The August 17, 1999 Izmit (Turkey) earthquake (M_w=7.4) will be remembered as one of the largest earthquakes of recent times that affected a large urban environment. The shaking that caused the widespread damage and destruction was recorded by a handful of accelerographs operated by different networks in the earthquake area. As rebuilding in Turkey starts and picks up speed, the recorded ground motions that adversely affected the built-up environment are being studied intensively to explain the past disaster and for the implications for future earthquakes. New forecasts of future large earthquakes in the ever growing urban environment of Istanbul and its vicinities necessitates better preparation and extensive planning to upgrade and retrofit existing infrastructures. To successfully execute these endeavors, it is imperative to have better defined ground motion characteristics. This paper aims at discussing the issues related to the engineering implications of the recorded ground motions. The main shock records show that the peak accelerations, even those from near-field stations that exhibit the characteristic near-fault pulses that cause large displacements, are smaller than expected. On the other hand, smaller magnitude aftershocks yielded larger peak accelerations. This is attributed to the sparse networks, which possibly missed the recording of larger motions during the main shock - particularly in the heavily damaged areas of South Izmit Bay. Aftershocks recorded at such areas are used to estimate the missed main shock ground motions. Accordingly, the estimated motions in South Izmit Bay are approximately 1 g highly polarized in N-S direction. These attributes have significant engineering implications to be considered during the rebuilding phase and beyond.