed.
The significance of this framing is practical. A client who manages safety well — who maintains visibility of how work is designed and carried out, who ensures independent supervision, who requires risk registers to be live documents rather than filing exercises — is a client who is managing their project well. Safety performance is a lead indicator of project management quality, not a separate compliance obligation. The converse is equally instructive.
Specifically, before procurement begins, a client should be able to answer the following questions — or at least understand the implications of not being able to.
Question one
How does the client retain visibility and control of health and safety obligations?
Many decisions to eliminate risks can only be made early in the project lifecycle — before a contractor is appointed, before methods are fixed. Once made, they may commit the project to a set of risks that subsequent parties can only minimise. A procurement model should retain the client’s visibility of how work is designed and carried out. Visibility does not transfer duty but it should improve the practice of health and safety risk management. The duty to manage risk so far as reasonably practicable has specific meaning: a control measure meets the standard if its costs in time, money and effort are not grossly disproportionate to the benefit achieved. The test works in both directions — it asks what was done to manage the risk, and why no more was done. A client who cannot answer the second question has not discharged the first. The duty framework described here reflects modern health and safety legislation in jurisdictions including Australia, New Zealand, and the UK. The specific legislative structure varies, but the underlying principle — that influence over design carries accountability for its consequences — is broadly consistent across comparable frameworks.
Question two
What are the ground conditions along the alignment, and what are the key uncertainties?
If the answer is “the contractor will investigate as part of the design and construct contract”, the ground risk has not necessarily been managed. It has been passed to parties competing on their willingness to absorb uncertainty. The contractor who prices that uncertainty rigorously may lose to one who does not — the winner’s curse applied to procurement. Due diligence becomes a competitive disadvantage. The result is not risk transfer. It is competitive gambling, dressed in contract language, with either the public or a failed contractor funding the outcome.
Question three
What procurement model is appropriate for the project conditions and risk profile?
The choice of procurement model is a risk management decision. Design and construct can be appropriate where scope is well defined, complexity is well understood, and the contractor genuinely has the design capability to manage the risk. It is deeply inappropriate where those conditions do not hold. The deeper issue is design control. Who controls the detailed design may control the quality of the outcome — the technical decisions that determine not just whether the project is delivered, but whether what is delivered performs as intended over its design life. Quality, durability, maintainability, and operability are all shaped by design choices made during construction. Under a design and construct model those choices belong to the contractor, subject to the reference design specification and associated performance requirements. Where those requirements are well defined and rigorously enforced, the client retains meaningful influence over the outcome. Where they are not — because the reference design was prepared quickly, or because the performance requirements are ambiguous — the contractor has both the commercial obligation to their stakeholders and the contractual latitude to make design choices the client may not have intended. That is not bad faith. It is the predictable consequence of an incomplete specification.
Question four
Is the programme realistic — for design development, investigation, procurement, construction, systems integration, and commissioning?
Programme compression is one of the most consistent contributors to poor project outcomes, and one of the least visible in the project approvals process. The civil construction programme is usually the most visible element — it has a sequence of structures, a handover date. What comes after it is less visible and frequently less well planned: systems integration, fit-out, testing, commissioning, and the validation required before an asset can operate. The Elizabeth Line is the most extensively documented recent example — civil completion and operational opening were separated by years, and the reasons were systemic rather than incidental. The pattern is consistent across major underground transit projects. Commissioning in particular is routinely underestimated — it requires the whole system to work together, in an environment that has never previously existed, to a standard that must be demonstrated before passengers or users are admitted. A programme that does not honestly account for that complexity is not a programme. It is a political commitment dressed as a schedule.
Question five
What is the project’s cost range — across civil works, systems, fit-out, interfaces, and programme — and what does the contingency actually cover?
Most projects are presented with a cost range. The difficulty is that ranges are routinely compressed in public communication — by proponents seeking approval, by opponents seeking to alarm. By the time construction starts, the contingency is often already under pressure and the lower bound has become the budget. The ITA Code of Practice recommends that cost and time estimates be tested against the risk profile through sensitivity analysis. A contingency that has not been tested against the risk register is not a contingency. It is a number.
Question six
Who in the client organisation has the technical competence to evaluate what is being proposed — and to recognise when something is wrong?
This is the informed client question. Competence can be procured. But it must be procured before the tender, not during it — and the advisors must be independent of the parties whose proposals they are evaluating. The market for independent tunnelling advice is not large. The same scrutiny applied to selecting a contractor should be applied to selecting an advisor — including verification of references from previous clients on comparable projects.
Question seven
What are the scope, systems, and interface requirements beyond the civil works, and have they been adequately defined and costed?
The civil structure is the delivery vehicle. What has to work inside it — track, power, signalling, controls, operational systems, interfaces with existing infrastructure — is where major underground projects have repeatedly lost time and money. Flyvbjerg’s research on infrastructure cost overrun documents the pattern at scale; the Elizabeth Line and the Channel Tunnel’s fixed equipment programme are among the most thoroughly examined cases. The civil works and the systems are rarely procured or managed as a unified scope. When the interface between them is inadequately defined before procurement, the consequences emerge during commissioning — at the point of maximum pressure and minimum flexibility.
The safety duty
What procurement cannot transfer
Modern health and safety legislation in most jurisdictions places primary duties on the party with the greatest influence over how work is designed and carried out. For underground construction, that party is frequently the client — particularly before the contractor arrives, when the alignment, procurement model, and risk allocation have already been determined.
The practical implication is significant. A client who selects a design and construct procurement model transfers primary control of the design and construction process to the contractor. But in doing so they do not transfer the statutory duty to ensure, so far as reasonably practicable, that the work can be carried out safely. That duty remains with the client. It attaches to the decisions the client made — about alignment, about procurement model, about investigation adequacy — before the contractor was appointed.
The informed client asks: does my procurement model give me the visibility and control I need to discharge my health and safety obligations? If the answer is no, something needs to change — whether in the procurement model itself, in the oversight arrangements, or in the contractual provisions for independent supervision.
Ground risk
Why it never goes away
Ground risk is among the defining characteristics of underground construction. It is not one risk among many in the way it is in surface construction — it is the risk profile that makes tunnelling different from surface construction. And it has a property that the standard forms of contract in common use do not adequately reflect: it cannot be eliminated. It can be investigated, characterised, modelled, mitigated, and allocated. It cannot be made to go away by writing a contract that says the contractor accepts it.
The evidence for this is consistent, even if the industry’s failure data remains fragmented. Contracts that purport to transfer all ground risk to the contractor are frequently associated with disputes, delays, and cost overruns — a pattern visible in the published case literature and in the experience of practitioners, even without a unified activity dataset against which to measure frequency. The contractor prices the uncertainty — or cannot price it and hopes for the best. Someone pays either way: in the contract price, in the variation claims, or in the litigation that follows. Full ground risk transfer is not a risk management strategy. It is a deferral of the conversation about whose money it is.
The Geotechnical Baseline Report is the most effective instrument the industry has developed for honest ground risk allocation. Its purpose is not to protect the client from the contractor or the contractor from the client. Its purpose is to establish what is known, what is uncertain, and who carries what — before construction begins, when there is still time to make rational decisions.
The client who invests in understanding the ground, the scope, and the systems before procurement is not spending money on preparation. They are reducing the cost of everything that follows.
Caveat emptor. Caveat venditor.
The obligation that runs both ways
The procurement competition is where price is established. The contract is where collaboration begins. A client who wins the price and then manages the contract as an adversarial exercise — whether by withholding information, restricting access, or treating every variation as a claim to be resisted — has not protected their position. They have created the conditions for the outcome they were trying to avoid.
The disclaimer arrangements — “issued for information only”, “contractor to satisfy himself”, no warranty as to accuracy — are attempts to preserve the default common law position: you bid it, you build it. The difficulty is that this principle assumes the contractor had a fair basis to price what they were bidding. Where the client held material investigation data and structured the contract to prevent reliance on it, the risk allocation is inequitable in principle and unreliable in practice. Courts have shown a consistent willingness to look behind the contract language in these circumstances.
The practical point is simpler than the legal one. The procurement decisions that created the adversarial conditions were made years earlier, often by people who have since moved on. By the time the dispute is live, both parties have damaged assets, blown budgets, and legal costs that are rarely fully recovered.
The informed client’s interest and the contractor’s interest are aligned once the contract is placed. Their failure is your failure. The obligation at that point is not to extract the maximum from the contract. It is to help the contractor achieve their commercial goals within the framework that has been established — because that framework, if it was well designed, is the mechanism through which the project succeeds. Collaboration is not generosity. It is rational self-interest, correctly understood.
What good looks like
The attributes of an informed client organisation
The ITA Code of Practice identifies the client as the most important factor in project culture and risk management. This is not flattery. It is a structural observation: the client sets the conditions within which everyone else operates. A client who is technically informed, commercially realistic, and committed to collaborative risk management creates the conditions for a good project. A client who is not creates conditions that are harder for everyone to manage — and more expensive for the public to fund.
The attributes of an informed client are not mysterious. They include: a clear and realistic understanding of the project’s full risk profile before procurement — scope, systems, interfaces, and ground conditions where relevant; investigation and definition work proportionate to that risk profile; a procurement model chosen for the project conditions rather than for administrative convenience; independent technical advice procured at arm’s length from the parties whose proposals it evaluates; a contract that allocates risk explicitly rather than through silence and inference; and the organisational competence to recognise and respond to problems during construction.
Most of these attributes can be procured. Technical competence, independent review, GBR preparation, risk assessment — none of these require the client to have a tunnelling department. They require the client to understand that the investment is necessary and to make it early enough to be useful. The decisions that most need technical support are made before the tender is issued.
Institutional memory deserves specific mention. The gap between major underground projects in any single organisation can be a decade or more. The people who managed the last project have moved on. The lessons learned — if they were captured at all — are in documents that nobody reads. The procurement of institutional memory, through independent advisors with experience of comparable projects, is one of the most valuable investments an occasional or first-time client can make.
What the record shows
And what it asks of you
The tunnelling industry has the technical knowledge to build extraordinary things in extraordinarily difficult ground. The failure record — which is probably small as a proportion of projects built, but which cannot be demonstrated to be so because no activity dataset exists against which to measure it — is not primarily a record of technical failure. It is a record of system failure: procurement models mismatched to project conditions, inadequate investigation, fragmented responsibility, adversarial risk transfer, and the absence of an informed client and the right questions asked before the project started.
The informed client is not a luxury. In underground construction, they are the most important risk control the project has — and the one most consistently absent from the risk register.
But none of this is new. The diagnosis has been available for decades. What this article attempts is not discovery but reframing — to place the informed client at the centre of the analysis, where the evidence suggests they belong.
References
- Flyvbjerg, B., Bruzelius, N. and Rothengatter, W. (2003).Megaprojects and Risk: An Anatomy of Ambition.Cambridge University Press, Cambridge.
- Flyvbjerg, B., Holm, M.S. and Buhl, S. (2002). Underestimating costs in public works projects: error or lie?Journal of the American Planning Association,68(3), pp. 279–295.
- International Tunnelling and Underground Space Association (ITA-AITES) and IMIA (2023).Code of Practice for Risk Management of Tunnel Works.3rd ed. ITA-AITES, Geneva.
- ITA Working Group 5 (2020).Client Organisation Recommendations for Underground Construction Health and Safety.ITA-AITES, Geneva.
- Newns, W.R. (ed.) (2026).A Guide for Geotechnical Baseline Reports in New Zealand.2nd ed. New Zealand Tunnelling Society, Auckland.
- Essex, R.J. (ed.) (2022).Geotechnical Baseline Reports for Construction.2nd ed. American Society of Civil Engineers, Reston VA.
- National Audit Office (2019).Completing Crossrail.HC 2106, Session 2017–19. National Audit Office, London.
- National Audit Office (2021).Crossrail — a progress update.HC 299, Session 2021–22. National Audit Office, London.
- Mitsui Construction Co Ltd v Attorney General of Hong Kong (1986) 33 BLR 1 (PC).
- ISO 31000:2019 — Risk Management — Guidelines. International Organization for Standardization, Geneva.
He aha te mea nui o te ao? He tāngata, he tāngata, he tāngata.
What is the greatest thing in the world? It is people, it is people, it is people.