Incremental approach is the best strategy. This is less to do with technology & infrastructure but more with the stamina in business to define and think through on what they need and why they need. This also gives the learning experience to the team, as every data warehouse project is little different. This will also test management's culture & capability to use the data warehouse information. The conventional steps are to create a set of data marts first. While creating the data marts, do try (as much as possible) to create standard set of dimensions and measures across the data-marts. After few data-marts are created and the readiness of organization improves, one can go for creating a data warehouse. This is called Bottom-Up Data Warehouse approach.

When we say Incremental Strategy, it does not indicate another extreme. There has to be some level critical mass for a data-mart. A data-warehouse initiative should analysis all the business requirements of that function. For example in sales management function, one should analysis sales revenue, sales profitability, sales campaign, sales compensation, sales channel and sales process, before short listing the themes one is focusing on. You may not immediatly build data-marts on all these areas in a function, but you should analyze them before you start your dimensional modeling. Data Mart initiative at a functional level should also not be too narrow, as it puts a significant constraints on future growth.

These are been some views of creating a large enterprise level data warehouse, and then create data-marts as per the requirements of a function. The Data Warehouse in this case is typically normalized. This is called Top down data warehouse technique. Unless we have a very stable and mature organization, we should go for Bottom-up approach. It may be worthwhile to refer to Integration Strategies for Stand-Alone BI environments.