Purpose: To quantify errors introduced by rescaling measured dose to compare measured dose distributions with calculated dose distributions for relative dosimetry. Methods: Some dosimeters such as radiographic films require a relationship between dosimeter readings and actually delivered doses. The relationship (or characteristic curves) depends on the manufacturing and environmental conditions under which the dosimeters are made or stored. To compensate for the difference in radiation response among different batches of dosimeters, the measured dose is often rescaled by normalizing the measured dose to a specific dose. This procedure allows us to skip the time‐consuming dosimeter calibration. In this study we derived a mathematical formula of the error in the dose estimated from measurements when this dose scaling procedure is applied to the original dose data. To quantify the errors, we used the characteristic curves experimentally obtained for four different batches of BANG3 polymer gels. A linear equation was used to relate a spin‐spin relaxation rate to a known dose. We chose single calibration equation to estimate doses. The scaled doses were then compared with correct doses which were obtained by using the true calibration equation specific to each batch of polymer gel. Results: The error was smaller than 2% for doses within 1 Gy of the 5‐Gy normalization dose. The error considerably increased as the dose decreased below 5 Gy; but the error was almost constant for doses higher than 5 Gy. The error for a batch requiring a dose scaling factor of 0.87 was larger than the errors for other batches requiring smaller dose scaling factors of 0.93 or 1.02. Conclusions: Dose scaling introduces a large error for doses far from the dose used for normalization. The error is larger for doses lower than the normalization dose. The error increases as the dose scaling factor deviates more from unity.