BACKGROUND: Reducing the portion sizes of foods available in restaurants and cafeterias is one promising approach to reducing energy intake, but there is little evidence of its impact from randomised studies in field settings. This study aims to i. examine the feasibility and acceptability, and ii. estimate the impact on energy purchased, of reducing portion sizes in worksite cafeterias. METHODS: Nine worksites in England were recruited to reduce by at least 10% the portion sizes of foods available in their cafeterias from targeted categories (main meals, sides, desserts, cakes). In a stepped wedge randomised controlled pilot trial, each site was randomised to a date of implementation, staggered fortnightly, following a baseline period of four weeks. Impact on energy purchased was analysed using generalised linear mixed modelling. We also assessed feasibility, acceptability, and fidelity of intervention implementation. RESULTS: Data from six of the nine randomised sites were analysed, with three sites excluded for not providing sufficient data and/or not implementing the intervention. The extent to which the intervention was implemented varied by site, with between 6 and 49% of products altered within targeted categories. Feedback following the intervention suggested it was broadly acceptable to customers and cafeteria staff. For the primary outcome of daily energy (kcal) purchased from intervention categories, there was no statistically significant change when data from all six sites were pooled: percentage change - 8.9% (95% CI: -16.7, - 0.4; p = 0.081). Each of these six sites showed reductions in energy purchased, ranging from - 15.6 to - 0.3%, which were borderline statistically significant at two sites (respective percentage changes (95% CIs): - 15.6% (- 26.7, - 2.8); - 14.0% (- 25.0, - 1.2)). Secondary outcome data are suggestive of a compensatory increase in energy purchased from food categories not targeted by the intervention, with no overall effect observed on energy purchased across all categories. CONCLUSIONS: The results of this pilot trial suggest that reducing portion sizes could be effective in reducing energy purchased and consumed from targeted food categories, and merits investigation in a larger trial. Future studies will need to address factors that prevented optimal implementation including site dropout and application across a limited range of products. TRIAL REGISTRATION: ( ISRCTN52923504 ). Registered on 20th September 2016.