Abstract Extraction of plastic particles from soil is challenging and, thus, exceptionally little spatial information on plastic distribution at the field scale has been gathered. However, for environmental risk assessment, adequate sampling should complement coherent plastic profiling. In this study, we investigated the spatial distribution of mesoplastics (MePs; from >5 mm up to 130 mm) in arable soil (Haplic Cambisol) managed intensively by 12 years of compost application. Geo‐referenced samples ( n = 128) and five different sampling designs ( n = 45) of variable sampling volume (from 2 to 300 L) were collected at a three hectare study site in Northern Germany (0–30 cm soil depth). Soil properties such as pH and soil organic carbon (SOC) were measured to evaluate dispersion measures of these data. In total, we found 259 MePs with a predominance of transparent packaging foils made of polyethylene and coloured fibres of polypropylene. Average particle metrics were a projection area of 47 (3–400) mm 2 , a Feret diameter of 18.5 (5.4–130) mm and a mass of 1.89 (0.11–221) mg. Caution is advised when measuring the particle mass due to still strongly adhering soil material, especially for fibre bundles with 0.544 mg soil mg −1 particle. We recommend using a 0.1 mol L −1 tetrasodium pyrophosphate solution to purify MePs by removing attached soil before weighing for further environmental risk assessment. The MePs count with a median value of 0.50 (0–3.2) particles kg −1 and median mass of 2.26 (0–221) mg kg −1 featured the highest coefficient of variation (CV) with 103% and 187%, respectively. This is 10–20 times larger in comparison to the CV of SOC (9.2%) and even 50–93 times larger than CV of soil pH (2.2%). This leads to the need of larger sample numbers to delineate plastic metrics in comparison with soil properties to identify a reliable mean value of the field within a predefined allowable error. Mesoplastics in the soil were characterized by a pure nugget effect variogram (no spatial correlation), revealed no intrafield variability and the sample volume yielded inconclusive results. Sampling for plastics in soil should either (i) drastically increase the sample number for a single field or (ii) communicate transparently that the allowable error is by far enhanced in comparison with classical soil properties like pH and SOC. More systematic studies featuring geo‐spatial analysis of MePs and smaller‐sized plastics in soils are required to propose adequate sampling designs across multiple land uses and plastics fingerprints. A larger database would, thereupon, pave the way for best‐practice guides on how to treat ‘outliers’ and search for robust estimators for spatial mapping of plastics in soils.