A recent TCC decision has provided guidance on the bringing of representative defect claims based on sample evidence and expert statistical analysis. This appears to be the first time the TCC has considered the use of statistical evidence to support such claims. The case is likely to be of relevance to parties considering allegations of endemic failings with regard to specific aspects of construction work.
Amey LG Limited v Cumbria County Council
Amey was contracted by Cumbria to provide highways maintenance and associated services for a term of 7 years. The relationship between the parties subsequently deteriorated and following the expiration of the contract, Amey commenced proceedings to recover sums deducted by Cumbria from Amey’s final monthly payment application. In response, Cumbria advanced a number of counter-claims, including a claim for the cost of remedial works to repair a proportion of patching and surfacing works carried out by Amey, which Cumbria claimed to be defective. Cumbria advanced this claim on the basis that it had examined a sample number of patching and surfacing works undertaken by Amey, and that its conclusions as to the defective nature of those samples could be extrapolated to the entirety of the works of that nature undertaken by Amey over the contract period.
The extrapolation of Cumbria’s claims meant that the financial value of the losses claimed was much greater than those connected specifically to the samples examined by Cumbria. For one claim, the cost of remedial works for the sampled items was approximately £22,000 but would rise to approximately £1.69 million when extrapolated to the rest of the works.
The key questions for the court to determine in relation to extrapolation were:
- Whether Cumbria was entitled to advance its case based on a sample of evidence.
- Whether the method of sampling used by Cumbria was acceptable to advance its extrapolation case.
- If the method was acceptable, whether the statistical evidence in relation to the sample set was sufficient to discharge the legal burden of proof in relation to its claim.
The court accepted that the substantial quantities of patching and surfacing works carried out by Amey under the contract made it impractical for Cumbria to have inspected every item of work and to have pleaded and proved its case in relation to each allegedly defective item separately. Cumbria was therefore entitled to advance its case on the basis of sampling. It is unclear from the court’s decision whether it would still have been permissible for Amey to rely on sampling if it was not impractical, but simply more expensive or time-consuming, to prove each item of defective work separately.
With regard to the second issue, the court rejected Cumbria’s initial position that its sample evidence could be extrapolated with a 95% confidence rate across the whole of the works. The court noted that it was well understood by statisticians that this level of confidence could only be demonstrated mathematically if the sample evidence was obtained by a genuinely random sampling process. Cumbria ultimately accepted that its sample was not sufficiently random and, whilst this ruled out proof of a 95% rate of confidence, the court found that there was no principle of law or statistical theory to suggest that such a claim could only be established by statistically random sampling. Cumbria was therefore entitled to rely upon its sample evidence but was required to demonstrate that, whilst it may not be statistically random, it was still sufficiently representative of the whole of the works.
Ultimately, Cumbria failed to show that its sample evidence was sufficiently representative:
- The sample was initially obtained to ascertain the presence or absence of defects and not a general sample of the works carried out by Amey. Accordingly, the sample was not being used for the intended purpose of its collection.
- The sampling process had been extended over a lengthy period of time.
- Patches from the sample which could not be located on a GPS were excluded from the statistical analysis, thereby excluding samples from works carried out earlier in the project before GPS was being used.
- The sample excluded patches classified as “pre-surface dressing patches” which had subsequently been covered by surfacing.
The court held that each of these matters demonstrated an opportunity for the sample to be infected with bias. Cumbria was also found to have failed to have proper processes for collection of the sample and did not take suitable steps to mitigate and/or avoid the possibility of bias. In these circumstances, it was held that it would be unsafe to extrapolate the sample evidence relied upon by Cumbria.
Conclusions and implications
This case provides important guidance for the bringing of representative defects claims under construction contracts. The need for such claims can arise wherever a construction project involves repetitive work, such as welding, bricklaying, glazing or road repairs as in the present case.
Although it will always be preferable for a claimant to prove each item of defective work, the present case shows that where this is impractical or impossible, evidencing a claim by reference to a statistically random or sufficiently representative sample will be permissible. The difficulties involved in doing so should not be underestimated, however. Care is needed from the outset to ensure any sample evidence collected is genuinely random and/or sufficiently representative and that all possible steps are taken to avoid the sample being affected by bias.