Purpose High intensity occupational physical activity (OPA) seem to aggravate health and increase risk of sick leave and early retirement. Most intensity of OPA monitoring has been self-reported, e.g. by rating of perceived exertion (RPE). However, no studies have investigated the precision and risk of bias in RPE reporting during free-living OPA. This study investigated the agreement between OPA intensity in percentage of the heart rate reserve (%HRR) estimated from RPE and device-measured heart rate (HR), and potential bias factors on this agreement. Methods The CR10 scale measured RPE at work. The Actiheart monitor measured HR during 24-h a day for 2-4 days. Both RPE and device-worn HR were converted to %HRR. The difference between both %HRR and their limits of agreement was determined in a Bland Altman plot. To detect bias factors, the difference between both %HRR was regressed on age, sex, cardiorespiratory fitness, occupational lifting, medication, consequences of musculoskeletal disorders and the interactions between these factors with device-work %HRR. Results Six hundred and twenty-three participants were included in the analysis. Mean difference between RPE-based and device-worn %HRR was 54.6% (SD 19.5). The limits of agreement were wide (11.6-90.1%HRR). Age (0.48%HRR, 95% CI 0.18-0.79) occupational lifting (9.84%HRR, 95% CI 3.85-15.83) and cardiorespiratory fitness (0.41%HRR, 95% CI 0.03-0.79) significantly biased the agreement between the estimations of OPA intensity. Conclusion RPE overestimated OPA intensity, and was biased by several factors. Device-worn %HRR should be preferred when evaluating OPA intensity among workers with physically demanding jobs.