This paper demonstrates the application of full Bayes multivariate linear intervention models to evaluate the impact of the Signal Head Upgrade Program in the City of Surrey, British Columbia. The program was implemented by the Insurance Corporation of British Columbia (ICBC) and is aimed at reducing the number of collisions by improving signal visibility. Specific improvements include: upgrading signal lens size, installing new backplates, adding reflective tapes to existing backplates, and installing additional signal heads. The evaluation utilizes full Bayes multivariate linear intervention models with a multiple regression links representing time, treatment, and interaction effects as well as the traffic volumes effects. The models were utilized to determine the safety impacts of the program on each combination of collision severity and collision time of occurrence (i.e. severe/day-time, severe/night-time, PDO/day-time, PDO/night-time) in the City of Surrey. The results indicated the effectiveness of the program in improving the safety of the treated intersections as there were reductions in all types of collisions at the treated sites after the improvement. However, such reductions were statistically significant only for night-time severe collisions and day-time PDO collisions.