Background Physical activity is an important determinant of human health but it is inherently difficult to measure. Global surveillance systems for physical activity have so far only included self-report measures, which capture only a small subset of daily activity and are limited due to issues of recall bias. Wrist-worn accelerometry offers a reasonably cost-effective objective method of measuring physical activity during free-living with proven feasibility in large-scale population studies. The key objective of this project was to pilot wrist-worn accelerometry within a surveillance setting in order to inform the implementation of this methodology into the global WHO STEPS programme. Method development and implementation Accelerometry protocols were developed and deployed within an existing STEPS survey in two regions of Malawi (Dowa and Lilongwe). This also included developing training for local staff. Survey information was collected on tablets. Accelerometers could only be set up on PCs, so the protocol was adapted to do this in advance of recruiting participants. For this, an alternative Participant ID linkage system was developed to enable linking accelerometer files to the rest of the survey data. On the whole, the implementation was successful. During the process evaluation, some issues were identified. For example, black wrist straps were culturally associated with the Devil by some participants. A total of 499 participants were recruited for Step 1 and 2, of whom 446 returned for Step 3 measurements which included accelerometry. The accelerometry data collection was well accepted by both fieldwork team members and study participants, with only four participants (<1% of those eligible) declining to wear the device. There were no major technical issues with devices, although a small number of wrist straps were damaged and 13 monitors were lost (3% of deployed). Of 456 accelerometer files retrieved, 410 files (90%) could be linked to survey participants. All but two accelerometer files could be processed with standard techniques to produce participant-level summary results. Sufficient valid data (defined as at least 48 hours of monitor wear time with reasonable diurnal representation) were available for 386 survey participants (87% of eligible). Results Objective levels of physical activity in Malawi from this pilot study were about 50% higher than levels observed in the UK using similar methodology. Rural dwellers were more active than urban dwellers, particularly in the morning hours of the day. Men had higher activity levels compared to women, and there were decreasing trends with advancing age. Conclusion and recommendations This pilot demonstrated that it is feasible to implement wrist-worn accelerometry within the STEPS program in settings such as Malawi. Detailed description of objectively measured physical activity patterns could be produced from nearly all accelerometer files retrieved, including behavioural indicators known to be important for human health. In future surveys, culturally specific issues that could impact data collection should be identified early in the fieldwork planning stage and changes to the protocol made. Experiences from this pilot have led to development of software platforms that allow accelerometers to be set up from Android tablets at the point of issuing the device to the participant, which would simplify future fieldwork training and reduce risk of data linkage error.