Applications of NIR spectroscopy in the QA/QC environment are becoming increasingly commonplace and more stringent. Current NIR spectrometers must be highly stable and very sensitive to facilitate the development of robust and reliable calibration models. The development of a good calibration model is generally very time consuming. It is imperative that a calibration model once developed be useful for a long period of time and be transferable to other instruments with as little additional processing as possible. Many large scale research facilities have a central location where method development is performed, as well as satellite laboratories where analyses employing those methods are carried out. In such situations, the need for rapid implementation of a centrally developed calibration model in a satellite facility with as little rework as possible is evident. Ideally, this would take the form of a direct transfer of the calibration model from one instrument to the next, with no mathematical manipulation or recalibration at all. A further need for direct calibration transfer arises from maintenance and repair procedures. Commonly, the replacement of optical components within a spectrometer will cause a previously functioning calibration model to behave erratically or to no longer function at all. It is highly desirable that this situation be avoided and that calibration models continue to function following such routine maintenance or repairs. Effectively, this amounts to a need for direct transfer of the calibration model from the pre-repair instrument to the post-repair instrument. Several key factors have been identified as having influence over the ability to perform direct calibration transfers successfully. The wavelength precision, photometric precision and absorbance scale linearity of a given instrument over time are all significant factors. The wavelength accuracy and photometric accuracy among different instruments are also significant. The development and validation of a stable and robust calibration model is certainly another factor. Instruments with shortcomings in any