1995 start-up Broadbase Information Systems Inc, San Mateo, California is going to stand up and start shouting about its packaged Windows NT-based data mart solution it claims is cheaper and faster than equivalent Oracle Corp or other third party offerings. A former Oracle 7 decision support engineer, president and CEO Mark Kremer – who also created Verity Inc’s retrieval system in a previous role – has put the company together; Bruce Armstrong, is SVP sales and marketing. Armstrong was formerly vice president for Sybase’s data warehouse products and prior to this a Teradata executive. Broadbase’s venture backers, Benchmark Capital and Mohr Davidow Ventures are hoping the company can capitalize on International Data Corp estimates that companies can expects a return on investment for successfully implemented data warehouses of up to 400%. However, due to the complexity of these solutions, getting an enterprise warehouse up and running averages $3m, Broadbase figures, beyond the reach of mid-size companies are departments. Its product includes the Broadbase Server, a query engine optimized for decision support; an application layer providing multiple types of analysis; a dynamic connectivity/transformation layer and automated administration and management. All of the components are integrated utilize SQL92, ODBC, COM, SQL Server and Windows NT. It also supports Java, which is embedded directly into the server as the procedural language. The company claims its database implementation has been designed to maximize decision-support query performance. Query speed can be regardless of data or query type. Broadbase stores data in a column structure, since most decision-support queries are based on columns analyzing multiple records. It says this eliminates the time- and resource-consuming full-table scans that often result from using row-based structures of OLTP systems. Supposedly it results in reduced I/O and increased cache and enables Broadbase to compress data and indices decreasing required disk space and improving performance. The Broadbase server utilizes selective decompression, keeping the data compressed in memory and only decompressing the required result set. Its DSS engine is claimed to perform look-ahead indexing that automatically selects the optimal access path, eliminating the requirement for having to understand the nature of the data and/or queries. Since many queries will supposedly also require multidimensional analysis, Broadbase says it’s also offering provides dynamic cubing – the ability to create a Broadbase Cube to precalculate aggregations and improve the response time of queries which are multi-dimensional in nature. Broadbase has integrated a query and reporting tool – it doesn’t say which one. No prices.