If you go into Google youÂ can find most things remarkably quickly amongst the vastness of the internet, so why can’t you find yourÂ sales data inside a large company?Â This perfectly reasonable question has prompted some BI vendors to team up with Google in order to put a Google search front end onto enterprise data.Â Sound too good to be true?Â Sadly I fear that it is.Â The search capabilities of Google are superb at searching for keywords on websites, and enable you to quickly zero in on what you are looking for provided you can make your search keywords precise enough.Â Unfortunately the same technique does not translate well to the semantic nuances of enterprise data, where finding a database with “price” data in it unfortunately does not give you sufficient context (which for which product, under which commission scheme,Â on what date, within which sales area, etc?).Â Moreover a search engine does not yetÂ generate the SQL to get the wretched data out of the corporate databases where the answers lie.Â Hence to put a Google front end on to a BI tool you are probably going to have to run a bunch of reports, give them some tags and publish them as web pages – Google will certainly be able to deal with that, but then is this so much better than just picking the report you want out from a list anyway?
AndrewÂ Binstock writes a useful article about this in InfoworldÂ but perhaps glosses over the magnitude of the problem in terms of finding answers to data on an ad hoc basis.Â Indeed early implementations essentially throw the problem back to a BI tool, which generates results in a form that a search front end like Google can use.Â Usually this is not the biggest problem anyway, as it easy enough to put menus together with the top 20 or whatever regular reports for users to choose from.Â I can see a real use for this when the sheer number of canned reports gets out of hand though.Â If you have thousands of reports to trawl through then having a search front end could be genuinely useful.Â But the lack of semantic understanding needed of enterprise definitions will make it just as hard for a search tool to make any sense of a mass of numbers as a BI tool, which relies on either some from of front end semantic layer (as Business Objects uses) or assumes the existence of a data warehouse where the semantic complexity has been pre-resolved into a single consistent form.Â As the article correctly points out, the only way to fix this is through better metadata.Â Indeed, greatly improved master data definitions could find a further use as tags to help search engine front-ends make more sense of large numbers of pre-built corporate reports. Unfortunately the nirvana of ad hoc access to corporate data viaÂ an intuitive search front-end seems to me no closer than before.Â
What is certainly true is that the BI vendors can use these Google front-ends to make pretty demos to try and sell more software.Â However they do run a hidden danger in doing so.Â Given that at present the Bi vendors compete partly through the ease of use of their graphical interfaces, by handing over the user interface to Google they may be in danger of commoditising part of their competitive advantage.Â If you have a simple search front-end, who knows whether the report originally came from Business Objects, Cognos, or Information Builders?Â I wonder whether the BI vendors have really thought through the danger to their own businesses that this seemingly innocent search front end could become.Â By jumping on the Google bandwagon they could be unleashing something that removes their direct contact from the end user, a key element in differentiation.