'Farms like mine': a big data method in peer matching for agricultural benchmarking.

To find opportunities to improve, in efficiency or performance, farms are often compared on the basis of standard typologies (i.e. categorisations). For example the EU "specialist-cerealsoilseeds-pulses" farm type, known in Britain as "cereals" farms. These categories, being aggregates, contain significant numbers of atypical enterprises. For example, in 2017 there were 30 cattle and 69 sheep on the average "general-cropping" farm in England. This means that comparators are averages across farms with widely divergent scales of different enterprises (and hence farm characteristics), that are not relevant for the comparison. Furthermore, farmers may not necessarily even know their own farm "type" when undertaking benchmarking or comparative analysis. We therefore present a novel method that matches a specific farm against all farms in a survey (drawing upon the Farm Business Survey (FBS) sample), and then selects the nearest "bespoke farm group" of matches based on distance (Z-score) away. Across 34 dimensions, including almost all the enterprises characteristic of English farms, as well as tenure and geographic proximity. Means and other statistics are calculated specifically for that bespoke farm comparator group, or "peer set" of 25 farms or more if less than 1 Z-score away. This generates a uniquely defined comparator, for each individual farm and gives a substantially improved key-performance-indicators for benchmarking purposes. This methodology has potential to be applied across the full range of FBS farm types and across a wider range of contexts.