The measurement uncertainty is a non-negative parameter associated with the result of a measurement that characterizes the dispersion of the quantity values that could reasonably be attributed to the measurand. In the present study measurement uncertainty is estimated using the GUM (ISO/IEC Guide 98: 1993 Guide to the expression of uncertainty in measurement) bottom-up approach. The steps were followed: specifying the measurand; identifying all the associated sources of uncertainty; quantifying the uncertainty components; combining the uncertainty components; determining the extended combined standard uncertainty; reviewing the estimates and reporting the measurement uncertainty. In this process, the major uncertainty components with greater impact were identified; try to eliminate or to reduce the impact of these components can improve measurement methods. Examples were the determination of aluminum in starch and bread crumbs by inductively coupled plasma-mass spectrometry (ICP-MS). The uncertainties of aluminum contents were from measurement repeatability, variability of calibration curve, standard stock solution, dilution, solution volume and sample weighing. The data indicated that the major contributions to the uncertainty budget originating from urel(cAl)1 (the relative standard uncertainty of aluminum content derived from linear least squares calibration), urel(cAl)3 (the relative standard uncertainty of aluminum content derived from the dilution of the standard stock solutions) and urel(rep) (the relative standard uncertainty derived from the repeatability). Based on the analysis of the main individual contributions of each uncertainty source to the total uncertainty value, several modifications were proposed. Firstly helium collision mode was replaced by no gas mode to improve the sensitivity of mass spectrometry. Secondly the number of measurements was increased. Thirdly let the mean of data points in the calibration closer the measurand. Finally the relative error smaller gauges were used. After these modifications, urel(cAl)1, urel(cAl)3 and urel(rep) were from (0.035 8, 0.013 2, 0.008 5) down to (0.006 0, 0.010 5, 0.003 3), respectively; the combined relative standard uncertainty of aluminum was from 0.039 down to 0.013; the expanded uncertainty from 1.8 down to 0.4 mg·kg-1(coverage factor k=2). The improvement effect was significant.