Put your SSIS packages in source control.
We have found that using a datasource of select field1, field2 from mytable is faster than using the option to select all records from the table. Made a huge difference in one of the package we did where we preprocessed data into a staging table for an export.
Build a standard import package and a standard export package that already has the variables you commonly need and the connections you commonly need and error handling you commonly use. It is much faster to build a package by opening a template package and saving under the new name, then adjusting for particular circumstances than it is to build everything from scratch each time.
A couple of things not specific to SSIS, if you send an export to another company, make sure to copy the file to an archive location as well as the ftp site. This will help tremendously when there is a question about what was sent when. Never import data from another source without looking at whether the data needs to be cleansed. An address missing the city and state is useless, an email that says "Talk to his secretary" is also useless, don't add this stuff to your database. Don't try to add records that will fail insert. If first name is a required field in your database and the import file doesn't havea value for one of the records, either add a default value ("Unknown") or don't try to insert the record.
Log errors and steps processed. This will help tremendously in figuring out where something went wrong when that package that has run flawlessly for six months suddenly fails.
Use variables and configurations to change things at run time. This lets you have test locations and prod locations, so you aren't putting the test file you don't yet want the client to see on his ftp site.
Use the trash destination (you can download it here - http://www.sqlis.com/post/Trash-Destination-Adapter.aspx) as your destination until you have tested enough to know that what you are sending to the destination is what you intended to send. In conjunction with this, use the Data Viewer to see what you data looks like at various parts of the data flow.
