It is currently common to use spatially unresolved multi-filter broad-band photometry to determine the masses of individual stellar clusters (and hence the cluster mass function, CMF). I analyze the stochastic effects introduced by the sampling of the stellar initial mass function (SIMF) in the derivation of the individual masses and the CMF and I establish that such effects are the largest contributor to the observational uncertainties. An analytical solution, valid in the limit where uncertainties are small, is provided to establish the range of cluster masses over which the CMF slope can be obtained with a given accuracy. The validity of the analytical solution is extended to higher mass uncertainties using Monte Carlo simulations and the Gamma approximation. The value of the Poisson mass is calculated for a large range of ages and a variety of filters for solar-metallicity clusters measured with single-filter photometry. A method that uses the code CHORIZOS is presented to simultaneously derive masses, ages, and extinctions. The classical method of using unweighted UBV photometry to simultaneously establish ages and extinctions of stellar clusters is found to be unreliable for clusters older than approx. 30 Ma, even for relatively large cluster masses. On the other hand, augmenting the filter set to include longer-wavelength filters and using weights for each filter increases the range of masses and ages that can be accurately measured with unresolved photometry. Nevertheless, a relatively large range of masses and ages is found to be dominated by SIMF sampling effects that render the observed masses useless, even when using UBVRIJHK photometry. A revision of some literature results affected by these effects is presented and possible solutions for future observations and analyses are suggested.