I will use Bill Graziano, Ken Henderson, and Fernando Guerrero as my examples for this:
Pick a good topic, cover it thoroughly, and try to provide more than what's expected (Graz, Ken, Fernando)
Create good slides that are descriptive enough so that notes aren't required (Graz, Ken, Fernando)
Provide handouts anyway, in case you want to take notes (Graz, Ken)
Allow for a Q&A session (Ken), or allow attendees to ask as you present (Graz). Reward questions with food (Graz). :)
Complement, but don't repeat verbatim, your previously published material (Graz, Ken, Fernando)
Practice speaking clearly and authoritatively on the topic (Graz, Ken, Fernando)
DEMONSTRATE, don't just...
I just got out of a presentation on “Performance Tuning and XML Support in MS SQL Server 2005”. Here are some high/low-lights:
Zero discussion on performance tuning, except for “use showplan, it's great”. Don't describe anything new about showplan either
List various new and improved locking, query, and join hints, and....don't discuss them in any detail
Provide almost no demos. And then, they're demos that don't demonstrate anything; they're just one line queries
Finally got into some depth on XML features, but don't describe how they work or what's new about them. I think I now know LESS about XML support in SQL2005 than I did before...
Fabian Pascal responds to an article on Slashdot about XML and Relational Databases. Slashdot discussion and feedback on it is especially entertaining. More kind words from Fabian add to the verbal melee.