When Yossi Caspi was head of data warehousing operations at drinks maker Coca-Cola in Israel , he had the kind of problem that is all too familiar to business intelligence (BI) professionals.

Simply to generate the required weekly sales reports, Coca-Cola’s data management systems had to churn for the best part of 18 hours each weekend. If anything went

wrong during that batch run, the vital reports would be absent from decision-makers’ desks on Monday morning – and Caspi would shoulder the blame.

The problem was that the relational database underlying the BI system, like most such standard databases, was particularly poor at data aggregation, a task that involved pulling together data from every corner of the company.

But then Caspi met Guy Shaked, the vice president of research and development at a struggling parallel computer developer. Together the two sketched out a technology that would solve the problem and, shortly afterwards, quit their jobs to found HyperRoll.

HyperRoll’s technology tackles the problem by eliminating the need to compose or maintain summary tables – a key part of building BI query databases. Instead, its real-time engine creates summary tables on the fly by using several proprietary mathematical algorithms that aggregate data much faster than any relational database, says HyperRoll CEO Rani Cohen.

He claims that HyperRoll’s technology can speed up complex querying of data by a hundred times or more as a result. Implementation is similarly fast – just a couple of days, says Cohen. He points to Swedish airline SAS where the financial consolidation process was reduced from nine hours to three minutes, after an installation lasting just one day.

Other customers – including Kellogg Co, Mastercard and Comdisco – were also impressed. But to enter the BI mainstream, HyperRoll has to ensure it stays well ahead of the major database vendors as they respond to the kind of requirement that plagued Coca Cola.