Stellar spectra derived from current multiple-object fiber-fed spectroscopic radial-velocity surveys, of the type feasible with, among other examples, AUTOFIB, 2dF, HYDRA, NESSIE, and the Sloan survey, differ significantly from those traditionally used for determination of stellar abundances. The spectra tend to be of moderate resolution (around 1 Angstrom) and signal-to-noise ratio (around 10-20 per resolution element), and cannot usually have reliable continuum shapes determined over wavelength ranges in excess of a few tens of angstroms. Nonetheless, with care and a calibration of stellar effective temperature from photometry, independent of the spectroscopy, reliable iron abundances can be derived. We have developed techniques to extract true iron abundances and surface,oravities from low signal-to-noise ratio, intermediate resolution spectra of G-type stars in the 4000-5000 Angstrom wavelength region. The theoretical basis and calibration using synthetic spectra are described in detail in another paper (Jones et al. 1995). The practical application of these techniques to observational data, which requires some modification from the ideal case of synthetic data, is given in the present paper. An externally derived estimate of stellar effective temperature is required in order to constrain parameter space sufficiently; a new derivation of the V-I-effective temperature relation is thus an integral part of the analysis presented here. We have derived this relationship from analysis of available relevant data for metal-poor G dwarfs, the first such calibration, We test and calibrate our techniques by analysis of spectra of the twilight sky, of member stars of the cluster M67, and of a set of field stars of known metallicity. We show that this method, combined with our new color-temperature calibration, can provide true iron abundances, with an uncertainty of less than 0.2 dex over the range of metallicity found in the Galactic thick and thin disks, from spectra obtained with fiber-fed spectrographs.