Formal performance measurement systems have become a prominent feature throughout most of the public sector. The NHS is no exception and the government is developing and expanding on existing performance data to produce a new national framework which is to be at the heart of the 'performance-led' NHS. The success of formal performance measurement systems depends in part on the degree to which they can capture adequately relevant information within a quantitative framework. In this paper, we explore the use of formal, quantitative 'hard' information and informal, subjective 'soft' information in the assessment of the performance of NHS hospital Trusts by external organisations. Our empirical work is summarised into 3 main themes: (i) the use of 'soft' information as a complement to 'hard' information; (ii) the use of 'soft' information as a substitute for 'hard' information; and (iii) the use of 'hard' information as a safety net in the assessment of performance. We argue that 'soft' information plays a valuable role in the assessment of performance of NHS Trusts and note that this is also reflected in current practice in the private sector. We argue that one of the main functions of 'hard' information in performance assessment is to act as a safety net in order to identify laggards by highlighting poor performance. Rarely is it used as a means to encourage good performance or to identify best practice. Whilst the safety net function may indeed be useful, if the new national performance framework is to achieve its aim of promoting good performance, the limitations of formal systems will have to be taken into account. We argue that the advantages of 'soft' information in the NHS should be acknowledged, rather than trying to force it into a formal framework where the benefits may be nullified. However, we also warn against excessive reliance on 'soft' information and suggest that a balanced system which allows both 'hard' and 'soft' information to flourish is the optimal solution to performance assessment in the NHS.