Quantrix Modeling Methodology

"Uncontrolled and untested spreadsheet models pose significant business risks. These risks include: lost revenue and profits; mispricing and poor decision making due to prevalent but undetected errors; fraud due to malicious tampering; and difficulties in demonstrating fiduciary and regulatory compliance.

"These risks are ignored due to a widespread failure to inventory (keep records of), test, document, backup, archive and control the legions of spreadsheets that support critical corporate infrastructure."

        - European Spreadsheet Risks Interest Group (EuSpRIG)

Overview

The Quantrix Modeling Methodology (QMM) provides a systematic approach to creating models with Quantrix Modeler.  Regardless of skill level or experience with Quantrix, business modelers should consider using a quality process as it will save time and effort when designing models and often increase the utility and impact of models.  The process begins with a simple question and progresses to the new business understanding or insight.  QMM also ensures on-going understanding of the current and forward-looking business environment through the data updating loop (DataLink).

History

QMM traces its origin to a proven standard process model.  In 1996, an initial committee with representatives from DaimlerChrysler of Germany, SPSS Inc of the US, and NCR Systems Engineering of Denmark began to create a structure to facilitate the immature data mining market.  The committee received funding from the European Commission to further their research and activities and sought input from data warehouse vendors and management consultancies. The committee created the Cross-Industry Standard Process (CRISP) which was intended to be an industry, tool and application-neutral structure to build low-defect data models.  When CRISP 1.0 was released, there were nearly 300 organizations involved in the CRISP special interest group.   Quantrix has employed this structure as a starting point for creating a modeling methodology that conforms with the Quantrix Modeler paradigm.

The Methodology

The Quantrix Methodology diagram (above) is further explained by the following general descriptions:

Business Understanding (A)
This initial phase focuses on understanding the business objectives and questions, and then converting this objective/question into a data modeling problem definition. The result is a preliminary plan designed to achieve the objectives.

Data Review (B)
The data understanding phase starts with getting familiar with the data sources, identifying data quality issues, initial data collection, and then discovering first insights, and interesting subsets, to form hypotheses regarding the desired business understanding.  The result is identification of the categories and items within the data.

Develop Main Logic (C)
In this step, the key formulas that drive your model are created.  This may include formulas for summary across a time dimension, key algorithms or expressions, and other significant “housekeeping” requirements.  A modeler typically will limit the number of items during the main logic development in order to make testing easier (see Step D).

Testing Main Logic with Sample Data (D)
Through the use of simple inputs or a small sub-set of the data, the modeler will test the formulas to ensure correct performance.

Refine and Debug Logic (E)
Errors recorded in Step D are corrected. The items that were not initially created during Step C, are now added.  In addition, improvements to the logic can be identified and implemented.

Importing Entire Data Set (F)
Using DataLink™, the entire data set is imported and maintained. Once the model has been validated, future imports of the entire data set can be performed repeatedly for continuous business understanding (See Step H).

Review, Validate and Create Presentations (G)
A validation function is created within the model for large data sets to ensure that the data is being properly manipulated.  Create reports and presentation items that elucidate the model results.  These reports or output items are compared to known sources or other third party data. 

Analyze Results, Trends, Previously Hidden Relationships, and Points of Interest (H)
The modeler is now in a position to analyze the results based on the outputs from Step G .  These results provide insights into trends, previously hidden relationships and other points of interest. 

Answering the Question/Achieving the Business Understanding
The analysis answers the business question and achieves the business understanding.  The equation below illustrates the model combined with analysis supports the decision-making process.

        Model + Analysis ⇒ Decision

Careful analysis of the model outputs completes the decision equation.

How to Learn More

For more information about the Quantrix Modeling Methodology or to inquire about how this quality process can be implemented in your organization, please contact Professional Services.

---