How does one resolve a dispute involving thousands of individual items where it is impractical to deal separately with each one? A recent decision in the Technology and Construction Court provides some guidance - but raises further questions.
The case is Standard Life Assurance Limited v Gleeds (UK)(a firm) and Others (December 2020, TCC). Standard Life had engaged Costain as its main contractor for the development of a large residential and retail development in Berkshire. It had paid some £146 million in settlement of Costain’s final account.
Standard Life’s’s case was that about £38 million of that sum was due to the negligence of its professional team. It alleged that various members of that team (who were now the defendants) had been responsible for the issue to Costain of some 3600 instructions. It was those instructions, said Standard Life, that had caused Costain’s final account to increase by £38 million.
A clear logistical problem was that the court could not be expected to try 3600 individual instructions. However, Standard Life had analysed some of these instructions in detail, and had given the defendants full particulars of 122 of them. They maintained that the court’s findings on these representative samples could then be extrapolated so as to draw inferences in relation to the remaining instructions.
Three of the defendant firms objected to this approach. They argued that sampling was not appropriate in principle for a professional negligence claim. They also objected to Standard Life’s proposed samples, which they said were tilted towards the high-value variations. Finally, they criticised the way in which the case had been pleaded. They asked the court to strike the claim out altogether.
The judge rejected this application. Instead, he proposed giving outline directions, subject to refinement, which were tailored towards ensuring that the sample was a valid one.
The draft directions allowed the three defendants to nominate between them some 160 variations that would make up the sample. First however, Standard Life would have the opportunity to whittle down its claim to the variations on which it genuinely thought it could rely. There was a subtle point to this. Standard Life would be incentivised to get rid of its weaker claims, for fear that the defendants might otherwise nominate those claims as the samples.
This case involved an employer seeking to pass on to others its own liability for a contractor’s claim. Could a contractor that wanted to make a claim for loss and expense or increased costs, based on hundreds or thousands of variation orders, also argue for extrapolation by sample?
Such a claim would be a global one. Global claims have long been criticised by employers as failing to match cause to effect. But it has been accepted - for example, in Walter Lilly v Mackay (TCC,2012) - that they are permissible in principle, albeit the claimant must overcome steep evidential barriers.
A contractor’s claim for loss and expense would be very difficult to mount on the basis of producing samples of variation orders, or other events, that could be said to be extrapolated to apply to other orders or events. A claimant would need to show a high degree of uniformity among the orders or events relied upon. Such degrees of uniformity do not usually apply to variation orders or other instructions. That is not to say that such a claim is impossible in law.
Sampling is, however, more suited to defects cases. A court may readily conclude, for example, that the presence of faulty welds in a reinforced concrete slab is indicative of similar faulty welds elsewhere in the slab.
Proof by way of extrapolation of samples could theoretically apply to a loss and expense claim, but is more likely to be confined in practice to defects cases and to the sort of situation that arose in Standard Life (where liability for a contractor’s claim was being passed on). And in such cases, if sampling is agreed or allowed, there is much to be said for the method proposed by the judge in that case: namely, giving the defendant a large say in selecting the applicable samples.