Anyone who’s ever been tasked with designing and implementing a large-scale business intelligence (BI) deployment knows there are many moving parts.
While the final result may meet all of an organization’s BI and analytical needs at the moment, requirements change over time, and complex BI software often lacks the flexibility to adjust. To help meet changing conditions, Evelson recommends that organizations look to metadata-generated BI applications.
Metadata-generated BI applications let developers define the metadata or, in this case, the business rules that govern the 30 components at a “higher, semantic level,” Evelson explained. Instead of manually and separately creating all 30 components, like data integration and data quality jobs, these applications automatically generate them once the rules or metadata are defined, greatly reducing the effort needed to make changes to the BI system.
The problem, Evelson said, is that even “a tiny change at the source system may explode and mushroom into hundreds of changes in my BI application.” With conventional BI applications, that could require developers and BI workers to manually make the changes to each of the apps’ many components. Metadata-generated BI apps solve this problem.
Take an analytical CRM application that analyzes customer behavior, for example. If a change is added to the source CRM system – say, adding a field for customers’ age so the sales rep can track trends by age group – then numerous changes are needed to accommodate the analytical app.
“To make sure my analytical CRM app encompasses this new change, I have to take a look at my data sourcing routines, my data integration routines, my data cleansing routines, my data warehouse model, all of my metrics and aggregations,” Evelson said. “There may be hundreds of reports and dashboards affected by that one tiny change.”
With a metadata-generated app, Evelson said, “I push a button and generate the entire set of 20 or 30 components.”
While metadata-generated BI applications are also a good choice for developing new analytical applications, he said, they allow developers to quickly build BI prototypes that can be easily rolled out to select end users for evaluation. Any changes the users request can then be easily made.
The other benefit is that a BI prototype built using metadata-generated applications can be easily put into production when the proof-of-concept is complete, Evelson said.
In addition, he recommends that organizations consider these applications to address sudden competitive threats and other unexpected analytic needs.
Of course, they come with their own drawbacks and are not ideal in all environments, Evelson warns.
“Clearly, not all BI projects and initiatives fit well into the agile [metadata-generated] BI approach,” he wrote in a recent report on the topic. “Customer billing statements, monthly income statements, balance sheet reports, and all types of regulatory reporting are probably not good candidates for agile BI.”
Also, only a handful of vendors currently offer metadata-generated BI applications and “are 100% dependent on that vendor’s technology,” creating the risk of vendor lock-in, Evelson said. And some of the vendors are rather small, so “have a contingency Plan B that you prepare for and refresh at least once a year, and be ready to implement on a moment’s notice should the unthinkable occur and your agile BI platform provider goes under.”
Evelson identified Kalido, WhereScape, Endeca Technologies and Balanced Insight as among the handful of metadata-generated BI vendors.
He also recommends using metadata-generated BI apps in conjunction with, not as a replacement for, traditional BI applications.
“Use both approaches in tandem and purposefully adopt initiatives that align with their strengths,” Evelson wrote. “Running strategic initiatives in parallel with tactical rapid prototypes will capture the benefits of both approaches and increase optimization.”