Lubricants are used in rolling process mainly to reduce the friction. Thus, the measurement of friction can give the idea about the performance of a lubricant. However, the measurement of friction by no means is an easy task. In the past, various methods have been employed for measuring the friction in rolling. Some of these methods require damaging of the surface of the rolls. The methods based on the measurement of roll force, roll torque and the slip can be easily used, but their reliability is dependent on reliability of measuring devices and the mathematical model. A possible way of measuring the average coefficient of friction in rolling is to measure the exit temperature of the strip. It can be easily done by means of temperature sensors. In this work, an inverse method of estimating the approximate value of friction coefficient is proposed based on the exit temperature measurement. The inverse model makes use of a direct model of temperature determination, which is based on finite element analysis and analytical models available in the literature. For a given exit temperature, the inverse model searches the appropriate value of friction coefficient using golden section search algorithm. The methodology is tested by carrying out a number of numerical experiments on the cold strip rolling. Some preliminary experiments have been conducted. It is planned to carry out more experiments in future. Although the direct model used in this work is highly approximate, the entire methodology displays its high potentiality in an industrial setting. In any case, the methodology can compare two lubricants with respect to their ability to reduce the coefficient of friction, even if the estimated coefficient of friction may be approximate.