Say you have a table that looks like:

lotsodata(client int not null, recorddate datetime, usefulgrouping varchar(20), datum1 int, datum2 int, datum3 char(10), datum4 int);

Normally we search this table by: client, recorddate, usefulgrouping

That would lead us to a logical clustered index of the same.

For different reasons, we sometimes look up data like: client, recorddate, datum2

What would be the best way to index for that situation? (running SQL2K if it makes a big difference)

Take this question to the SQL Server forum. They can help you specifically with these indexing issues in SQL2K.