The missing ‘Next Big Things’

This page is part of the free content of The BI Verdict, which represents a small fraction of the information available to subscribers.

By purchasing a subscription to The BI Verdict, you and your team will gain access to the most thorough single source of information available for business intelligence buyers, including:

• Detailed reviews of over 20 leading BI tools
• Factbases with feature-by-feature product scores and notes
• Customer Verdicts revealing product strengths and weaknesses
• KPI Dashboards comparing leading solutions on 27 key criteria
• Access to the key findings from the latest edition of The BI Survey
• A series of in-depth market analyses and commentaries

Click here to see the full table of contents for The BI Verdict

1.

Executive information systems

EIS was a much-hyped, forward-looking development of the mid-1980s. Sporting attractive graphical user interfaces, new-fangled mice and touch screens, EISs looked stunningly modern compared to the dull early DOS business PCs and mainframe terminals of the era. They also included colorful business graphics, merged structured and unstructured data from multiple sources, ad doc multidimensional analysis, external as well as internal data, hot-links, KPIs (key performance indicators) and CSFs (critical success factors) — all ideas that were here to stay, unlike the early EIS products themselves.

There were essentially two problems that were to kill these products off: the ambitions were beyond the technology of the day, making the products expensive, slow and hard to administer; and they fundamentally misunderstood the way that most senior executives work. Contrary to the notions of the bright techies who created the products, senior executives spend their time in meetings, reading printed reports and making presentations, not analyzing data. When they need data analyzing, they delegate the task to younger, cheaper specialists, rather than doing it themselves. These analysts need powerful tools, rather than the patronizingly simple user interfaces of an EIS supposedly aimed at the boss.

Although the EIS term has faded, many of the ideas persist in today’s dashboard products; in the The OLAP Survey 6, fewer than 20 percent of sites reported that they had an EIS application, but more than 30 percent say they use the OLAP tool to power a dashboard. Just two years earlier, EISs were ahead of dashboards.

2.

Balanced scorecard tools

Some vendors — particularly Gentia — saw a big opportunity for themselves with the balanced scorecard application of the 1990s. But what they overlooked was the full balanced scorecard described by Kaplan and Norton is much more than a software application: it is a management system which must be deeply embedded into an organization’s processes. Implementing a full-blooded balanced scorecard is not a casual or IT-driven task, and it will surely be driven by a major management consulting project. Only at the end of this protracted process is a software application deployed. A few organizations have indeed been through this full process, but many others who have deployed what they call balanced scorecards actually have a pale imitation, which is little more than a simple dashboard or EIS. The hoped-for bonanza for the software industry never happened, and vendors like Gentia who had depended on it have disappeared. Even in surveys done in the last years specialized BSC- or BI-applications were never used in more than 30% of the sites surveyed to implement a BSC. Excel is still the predominant platform for balanced scorecard applications.

3.

BI extranets

Fueled by dotcom euphoria, BI software vendors saw BI extranets as a great business opportunity for themselves. Not only could they aim to sell their customers a license for every white collar employee in the company, but also many more seats for their customers and suppliers. It wasn’t just vendors who believe this: more than 70 percent of respondents to the first edition of The OLAP Survey in 2000/1 expected to have a BI extranet in place within two years. By 2006, expectations were much lower, with only 43 percent expecting to have a BI extranet by 2008, but the actual percentage of companies with a BI extranet has remained firmly in the 15-16 percent range. Presumably too few companies saw a compelling business need to provide on-line analyses of their data to outsiders.

4.

Data mining and AI guided analysis

In the end of the 1990’s “Desktop Data Mining” was supposed to bring the merits of new algorithmic developments in the field of knowledge discovery in databases to the manager’s fingertip. Major vendors like Business Objects and Cognos proudly included products in their offering. Business Objects only licensed the data mining technology, but Cognos also acquired a company, Forethought.

But what the vendors didn’t allow for was that providing dumbed-down data mining methods to regular BI users is still too complicated and even dangerous in the results. The selection of an analysis method, its correct execution and the validation of results are important steps in all data analysis processes. The more complex the methods get, the higher the expectations toward the user’s skills. The dilemma is that data mining tools feasible for business users have to be very narrow in the methods they offer, which makes them most often useless. Proper data mining workbenches should only be used by statistically trained persons, and such people already have access to suitable tools from companies like SAS Institute and SPSS. Nevertheless, several vendors, including Hyperion, Microsoft and MicroStrategy, still provide simplified data mining linked to OLAP engines.

5.

Pre-built analytical applications replace data modeling

The success of analytical applications with pre-defined data models for specific functional areas or vertical industries is very limited. “Business Content” (SAP), “Rapid Marts” (Business Objects), “Performance Applications” (Cognos) and other predefined models often only serve as demonstration or prototyping aids but do not free the implementers from designing a proper data model. If this is forgotten the results can be disastrous: no control over objects in the application, unstructured growing data model, low performance and maintenance efficiency and in the end a major redesign project.
One reason for unmet expectations is that it is quite costly for a software vendor trying to be a business expert. For example Informatica entered the analytical applications market in the beginning of the 2000s with “Informatica Warehouse” but was quickly to discover that the required knowledge about vertical industries and functional areas to customize and update these applications stood in no comparison to the sales success. After a few years the applications business was handed over to the Informatica partners, who can bring in their special expertise and develop data models and data extractions routines on the platform provided by the software vendor - a better business model where each partner concentrates on what he does best. Most other BI vendors now recognize that vertical expertise should come from partners, but several continue to market horizontal applications. Apart from the long-established planning, budgeting and consolidation applications, this is usually with limited success.

6.

Database-embedded OLAP

When the major database vendors, starting with Oracle in 1995, began invading the OLAP industry, it seemed only a matter of time before they embedded OLAP directly into their relational database engines. Indeed, Oracle promised nothing less. By taking advantage of the advanced technology of the RDBMS engine, the theory was that the database vendors would be able to deliver products that were more scalable, faster and better integrated with other applications. It was assumed that this would squeeze the independent OLAP vendors.

Logical and apparently inevitable it may have seemed, but it didn’t happen. Oracle came closest to doing it, but the Oracle OLAP Option still uses a version of the venerable Express engine. This does store its multidimensional data in the relational database, but as a MOLAP, stored as a set of blobs (which cannot therefore be queried by the relational engine). In effect, the former Express multidimensional database is simply stored in Oracle tables, but not in a useful form. Although SQL can be used to query the OLAP engine, this actually happens via an internal conversion to multidimensional queries based on Express. Microsoft bundles Analysis Services with SQL Server, but it doesn’t even pretend that the two are integrated at the product level. Despite years of rumors, IBM never did acquire Arbor Software (or Hyperion Solutions as it became), and it even dropped its reseller rights for Essbase, which had been marketed as DB2 OLAP Server. Indeed, it even dropped its ROLAP version of this product, and in the later stages of its OEM relationship, simply distributed Hyperion Essbase, which remained robustly MOLAP with no specific DB2 integration. It tried a different approach to adding OLAP into the database with DB2 Cube Views, but this has had very little market acceptance. Teradata sold its TeraCube product to MicroStrategy, which discontinued it, while CA ‘accidentally’ acquired two ROLAP products which were never integrated with its databases.

7.

BI Portals

Most of the multi-product BI vendors have grown through acquisition. As a result, their products have often had a less-than-integrated look and feel. One popular way of dealing with it has been to build a their own portals through which the various applications and documents can be accessed in a relatively consistent way. This was typically first implemented when portals were a hot topic, so the BI vendors duly promoted their integration layers as portals. But this turned out not to be a good idea: customers who wanted portals were more likely to opt for one from a portal specialist or a major vendor such as SAP, IBM or Microsoft. The BI vendors then re-labeled their products as BI portals, emphasizing that these were not meant to compete with industry-standard portals. But even this re-positioning was not successful, so it is now rare to find BI vendors claiming to offer any flavor of portal.

8.

Natural language query

Natural language query is another once-promising idea that has yet to deliver. Even in the early 1980s, it was possible to query IBM mainframe databases in English using a product called Intellect from AICorp, but the results were too unreliable and unpredictable to be useful for general BI purposes. Indeed, once users discovered a set of queries that did work, they were often embedded in a set of menus. Unfortunately, languages like English just don’t have the precision of a proper query language, so even the hugely increased power of today’s computers still doesn’t deliver reliable results. For example, Microsoft introduced English Query in SQL Server 6.5, enhanced it in the 7 and 2000 releases — and then quietly dropped it in the 2005 edition.

Now, BI vendors are trying to take advantage of Web search tools like Google, but the idea now is just to locate reports, not to support a universal query capability. It cannot be long until the next wave of trials will start on this new platform.

9.

New visualization methods

Business graphics are more colorful, easier to produce and faster than ever, but the popular basic chart types haven’t changed in decades. Numerous new visualization techniques have been introduced, but none have been widely adopted, despite the apparent need for them in business intelligence applications with large data volumes to report. One problem seems to be that the more advanced charts are not immediately intuitive for the reader; if users need special training to understand new chart types, they are not likely to become mainstream.
Among many failed ideas for improved business graphics were dimensional animation. The idea was to smoothly step through members of a dimensions, with each frame in an animation representing a different slicer dimension member. The idea was that changes and trends would be highlighted in this way, but the idea never caught on.

10.

CRM analytics

When CRM was hot, at the turn of the 21st century, BI vendors saw an opportunity to sell their products for “CRM analytics”. Several vendors produced pre-built CRM analysis applications, and applications to analyze Web logs were expected to be very popular. They weren’t: in The OLAP Survey 6, only 1.7 percent of respondents said that their organizations had BI applications used for clickstream analysis, down from an already low 4.8 percent three years earlier. Only 11.5 percent said that they had any form of CRM OLAP application, down from 17.9 percent four years earlier.

11.

Real-time BI

From time to time, BI and ETL tools vendors have promoted the idea of real-time BI. In some cases, they actually mean real-time in the accepted sense of the term (a sub-second delay between transactions being accepted and being incorporated in BI reports), but in other cases they either mean controlled latency or even something entirely different, such as fast reprocessing of planning cubes. But there is little evidence of demand for genuine real-time BI. Yes, some operational applications do need real-time reporting, but this isn’t business intelligence in the accepted sense, which are more broadly based. Normal BI applications are expected to have consistent data — while most users do want ‘up-to-date’ reporting, few users would welcome reports where the numbers changed from moment to moment, and there is little business need for it. And genuine real-time operational reporting is unlikely to be based on normal BI tools, which are not designed for integration with message buses for capturing the very latest transactions, and nor are they fast enough.

12.

BI for compliance

The introduction of strict new accounting regulations in both the US and Europe seemed like yet another lucrative opportunity for BI vendors. They reckoned that the requirement to comply with Sarbanes-Oxley, IFRS and Basel II would force their customers to implement expensive new reporting and analysis systems. While it is true that these regulations have indeed proved expensive, most of the expenditure seems to have been spent on audit consulting and specialized transactional applications, not BI and CPM projects.