It's Thursday. The CEO wants a plan to boost revenues by 5 percent by the end of the week, and your retail manager hasn't got a clue.
He logs into the wiki to check in with his team, and reads a post from a marketing VP: "Why are all the southerners dissing the product?" On a Google map, red flags are clustered in the south-east. As he mouses over them, windows pop up with snippets of text pulled from the company support forums and call center application.
The retail manager pages down to see a post by the support manager. "Looks like they're all complaining about problems with the 2000 model," he says. It's a bad start to the day but thanks to the advanced business intelligence (BI) interface that the IT team built, the retail manager begins to see a solution...
This is but one example of the type of BI we could see in the future. Experts hope that thanks to a mixture of Web 2.0 technologies, advances in hardware and new software algorithms, the discipline could be due for a renaissance.
One of the crucial aspects of BI that has to change is at the end-user interface, according to Gartner research VP Nigel Rayner. "To get insights from an interesting data pattern, you need more than reporting, querying and the traditional backwards-looking BI view," he said. "That's why we see these analytic applications emerging that give people a new way to model things."
Mashups--combining and enhancing data streams with Web applications--will revolutionize BI interfaces, predicted Mark Whitehorn, founder of BI consulting firm Penguinsoft. "BI might say that a cluster of data is interesting but you can't visualize that unless you write a whole application for doing it," he said. "But if you send that to a mapping system, you can suddenly see why the BI software found that cluster interesting."
Combining mashups with collaborative technologies could make BI even more powerful. One person rarely makes decisions based on business intelligence data, said Richard Neale, director of CIO marketing at SAP. "Providing facilities for people to collaborate on the decision-making process will be an important feature of the BI environment," he added.
This implies that the collaborators know what questions they want to ask the database, pointed out Boris Evelson, principal analyst at Forrester. But, he argued, this isn't always the case.
He offered an example of an employee wanting to find information about a single company, and just entering that firm's name into his BI app. "Wouldn't it be great if it could return the aggregations of all the dimensions of data that you have about that company?"
The system could then help the user whittle away information about the company until you had what you wanted. It might offer you subsets of information about the company as a client or a supplier, for example, and enable you to drill down still further. Forrester calls this concept "guided search".
In addition to innovations in BI interfaces, there are some coming advances in the underlying analytic engines that crunch the data. Rob Rose, who directs the office of strategy in BI and performance management at IBM, says that analyzing unstructured information will be a huge boon for BI users.
"You just know that customer satisfaction information isn't in the customer database. It's probably in a call center application, in a comments field somewhere, and the call center personnel used all kinds of creative wording and abbreviations to capture those comments," he said. "Being able to parse lots of unstructured content and structure it according to 'buckets of relevance' is an important step to be able to then group, analyze, filter, explore and trend this information."
Businesses would find it even more effective if this and other searches can be done directly in memory rather than by thrashing hard drives, advised Forrester's Evelson. With 64-bit computing, it becomes possible to address more memory, enabling companies to put larger data sets directly into RAM.
This has an obvious speed advantage but that isn't all, he argued. "There are some technologies out there that aren't multi-dimensional models in memory. They are just cross-referenced indices where everything is cross-referenced to everything else." This eliminates the need to build a data model, he said. Instead, it allows companies to just load flat file data into memory and begin exploring.
How would all these technologies help our harried retail manager in our opening tale?
He might type the name of the much-maligned 2000 model into his BI app and use guided search to gradually refine the details until he found that the products shipped to south-eastern stores were all made by one sub-contractor.
Using in-memory analytics, he might pull in supplier data and collaborate with the supply chain manager to see how much that supplier was charging compared to others, and how much money they could claw back from them due to poorly executed work. The supply chain, retail and marketing managers could collaborate to see how much a rebate and a discount upgrade to the 3000 model would cost, and suddenly create a campaign that turned a problem into an opportunity for extra revenue.
All of this could happen. Some of the technologies are already there. But unless businesspeople and IT work together to create best practices for the use of BI, sexy technology will get us nowhere, noted Gartner's Rayner. "There are some new technologies that will have an impact but part of the problem is also making better use of the old ones," he said.
Danny Bradbury is a freelance technology writer. This article first appeared on ZDNet Asia's sister site, Silicon.com.
Play video
There are currently no comments for this post.