The commonest way of geo-referencing farms as single points is using the location of the farmhouse as either read off a map or approximated by its postcode. While these two methods may be adequate for small farms, they are unlikely to be satisfactory for large ones, or alternatively when they are comprised of several discrete units or holdings.In order to investigate the best representation of the total farm polygon(s) by a single point, we undertook a study using nearly 500 actual farm boundaries in the county of Cornwall, England. For each farm, the farm boundaries were digitised, and its area and centroid determined using ArcView 3.2. A variety of point geo-referencing systems were tested to find the best single point location for a farm, as judged by the proportion of farm area captured. Whilst the centroid was found to capture the largest area, the main farm building was judged to be the best geo-referencing method for practical purposes. In contrast, the various systems of geo-coding using the farm postal address performed relatively poorly. Where there are separate parcels of land managed together in a single parish, they may be identified as a single unit, but if there are separate parcels in different parishes they should be identified as separate units.The implications of these results for Great Britain’s national animal health information system (VETNET) are discussed.