Background: The FDA have recently published draft guidance for the development of treatments for early Alzheimer's disease. Key features of this guidance are the advocacy of sensitive cognitive measures and a taxonomy of disease severity. Whilst desirable patterns of cognitive-functional improvement are included, specific measures, and the magnitude of required effects, are not described. Main section: We describe key elements of the guidance content, especially with regard targeting key cognitive domains and the means by which they might be efficiently indexed in the disease stages included in the guidance. We discuss also the opportunities to assess cognitive performance in 'Stage 2' and 'Stage 3' patients, as well as the possibilities for effectively assessing function in the latter category. In this section we review candidate cognitive assessments that we judge are capable of delivering on the guidance specification for sensitive neuropsychological measures. This includes detailed consideration of the ADCS-PACC and Catch-Cog initiatives. With respect to the magnitude of effects, we propose that standardised effect sizes of 0.3 represent a reasonable level of efficacy based on the observation that already marketed drugs on average deliver this level of improvement. Conclusions: We propose the use of cognitive measures in stage 2 patients to index the cognitive skills known to be compromised early in the Alzheimer's disease process. We recommend extending the traditional interest in episodic memory to include sensitive, reliable and valid measures of attention, working memory and aspects of executive function. We propose a focus on these additional cognitive abilities based on evidence that performance on tests of these domains is moderately well related to functional skills.