Decision support systems

Posted: November 2006 C: Business Intelligence Posted by: K. Panayotakis

A simple analysis of the functionality of a decision support system, is presented.

In an effort to analyze the performance of a Business, and develop strategy for the future positioning in a competitive environment, the dimension of time is crucial (this is why all data warehouse systems have a time dimension and maintain historical data).

In this article series, we describe three functionality categories of decision support systems: 1. Detailed capture of performance 2. Performance analysis 3. Modeling and prediction

Categories are presented in an increasing complexity and business value sequence. They do not necessarily represent maturity levels, since each category is developed and matured as it is enriched with new capabilities and adopted to the dynamically changing informational and analytical Business needs.

DSS systems may be developed for other purposes, apart from performance management. A very common example is the development of a customer intelligence system, in which the aim is to build customer insight in order to adopt products and services accordingly.

Detailed capture of business performance

A basic management principle is the following: ‘you cannot manage what you cannot quantify and measure’.

Within this framework, each Business aims to develop infrastructures which shall enable it, to capture detailed measurements of the current business process performance and gradually build detailed historical performance data. ‘How successful the Business was’ for a given time interval, is captured. Business performance is captured in a detailed mode for all ‘dimensions’ involved (indicatively): Business processes and activities, Products marketed, Customer touch points employed, Geography, Customer segments, Business Units involved.

Data analysis that takes place at the ‘business performance capture’ level, is usually predefined. Predefined Key performance indicators (KPIs) are regularly calculated and presented in business intelligence dashboards. Detailed facts are produced in a multidimensional data structure (also called data mart), like: o Detailed financial facts (revenue, expenses) o Detailed production facts (e.g. production activity performance) o Customer facts (e.g. number of contacts, number of complaints) o Customer churn data o Product quality facts o New product introduction facts, and presented in reports. A static performance capture takes place.

Even though many on-line transaction processing (OLTP) systems enable the direct production of reports, this approach does not assure: o a sufficient level of detail in the facts captured o the ability to perform time series analysis (no historical data available) o the option to evaluate data in a differentiated view, by ‘highlighting’ certain dimensions of the business process and hiding others o the integrity, clarity and accuracy of facts and the single version of truth o the ability to produce complex reports, without overloading the operational system.

The set of KPIs and reports, is expected to be enriched as new informational needs are identified and new reporting capability is developed (either based on the operational systems, or on external data sources). The development of a DSS system should follow an iterative method. The data model implemented should allow the gradual expansion of the DSS.

Business performance analysis

In part 1, we described the approach to capture detailed information on ‘how successful the business was’ for a given time interval. The next step is to analyze the performance “why has the business been successful” or in lower level of detail “why didn’t product A sell as expected”. Factors which have contributed to the given business results are analyzed.

Some analysis methods may be predetermined. However, as analysis proceeds, a need to apply ad-hoc analysis may arise.

The availability of detailed performance facts, captured in their full ‘dimensionality’ (e.g. time, geography, product, Customer or Customer group, Business unit involved), upgrades the analytical capability and allows complex fact processing in order to analyse complicates issues.

Analysis may involve specific issues, like: o Why is revenue on a specific product or customer group, below expectations ? o Profitability of a specific product or service – profitability comparison of various products or customer groups o Why productivity of a specific product is lower than planned o Analysis on the Customer profile to build customer insight o Analysis on Customer defection characteristics

o Analysis on production quality on a specific product o Analysis on a product introduction o Analysis on a marketing campaign execution result

Analysis may help identify deviations from expected measurement values and alert the Business. For example: o Reduction in the consumption of services by a Customer, may be an indication that this Customer is about to defect. o Revenue reduction of a Branch unit, may lead to actionable decisions

Analysis needs change dynamically, as the internal and external business environment change. The captured data level of detail, should strike a balance between allowing in depth analysis and keeping overhead cost related to data capture under control. The exclusive use of predefined reports or queries, is not the best practice.

A decision support system can contribute to the identification of trends, can enrich the view of facts (with richer dimensional data) and support the development of innovative strategy. The capability evolution, from simple capture of facts to the ability to analyze them in order to identify trends, is of particular value to the Business.

Modeling & forecasting

In parts 1 and 2 we described two levels of Decisions support systems (DSS) systems, which analyze the “what and the why” of business performance. In a third level of DSS functionality which is more advanced, analysis enables modeling of mechanisms, to predict the outcome of one action or alternative actions: ‘What shall be the outcome if a certain measure is taken’ (what-if analysis).

Predictions may relate to any of the following (indicatively): • How much shall revenue be increased, due to a marketing campaign • Which products must be promoted to which Customer groups – ‘propensity to buy’ scoring • Which Customers are about to defect (churn or attrition modelling) • Which tax payers are likely to evade tax (tax evasion risk scoring) o How would a product price-change affect sales volume (price elasticity analysis) o How would the application of a new tax policy affect state revenues

Certain risk management forms (e.g. credit risk modeling), belong to this DSS level.

Statistical analysis, data mining and modeling tools are usually employed. The development of an accurate predictive system which is information driven, leads to a sustainable competitive advantage which differentiates leading Organizations. While the ability to capture detailed results (level 1) is the background, the abilities to analyze and predict results (levels 2 and 3) enable the full exploitation of business information and the high competitiveness of an Organization.

DSS infrastructures, if developed and used aligned to the Strategy, contribute to the formation of an Organization which fully exploits information and takes actionable decisions based on the latter.