The North American Lightning Detection Network has been providing lightning data since 1998. Important applications, such as detection of lightning-caused forest fires, power line fault location and aviation safety procedures, have triggered a number of hardware and software upgrades for improvement of the network performance characteristics, including its detection efficiency and location accuracy. The performance characteristics of the network have been here evaluated by examining the reported North American Lightning Detection Network (NALDN) return-stroke data based on the lightning current measured at CN Tower during three major storms (2005, 2011 and 2014). Each of these three storms followed a substantial network upgrade in 2003–2004, 2010–2011 and 2013–2014, respectively. The objective of this investigation is to outline the evolution of the network and determine where there may be opportunities for additional improvements. (It is important to stress that the evaluation of NALDN in the vicinity of the tower is mainly based on tall-structure lightning, which is upward-initialed, due to the rarity of recording downward-initiated lightning at the extremely tall tower.) Since 1990, the CN Tower has been equipped to measure the lightning current derivative signals. Its 553-m height has allowed the recording of many hundreds of current derivative signals, mostly resulting from return strokes. Also, image recording systems have been used to capture images of strikes to the tower. The evaluation of the NALDN performance characteristics includes polarity, flash and stroke detection efficiency, location accuracy and peak current estimation. The 2013 deployment of LS7002 digital sensors with the enhanced embedded software has substantially improved the sensitivity of the sensors leading to greater flash and return-stroke detection efficiencies. Furthermore, the 2014 total lightning processor (TLP100), well designed with new algorithm and techniques, provides smaller time-of-arrival errors and better location accuracy. Based on this investigation, following the latest network upgrade, the polarities of the NALDN-detected return strokes were perfectly matched with the polarities of the corresponding return-stroke currents measured at the tower. In 2014, the NALDN detected every flash that contained at least one return stroke, as well as all recorded CN Tower return strokes, resulting in 100% flash- and stroke-detection efficiencies. The perfect stroke-detection efficiency in 2014 confirms a significant network improvement in comparison with the assessment in 2005 (55.3%) and 2011 (53.8%). In the 2014 evaluation, relative to the location of the tower, the NALDN is found to have a median absolute location error of 119m and an average absolute location error of 129m. The 2014 median absolute location error is found to be significantly lower than those for 2005 (289m) and 2011 (199m). Based on 2014 evaluation, the NALDN is found to overestimate the current peak measured at the tower by a factor of 3.89, which is due to field enhancement effect resulting from high speed propagation of the current within the tall tower. The presented analysis shows that the latest NALDN upgrade (2013–2014) following the 2003–2004 and 2010–2011 upgrades, has substantially improved the NALDN performance characteristics, especially in terms of stroke-detection efficiency and location accuracy.