Because stress is subjective and is expressed differently from one person to another, generic stress prediction models (i.e., models that predict the stress of any person) perform crudely. Only person-specific models (i.e., ones that predict the stress of a preordained person) yield reliable predictions, but they are not adaptable and are costly to deploy in realworld environments. For illustration, in an office environment, a stress monitoring system that uses person-specific models would require the collection of new data and the training of a new model for every employee. Moreover, once deployed, the models would deteriorate and need expensive periodic upgrades because stress is dynamic and depends on unforeseeable factors. We propose a simple, yet practical and cost-effective calibration technique that derives an accurate and personalized stress prediction model from physiological samples collected from a large population. We validate our approach on two stress datasets. The results show that our technique performs much better than a generic model. For instance, a generic model achieved only 42.5 ± 19.9% accuracy. However, with only 100 calibration samples, we raised its accuracy to 95.2 ± 0.5%. We also propose a blueprint for a stress monitoring system based on our strategy, and we debate its merits and limitations. Finally, we made our source code and the relevant datasets public to allow other researchers to replicate our findings.