I know I am a few months late to this game, but I wanted to give you a compliment and issue a small caveat.

Compliment: great article. It is reassuring, for me anyway as an RDBMS-agnostic designer, that I wholeheartedly agree with your physical modeling approaches. Sizing indexes by understanding column data types is critically important. Nulls matter, as does premature denormalization. I already apply 9 of the 10 suggestions in practice and all 10 in spirit, which leads me to the caveat.

Caveat: If you are in a Microsoft technology-only shop with .NET, C# and other such technologies then no worries. Otherwise be careful with using non-ANSI standard data types like "tinyint" and "bit". The issue isn't one of ANSI compliance for its own sake, but for the sake of compatibility with non-Microsoft external drivers, add-on reporting systems and integration tools.

Once again thanks for the confirmation on my physical data modeling approaches. - Todd