It's horses for courses really! Using native SAS tables is likely to be more efficient just because there's nothing in between SAS and the data. But it depends what your data sources are. We have quite a few in separate databases (SQL Server/Oracle/PostgreSQL) but most users extract from them and save the results locally for analysis. This has the added advantage that, as the sources are operational databases, the source of a specific analysis is always available as a snapshot of the data.
Having said that, our analysts are moving towards having a separate platform for hosting all their analytical data, using SQL Server and Sharepoint. This has metadata built in to its front end, so that it is searchable on different topics. It means the analysts aren't restricted to any single analytical tool. SAS is the main one but we have an increasing number of R users, and there have always been a few people who prefer other platforms such as SPSS, Stata and so on. It's also become quite important because there are a wide range of areas for analysis (eg. health, education, transport, justice), which sometimes have cross-overs between them.
... View more