Accurate measurement of control delay and queue length is important for effective traffic management at signalized intersections and to assess its performance. The analytical methods which are commonly used to estimate delays are primarily developed for homogeneous lane-based traffic conditions and require lot of inputs for delay estimation. If cycle lengths are not the same for a signalized intersection operated by traffic police, then use of analytical methods is restricted. The present study tries to overcome these limitations by proposing an approach in which the individual delay components were directly measured using travel times from video as input and electric poles on median as reference points. Traffic captured through videographic survey at a busy intersection in Vellore, India, was used to illustrate the proposed approach. Instead of conventional stopped delay concept, a new delay called ‘Stop & Go’ delay which is more appropriate for heavily congested intersections has been proposed and used. This ‘Stop & Go’ delay concept captures the time lost due to movement of queue in addition to frequent stoppings which is rather not possible with stopped delay definition. It was found that the control delays experienced by different vehicle types are not the same and four-wheeler exhibits longer delays than two- and three-wheelers. One of the main advantages of the proposed approach is it does not require many parameters to calculate control delay and only travel times of sample vehicles are sufficient. Another advantage with the proposed approach is it can be applied to all types of signalized intersections irrespective of whether it has a fixed time signal or manually operated one. With the proposed approach, one can get vehicle class-wise individual delay components which may not be possible with the existing methods available in the literature. In addition to delay measurement, the present study also developed polynomial regression models to predict queue length at intersections using control delay as independent variable. The results are promising and can be used to predict the queue length at signalized intersections in India. © 2020, The Institution of Engineers (India).