The safety of observing small non-functioning pancreatic neuroendocrine tumours (NF-Pan-NETs) remains under debate. This was a multicentre retrospective study of patients with small incidental NF-Pan-NETs. Survival of patients who underwent upfront surgery versus active surveillance was compared. The risk of death was matched with that in the healthy population. The excess hazard rate and probability of a normal lifespan (NLP) were calculated. Propensity score matching (PSM) with a 1 : 1 ratio was used to minimize the risk of selection bias. Some 222 patients (43.7 per cent) underwent upfront surgery and 285 (56.3 per cent) were observed. The excess hazard rate for the entire cohort was quantifiable as 0.04 (95 per cent c.i. 0 to 0.08) deaths per 1000 persons per year, and the NLP was 99.7 per cent. Patients in the active surveillance group were older (median age 65 versus 58 years; P < 0.001), and more often had co-morbidity (45.3 versus 24.8 per cent; P = 0.001), and smaller tumours (median 12 versus 13 mm; P < 0.001), less frequently located in the pancreatic body-tail (59.5 versus 69.6 per cent; P = 0.008, 59.3 versus 73.9 per cent; P = 0.001). Median follow-up was longer for patients who underwent upfront surgery (5.6 versus 2.7 years; P < 0.001). After PSM, 118 patients per group were included. The excess hazard rates were 0.2 and 0.9 deaths per 1000 persons per year (P = 0.020) for patients in the active surveillance and upfront surgery groups respectively. Corresponding NLPs were 99.9 and 99.5 per cent respectively (P = 0.011). Active surveillance of small incidental NF-Pan-NETs is a reasonable alternative to resection.