Portable XRF is widely applied in the investigation of contaminated sites as a fast-screening tool for heavy metals, but very few studies focused on the assessment of XRF test results to guide the sample selection in laboratory detection. In order to evaluate the guidance of portable XRF in practical applications, the XRF and lab measured data of six heavy metals of Cu, Pb, As, Ni, Cd, and Hg were selected from two sites. The linear fitting, Pearson correlation and differences between XRF and lab values were analyzed. The results showed that there was a significantly positive correlation between XRF and lab data in Cu, Pb, As and Ni. XRF data of Hg and Cd were higher than those of lab data, while the values of Cu, Pb, As and Ni detected by XRF were smaller than those of lab data with more than 50% of the samples. Among the six heavy metals, Cu and Pb performed better in the correlation analyzed and linear fitting results between XRF and lab data (correlation coefficient=0.5-0.8, R²>0.7). Moreover, Cu had better fitting results between XRF and lab data in XRF data high than 1 000 mg/kg (R²>0.8), the better fitting results of As and Pb were found in the range of 4.0-10.0 mg/kg and the detection limits to 80 mg/kg respectively. Furthermore, when the half screening values of heavy metals were used as the standard for selecting lab samples, the accuracy of predicting contaminated samples by XRF in Cu, Pb and As were 87.5%, 100%, and 67.9% respectively, which were an increase of 12.5%, 66.67% and 17.9% comparing to direct use of the screening values. Therefore, the screening criteria should be appropriately lowered in selecting lab detection samples when using XRF data in investigating contaminated sites.