Survey data are widely used to provide indicators of economic activity ahead of the publication of official data. This paper proposes an indicator based on a theoretically consistent procedure for quantifying firm-level survey responses that are ordered and categorical. Firms' survey responses are assumed to be triggered by a latent continuous random variable as it crosses thresholds. Breaking tradition these thresholds are not assumed time invariant. An application to firm-level survey data from the Confederation of British Industry shows that the proposed indicator of manufacturing output growth outperforms traditional indicators that assume time-invariant thresholds.