Chapter 15: Experts and the tribunal

A delay claim usually has two experts on each side. The delay expert, sometimes called the programming expert, analyses the programme and its updates against the site records. The question is what delayed completion and by how long. The quantum expert, often a quantity surveyor or forensic accountant, works out what that delay cost. The two reports are linked. The quantum expert prices the periods the delay expert finds, so if the weeks fall, the money falls with them.

Why are experts needed at all? What delayed a job is a question of fact, but a technical one. The critical path of a project with thousands of activities cannot be seen by looking at a bar chart. It has to be calculated from the logic of the programme (chapter 2). A tribunal needs an expert to run the programme, test its logic and read it against the records. But the question stays factual: what probably happened on this job? So the answer stays with the tribunal. Akenhead J said so in Walter Lilly (TCC, 2012): it must be for the Court to decide as a matter of fact what delayed the Works and for how long.

Two practical rules follow from that division of labour. First, the parties set the questions. The TCC Guide says the parties must identify the issues each expert deals with (14.7.2). Second, an expert cannot prove primary facts by stating them. If part of a report is relied on to prove what happened on site, that part is treated as factual evidence and belongs in a witness statement served with the other factual evidence (14.7.4). Whether an as-built programme built from the records falls into that category is a judgment call, not a settled rule.

Honest experts reach different answers, usually because of method. The same records, run through different methods, can give different critical paths and different periods of delay (chapter 7). The figure runs six methods over the guide's model project, which is invented and illustrative: thirteen activities, planned finish week 34, contract completion week 36.

Does the tribunal have to pick one expert? No. The burden of proof decides what happens when neither analysis is reliable. The party claiming time or money must prove the delay it relies on (chapter 6). If its expert's analysis fails, the tribunal is not bound to build a better one for it. In Tata v DBS (TCC, 2024) Constable J found that neither of the contractor's expert's two analyses was reliable [212]. The evidence gave no reliable basis upon which the Court could effectively carry out its own as-planned v as-built exercise, and the claim failed on the burden of proof [222]. The Court of Appeal later heard a different point only (DBS v Tata [3]).

That result was tied to what the contract required. The contractor had to prove the date by which it would have achieved a milestone. Where entitlement does not depend on a date, the judge said the court might have been tempted to adjust the evidence towards a middle answer. So Tata is not a rule that tribunals never repair an expert's analysis. It is a warning that they need not. In Walter Lilly the judge found that both experts had in reality been doing the exercise the court must do: a factual analysis of what probably delayed the works [381].

The expert is paid by one side but owes the duty to the court. The reason is practical. A court needs expert evidence because it cannot do the technical analysis itself. An expert who argues the client's case gives the court nothing it can rely on, because the court cannot check the work. So the Civil Procedure Rules say it is the duty of experts to help the court on matters within their expertise. That duty overrides any obligation to the person who instructs or pays them (CPR r.35.3(2)). Every report ends with a statement that the expert understands the duty and has complied with it (r.35.10(2)).

Practice Direction 35 fills in the detail. Expert evidence should be the independent product of the expert uninfluenced by the pressures of litigation (PD 35 2.1). The expert should not act as an advocate (2.2), and should consider all material facts, including those that cut against the opinion (2.3). HHJ Toulmin summarised the classic statement of these duties, from The Ikarian Reefer (1993), in LFEPA v Halcrow Gilbert (TCC, 2007) [46]-[52].

Is there a conflict between the duty to the court and the duty to the client? No. A client needs to know how strong its case really is before it pays for a trial. An expert who tells it what it wants to hear leaves it to learn the truth from the judge. Lord Dyson said in the Supreme Court in Jones v Kaney that there is no conflict between the two duties [99]. By a majority, the same case ended the old rule that experts could not be sued by their own clients for breach of duty (Lord Phillips [62]).

What does a breach look like? In Van Oord v Allseas (TCC, 2015) a quantum expert repeated, as his own views, assertions taken from his client's witnesses. Coulson J called that the complete opposite of what a responsible, independent expert is obliged to do [89]. In ICI v Merit Merrell (TCC, 2018) Fraser J found that another quantum expert had taken a partisan and unsustainable position in his client's favour [198]. Both were quantum experts. The duty is the same for a delay expert.

The usual consequence is loss of weight, not exclusion. HHJ Toulmin said that only in the most extreme case will expert evidence be rejected out of hand; otherwise the failure is the background against which it is weighed (LFEPA [60]). In practice that is enough. A judge who doubts an expert's independence discounts the evidence, and with it the case that rests on it.

Conflicts matter too. An expert who has worked for a client learns its case and its weak points. If the same firm then works against it on a connected dispute, the client cannot be sure that knowledge stays put. So what matters is what the expert promised in the retainer. Secretariat v A Company (CA, 2021) concerned a group of delay and quantum experts. It acted for a client in one arbitration, then took an engagement for another party in a related arbitration, and the client objected. The Court of Appeal decided the case on the terms of the retainer, which contained a conflicts undertaking. Whether an expert owes a duty to avoid conflicts depends on the terms of the retainer [66]. It made no difference that one engagement was for delay and the other for quantum, because the two issues are closely connected [121]. Read the retainer, and run the conflict check across every firm it binds.

Choosing the method is the expert's job. Where there is a range of opinion, the report must summarise the range and give reasons for the expert's own view (PD 35 3.2(6)). An expert is not confined to one of the methods the SCL Protocol lists (Thomas Barnes [110]). But a manifestly inappropriate method, or an unexplained departure from the method the expert says he is using, goes to weight (same paragraph). Chapter 7 has the detail. The lawyers should understand the choice and test it. They should not make it, because an expert who takes his method from the client invites an attack on his independence.

Expert evidence is the most expensive part of a delay case, and much of it is wasted if the experts answer different questions. The trial in Walter Lilly involved over 32,000 pages of documents, eight experts and costs of about £9m to £10m, which Akenhead J called obviously disproportionate to what is in dispute [2]. So the rules push the experts to agree what they can and to state the rest clearly. Expert evidence is limited to what is reasonably required to resolve the proceedings (CPR r.35.1), and no party may call an expert without the court's permission (r.35.4(1)).

The cheapest agreement is on method, and the best time is before either side has done much work. The SCL Protocol recommends that the parties try to agree an appropriate method of delay analysis before either side does serious work on an after-the-event analysis (11.8). It suggests that a failure to consult might count against a party on costs. The court's hook for that is its discretion over the conduct of the parties, including conduct before proceedings (r.44.2(4)(a), (5)(a)). No reported case has yet applied it, and the Protocol is guidance, not law (Thomas Barnes [109]). Treat the costs point as a risk, not a rule.

The experts should then meet. The TCC Guide says it is generally sensible for the experts to meet at least once before they exchange their reports (14.5.5). The lawyers may help with the agenda and agree it (14.5.2). Save in exceptional circumstances, and then only with the judge's permission, they must not attend or try to dictate what the experts say (14.5.4). The aim is to let the experts speak frankly and give ground without every remark being used against them at trial. So what the experts say in their discussion cannot be referred to at trial unless the parties agree (r.35.12(4)). Nor does their agreement bind the parties unless the parties expressly agree to be bound (r.35.12(5)). So an oral concession in the meeting cannot be put to the expert at trial. What the expert signs in the joint statement can be.

The joint statement records what the experts agree and what they do not, with a summary of their reasons for disagreeing (r.35.12(3)). The TCC Guide calls it a critical document, and its list of disagreements is likely to form an important part of the agenda for trial (14.6.1, 14.6.2). The lawyers must not negotiate, draft or revise it, directly or through their own expert (14.6.3). The experts may share drafts with them for information only. A lawyer with serious concerns about the statement raises them first with the other side's lawyers, not with the lawyer's own expert, and then with the court (14.6.4). The reason is the same as for the duty itself: a statement shaped by lawyers is advocacy, and the court would have no reason to trust it.

Each side must also be able to test the other's analysis. Where an expert relies on a spreadsheet or a software calculation, it must be disclosed to the other parties in its native format with the report (14.7.3). A PDF of a programme shows the bars but not the logic behind them. That the rule reaches a native file of a Primavera or MS Project programme is the better reading of 14.7.3, but no court has decided it. Chapter 14 covers native files and disclosure.

At trial the Guide allows an expert to open with a short presentation, whose purpose is not to introduce new evidence but to explain the existing evidence (14.8.1). How expert evidence is to be taken is considered at the pre-trial review (14.8.2). One option is concurrent evidence, known as "hot-tubbing" (PD 35 11.1). The judge leads the discussion and then invites the parties' advocates to ask questions (PD 35 11.4). The Guide adds that the experts are generally cross-examined on general matters and key issues before the concurrent session (14.8.2).

English delay disputes are decided in three main places. Adjudication is a statutory process that gives a fast decision, binding for now. The Technology and Construction Court gives a final, public judgment that can be appealed. Arbitration gives a final, private award, but only if the contract provides for it. Many disputes settle on the way. Chapter 16 compares how the JCT, NEC and FIDIC forms route disputes.

Adjudication exists because of cash flow. A contractor paid late for work it has done may not survive the years a trial takes. So the Construction Act gives a party to a construction contract the right to refer a dispute arising under the contract for adjudication (s.108(1)). Notice can be given at any time, and the contract must provide a timetable aimed at getting the dispute to the adjudicator within 7 days of the notice (s.108(2)(a)-(b)). The adjudicator must decide within 28 days of referral. The referring party can agree to extend that by 14 days, and both parties together can agree a longer period (s.108(2)(c)-(d)). Where a contract does not comply, the Scheme for Construction Contracts supplies the missing terms (s.108(5)).

The decision is binding until the dispute is finally determined by legal proceedings, by arbitration or by agreement (s.108(3)). Dyson J described the design in Macob (1999) as an intervening provisional stage in the dispute resolution process (as quoted in Carillion [26]). The Supreme Court recorded the motto that sums up the approach: "pay now, argue later" (Bresco [12]). It works because the courts, and in particular the TCC, enforce decisions quickly by summary judgment ([12]; TCC Guide 10.2.4).

The price of speed is accuracy, and the Court of Appeal has accepted it. In Carillion v Devonport (2005) it said that the need to have the "right" answer has been subordinated to the need to have an answer quickly [86]. So the courts enforce a decision unless the adjudicator answered a question he was not asked, or went about his task in a way that was obviously unfair [85]. Save in the plainest cases, a losing party should pay and argue later in court or arbitration [87]. Carillion was not a delay case; the rule is general. The Supreme Court noted a chorus of observations from judges and textbook writers that in most cases the decision becomes final in practice, because nobody challenges it, though statistics are hard to find (Bresco [14]-[15]).

What does that mean for a delay dispute? A full after-the-event delay analysis needs the native programmes, every update, the site records and an as-built programme. It is rarely realistic in 28 days, and the courts do not expect one. Expect a narrower and rougher analysis. In my experience many delay disputes are decided first in adjudication, and many are never taken further. The Act lets the adjudicator take the initiative in ascertaining the facts and the law (s.108(2)(f)). Under the Scheme he may ask for documents and, once he has told the parties, appoint experts or assessors (Sch Pt I paras 13(a), 13(f)). Where he reviews a certifier's interim extension of time, he must do the task the contract clause required (Mace v Baltic [50]; chapter 5).

Two procedural points catch delay parties. The first is jurisdiction. Suppose a party could reserve its position in general terms, take full part, and then, having lost, comb the documents for a new jurisdiction point to resist enforcement. The Court of Appeal said in Bresco v Lonsdale that the purpose of the 1996 Act would be substantially defeated if it could do that as a matter of course [91]. So a party that wants to challenge the adjudicator's jurisdiction must do so "appropriately and clearly". A general reservation is undesirable but may be effective, and much turns on its wording ([92]). This was guidance given obiter.

The second is serial adjudication. A decided dispute must stay decided, or the loser could refer it again until it wins. So an adjudicator cannot decide a dispute that an earlier adjudicator has already decided. The Court of Appeal warned in Sudlows v Global Switch (2023) that serial adjudications about delay carry a greatly increased risk of arguments about which side of the line a later dispute falls [1].

Arbitration works differently. It happens only if the parties have agreed to it, and its object is the fair resolution of disputes by an impartial tribunal without unnecessary delay or expense (Arbitration Act 1996, s.1(a)). The tribunal must act fairly and give each party a reasonable opportunity to put its case and deal with its opponent's (s.33(1)(a)). It may appoint its own expert, and the parties must then have a reasonable opportunity to comment on the expert's views (s.37). A challenge for serious irregularity succeeds only if the court considers that it has caused or will cause substantial injustice (s.68(2)). An appeal on a point of law is possible unless the parties have agreed otherwise (s.69(1)). It needs the other parties' agreement or the court's leave (s.69(2)), and leave is given only in narrow cases, for example where the decision is obviously wrong (s.69(3)(c)(i)). The Arbitration Act 2025 amended the 1996 Act and came fully into force on 1 August 2025.

The fairness duty in arbitration resembles the adjudication cases in the next section, but it has its own test. In Citylink (TCC, 2007) London Underground challenged under s.68 an arbitrator's award of an interim extension of time. Ramsey J said a tribunal should generally decide on the basis of the cases which have been advanced by each party [37(3)]. The arbitrator had accepted the contractor's own submission that the project did not suit critical path analysis, and had decided on the case pleaded, argued and dealt with in evidence ([152], [165], [167]). The challenge was dismissed [168]. Do not assume that the adjudication cases below state the law for arbitration.

Natural justice in adjudication means that each party must have a fair chance to deal with the points that decide the case. The reason is plain. A party cannot answer a case it never knew it faced, and a decision on a point nobody argued may be wrong in ways the parties could have shown. The Act lets the adjudicator take the initiative, so the line is not whether he used his own reasoning. It is whether he told the parties first. The breach must also matter. The remedy is refusal to enforce, and an unfairness that changed nothing caused no injustice. In Cantillon v Urvasco (TCC, 2008) Akenhead J said that any breach of the rules must be more than peripheral; they must be material breaches [57].

The founding delay case is Balfour Beatty v Lambeth (TCC, 2002). The contractor relied on 31 relevant events and put forward no critical path (as described in Try [53]). The adjudicator built his own as-built programme and critical path, by a method neither party had proposed, and awarded 35 days. He did not tell the parties how he was doing it. HHJ Humphrey Lloyd QC refused summary judgment. An adjudicator who intends to use a method not agreed or put forward by either party ought to tell them and get their views. He may not of his own volition use them to make good fundamental deficiencies in the material presented by one party ([33], BAILII numbering).

Later judges have refined the test. In Stellite v Vascroft (TCC, 2016) Carr J asked whether the adjudicator had strayed significantly outside the ambit of the materials and matters advanced before him without giving the parties a chance to comment [59]. She noted that the older phrase, a "frolic of his own", had been criticised. Fairness is calibrated to adjudication's short timetable, but it is not diminished [73]. An adjudicator may also decide on a basis neither party argued, if both knew of the material and the issues had been fairly canvassed (Roe Brickwork v Wates [24]).

Working inside a party's own method is generally safe. In RGB v Victory House (TCC, 2019) the contractor's delay expert had changed logic links in the baseline programme. On her analysis one relevant event, RE3, caused an additional 12 weeks after the effect of RE2 [14], [24]. The employer's expert said her analysis was meaningless because of the extent of the changes, but he ran no analysis of his own and did not have the programmes in native format [40]. The adjudicator asked for the native files. He rejected some of the changed links, re-ran the programme and gave no extension for RE2, which was sub-critical, but 272 days for RE3 [34]. The overall extension claimed was no different; what changed was which event was critical [38]. Jefford J accepted that at first blush the complaint might seem to have merit [36]. But the adjudicator adopted the methodology on which RGB's case was based [44] and did not launch an unrelated analysis of his own [46]. The decision was enforced.

One practical lesson follows, though it is my inference and not a holding. A party that attacks the other side's programme logic without running an analysis of its own widens the adjudicator's freedom to repair that logic.

Materiality is measured in money. Herbosh-Kiere v Dover Harbour Board (TCC, 2012) was a delay-quantum case rather than a critical path case. Both parties valued the delay resource by resource [8]-[9]. The adjudicator instead used a composite daily rate of £21,326, which neither party had argued and which he did not put to them [15], [33]. On the weather claim alone that gave £639,780, against £492,550 on the parties' method [30]. Overall he appears to have allowed in effect over £350,000 more than was being claimed [31]. The breach was material and the decision was not enforced [33]-[35].

Roe Brickwork is the mirror case. The adjudicator's own method produced an overhead and profit figure of £54,900.21, against the £121,625.09 the claimant sought [15]. Any departure from the parties' cases had no effect on the quantum of the claim that was adverse to the Defendant's position [37], and the decision was enforced.

Agreement also answers the complaint. In Try Construction v Eton Town House (TCC, 2003) the parties had agreed that the adjudicator's programming assessor could investigate. The judge distinguished Balfour Beatty, where the adjudicator without agreement or notice used an entirely independent analysis [58].

Two 2026 decisions show the line from each side. In Premier Modular v Maidstone and Tunbridge Wells NHS Trust, under NEC4, the adjudicator found a compensation event using dates from a revised programme that neither party said was an Accepted Programme. The deputy judge called this startling and material, and said it should have been put to the parties. The adjudicator had, in effect, invented a case for the contractor [42]. Summary judgment was refused because the employer had a real prospect of showing a breach [48]-[51]. That is not a final finding that the decision was bad.

In Coventry & Warwickshire v Avison Young, which was not a delay case, the adjudicator also took a point of his own. But he invited submissions from the parties and offered a meeting to address the point [82], and his decision was enforced [99]. Notice, not originality, is the line.

Cross-examining a delay expert tests the method, the assumptions and the fit with the records; the facts belong to the witnesses and the documents. An expert's conclusion is only as good as its inputs and its logic, so the aim is to show that one of them fails.

Start with what the expert was given and what he left out. In Thomas Barnes the contractor's expert did not even include the roof coverings as a potentially relevant critical path activity [133]. Had he read the records he had been given, the judge said, he could not have missed the late start of the roof works [131]. The omission did not cost the contractor its extension. The roof delay was concurrent with the Council's steelwork delay, which limited the money, not the time (chapter 8).

Next, find the assumption that carries the conclusion. In Tata one analysis was circular, because its efficacy as an approach itself depends on the very conclusion it reached [175]. It also answered a backward-looking question with a forward-looking method [199]. The judge saw some force in the point that the expert's second analysis, which picked different events and durations, showed the first to be inadequate [201].

Then ask what the site noticed at the time. No factual witness and no contemporaneous document had noticed the large critical delay the expert found [208]. The Protocol asks for the same common-sense check (SCL 11.2; chapter 7).

Then run the expert's logic against the contemporaneous updates, which needs the native files. Ask whether there was more than one critical path; in Thomas Barnes the Council's expert accepted that it was possible when the judge asked [142]. On method, ask two questions rather than one. Was the method manifestly inappropriate, and did the expert depart from his stated method without explaining why ([110])? A bare "you did not follow the Protocol" point rarely works [109]. Put a lack of independence only if you can prove it, as in Van Oord and ICI. In a hot tub the judge leads, so make credibility points in the ordinary cross-examination first.

A party that loses usually pays twice: its own costs and the winner's. In court and in arbitration costs follow the event: the loser usually pays the winner's legal costs. The reason is that a party forced to litigate to establish its rights should not bear the cost of doing so. The CPR state the general rule (r.44.2(2)(a)), and the Arbitration Act says the same for arbitrators (s.61(2)). The court keeps a discretion and looks at conduct, including conduct before proceedings (r.44.2(4)(a), (5)(a)). Adjudication is different. A contract term about who pays for an adjudication is ineffective unless it meets s.108A. It must be in the written contract and let the adjudicator allocate his fees, or be agreed in writing after the notice. Under the Scheme the adjudicator decides how his own fees are split (para 25).

The bill can also reach people who were not parties. A person who funds a claim and controls it, or stands to benefit from it, is in reality a party to it. If the claim fails, justice will ordinarily require that person to pay the winner's costs (Thomas Barnes (costs) [30], summarising the Dymocks principles). The court has full power to determine by whom and to what extent the costs are to be paid (Senior Courts Act 1981, s.51(3)). In Thomas Barnes (costs) (2026) the family who funded the contractor's failed claim were made subject to a non-party costs order. They were ordered to pay the Council's outstanding costs, jointly and severally ([1], [45]). The claim figure had not been independently valued [20]. The parties had been very far apart on delay, on the reports of their delay experts [22]. Chapter 1 tells the story in full.

How the experts analyse delay is the subject of chapter 6 and chapter 7. The records they work from are in chapter 14, and chapter 16 sets out how each standard form routes a dispute to a tribunal.