3: The BI Front-End
The previous chapter discussed the back-end systems that comprise the architecture of a BI system; the present one will investigate the front-end systems that enable users to interact and extract meaningful information from the system. The author stresses that both aspects must be done "right" in order for a BI system to deliver its potential value.
Business Query and Reporting
The informational tools provided to the user can be categorized as reports (a standardized analysis of the data that is periodically generated) and queries (an on-demand analysis to suit specific needs not served by the reports). The latter type enable business users to investigate and discover new information, and if that information is useful to have on an ongoing basis, it can become an established report, though in some instances, the query is for a one-off question that is unlikely to be posed again.
There are established formats for reports that have changed little over time, such as sales reports, operational overviews, and financial reporting. Another form of report that has become fashionable in recent years is the "dashboard", which brings together a significant amount of data for an at-a-glance analysis.
Ad-hoc queries are essential. Given the pace of change, it is not sufficient merely to have access to a fixed set of information, nor is it expedient for a business user to need to engage IT each time they wish to retrieve information from the system. A critical quality of the user interface is to provide business users with the ability to access data in familiar terms, without having to learn programming languages such as SQL, to perform analysis. If the tools are complex or designed in a system-centric manner, they will not be used and the BI system will fail to deliver value to the organization.
(EN: The author focuses on design, and does not mention another critical element: availability. This goes back to an earlier comment, in which senior managers are fearful of providing access to information for fear of losing power, and restrict access to preserve their personal control, a practice which is counterproductive to organizational effectiveness and efficiency.)
Production Reporting
Production reporting is often considered separately from other forms of business reporting because it is generally more granular than other reports: it must reflect the current state, in a high level of detail, and in real time.
It is also considered a highly sensitive area, in that it contains privileged information (the kinds of details that might appear in an invoice, bank statement, check or list of open orders) and there is the potential for a user or system to make a mistake that can have a financial impact on the organization, its vendors, and its customers. (EN: since BI provides read-only access to data, I don't see the latter as a valid risk.)
As a result, production systems are often safeguarded and subject to tight control and procedures, which reduces their effectiveness.
Online Analytical Processing (OLAP)
OLAP systems focus on analyzing and exploring data, and the term is intended to differentiate them from systems that use data for monitoring purposes. Said another way, OLAP seeks to determine the reasons behind the figures that reporting systems merely disclose.
OLAP systems perform statistical analysis on granular data to discover correlations between the phenomena they represent, though it is common for this functionality to be built into standard reporting tools (such that the user can click a link within a report rather than having to launch a separate application to conduct an investigation).
Some of the critical qualities of an OLAP system are:
- Multidimensional - The system must analyze numerical values from multiple dimensions (product class, time, geography)
- Aggregation - The system must be able to bring together and correlate data from multiple systems or data stores
- Fast - An OLAP system must be highly responsive. It is no longer acceptable for a decision-maker to have to wait until the next business day to receive the results of a query (ideally, results should be available within seconds)
- Interactive - The user should be able to run multiple queries, modify them, and slice-and-dice the data in a variety of ways (as opposed to the fixed format of reports)
The author describes some of the platforms - ROLAP, MOLAP, DOLAP, HOLAP - which use different methods to balance the need for performance versus flexibility. (EN: Good to know the acronyms, but the details seem incidental - the business should communicate its needs and IT can determine which solution is the best fit.)
Microsoft Office
Microsoft Excel remains one of the most popular interfaces to OLAP data - the user runs a query and retrieves output data in a table that is imported into Excel for further analysis - though current systems provide a wider variety of web-based viewing tools that have enhanced analytical capabilities and the ability to display information in ways that a spreadsheet cannot (e.g., visual representations of data)
There is some conflict over the use of Excel, and it's suggested that IT often attempts to disable this access to users, because it is counterproductive to a supposed "goal" of BI systems, which is to arrive at a universal "truth" by compelling users to accept the same methods of investigation. (EN: the counterproductive nature of this agenda is self-evident and needs no further elaboration.)
Dashboards and Scorecards
A "dashboard" is an at-a-glance display of the information that is considered to be critical to a specific objective. It is characterized by graphical representation or a relatively small number of analyses, generally from disparate systems, with click-through functionality to OLAP tools for more complex analysis.
The concept of the dashboard originated in the late 1980's but has become increasingly popular of late, and increasingly flexible. The author cautions readers against being dissuaded by previous bad experiences with dashboards, in that the implementations have often been flawed, but the tool has excellent potential.
The term "scorecard" is often used interchangeably with "dashboard", but it is a slightly different concept: a dashboard merely communicates data, whereas a scorecard compares current data to benchmarks or predictions. The two are often presented in a single interface, but the author wishes to stress that they are not the same.
Performance Management
Performance management tool have historically been considered as a separate application from BI, but the two are becoming more closely integrated because their information needs are closely related. Simply stated, there is little information in gathering and analyzing data unless it is also used in making decisions to manage and improve performance.
There is another "alphabet soup" of acronyms pertaining to roughly the same thing: BPM for business performance management, CPM for corporate performance management, EPM for enterprise performance management, OPM for operational performance management, FPM for financial performance management, etc. It's all the same thing, broken into specific facets.
In a fundamental sense, performance management is accomplished by determining which metrics are most germane to the success of an organization, establishing performance standards that must be met in future to achieve a desired goal, comparing the current metrics to those projections, doing a gap analysis to determine the reasons for any variances, and taking corrective measures to bring actual performance in line with the projected outcomes.
One key difference between BI and PM is that BI is focused on the analysis of historical data whereas PM is more concerned with planning and predicting future states. Generally, plans are made based on historical data (regression analysis) impacted by the intended consequences of future actions (e.g., an advertising campaign should improve sales figures, on top of what may be expected based on the historical trends).
PM is typically consolidated: a each business unit or product line within an organization performs its own analysis and planning, and the figures are rolled up into corporate projections and metrics, the ultimate expression of which tends to take on a financial tone (a projected future balance sheet, income statement, and statement of cash flows).
Analytic Applications
In 1997, International Data Corporation coined the term "analytic application" to describe software that functions independently of operational systems, aggregates data, and automates tasks related to a particular business process.
These systems were historically developed to provide information to decision-makers, but have been evolving toward automated decision-making with the ability an authority to make executive decisions. For example, a manufacturer may utilize an analytic application that considers inventory levels and production schedules to arrive at a conclusion about the quantity and timing for ordering component materials, and actually communicate orders to vendors to obtain them as needed, rather than feeding a report to a human manager who would attend to this task.
Emerging BI Modules
The author mentions some of the BI "modules" that are currently in the works: predictive analytics, enterprise search, advanced visual applications, mash-ups, and rich internet applications "to name a few" - however, the author merely names them, and doesn't go into much detail.
(EN: Mixed feelings here - merely rattling off a list without details is of limited value, and given that some of these are not specific to BI, it undermines the author's credibility, but the point is that BI, like any other technology, is in a state of flux, and there are dramatic new things on the horizon that may or may not materialize.)