MADM and Data Mining approaches to Performance Evaluation and Portfolio Planning for an Investment Holding Company

Introduction

In today’s competitive business situation, performance evaluation of firms is an extremely important concern of all the people who are typically stakeholders of the business game. In case of holding companies, this is a more important issue since the parent firm must permanently control the situation of its subsidiaries in their sectors to make appropriate investment decisions.

In 2017 a project named “Performance Measurement and Portfolio Analytics Platform for Holding Companies” was defined for a big Holding in Iran which manage more than 100 subsidiaries. As director of this project, and based on the project plan I proposed to develop Multi Attribute Decision Making (MADM) approaches for evaluating performance of firms considering financial and productivity criteria.

The main steps of this project include:

• Review and modify organizational processes in activities related to performance evaluation and portfolio management
• Designing solutions and tools for monitoring and evaluating performance
• Designing solutions and tools for planning of portfolio and subsidiaries Project Risk Management
• Designing solutions and tools for corporate governance
• Designing computer-based managerial dashboard

 The methods used in this project include TOPSIS, Promethee, Data Mining, Monte Carlo Simulation, Scenario Analysis, Strategic Portfolio Planning, …

Here some part of the results and outputs of this report will be provided to share the idea and have feedback to help me for performance evaluation and portfolio planning for big Holding companies with a lot of subsidiaries.

Because of Confidentiality of information, I cannot share the name of companies and total results.

 Promethee Method for Subsidiary Ranking

Based on mathematics and sociology, the Promethee and Gaia method was developed at the beginning of the 1980s and has been extensively studied and refined since then. It has particular application in decision making, and is used around the world in a wide variety of decision scenarios, in fields such as business, governmental institutions, transportation, healthcare and education.

Promethee_I is a partial ranking of the actions. It is based on the positive and negative flows. It includes preferences, indifferences and incomparabilities (partial pre order).

Promethee II is a complete ranking of the actions. It is based on the multicriteria net flow. It includes preferences and indifferences (pre order).

Ranking Companies

Let us suppose that one holding company wants to evaluate performance of its subsidiaries. We suppose that the decision-maker has identified 23 criteria (such as financial ratios, growths, and some qualitative criteria about management) as important for performance evaluation. Visual PROMETHEE can handle quantitative as well as qualitative criteria.

It is possible to un-check some box and the corresponding criterion or companies will be removed from the analysis and It is thus very easy to make what-if analysis and to see the impact of one company or criterion on the results of the analysis.

 There are two PROMETHEE rankings that are computed:

• The PROMETHEE I Partial Ranking is based on the computation of two preference flows (Phi+ and Phi-). It allows for incomparability between companies when both Phi+ and Phi- preference flows give conflicting rankings.
• The PROMETHEE II Complete Ranking is based on the net preference flow (Phi).

 On the Partial Ranking tab (left figure), the leftmost bar shows the ranking of the companies according to Phi+: Company 2 is on top, followed by companies 1, 6, 5, 3, and 4. The rightmost bar shows the ranking according to Phi-.

We can conclude that Company 2 is preferred to all the other companies. Company 3 is incomparable with 4 because it has a worse score on Phi+ and a better one on Phi-. This is confirmed by the complete ranking (right figure). Company 3 and 4 have also very close but negative scores. They are at the bottom of the complete ranking. While the complete ranking is easier to explain it is also less informative as the differences between Phi+ and Phi- scores are not visible anymore. Incomparability in the partial ranking is interesting because it emphasizes companies that are difficult to compare and thus helps the decision-maker to focus on these difficult cases.

An advantage of the PROMETHEE Diamond is that it is easy to visualize the proximity between Phi+ and Phi- scores globally. For example, as the company 2 cone overlaps all the other ones this company is preferred to all the other ones in the partial ranking.

 The PROMETHEE Network representation of the Partial Ranking will feel familiar to the users of older PROMETHEE software such as PromCalc or Decision Lab. Here the companies are represented by nodes and arrows are drawn to indicate preferences. Incomparabilities are thus very easy to detect.

In PROMETHEE Rainbow figure, for all companies we can see negative and positive slices as some criteria contribute positively and some negatively to its net flow score. This means that all companies have some weaknesses in some criteria with respect to the other companies.

GAIA analysis starts from a multidimensional representation of the decision problem with as many dimensions as the number of criteria. A mathematical method called the Principal Components Analysis is used to reduce the number of dimensions while minimizing the loss of information.

Here three dimensions are computed:

• U is the first principal component, it contains the maximum possible quantity of information,
• V is the second principal component, providing the maximum additional information orthogonal to U,
• W is the third principal component, providing the maximum additional information orthogonal to both U and V.

The standard GAIA analysis includes U and V only, as in older software such as PromCalc, Decision Lab, D-Sight or Smart Picker Pro. Visual PROMETHEE adds a third W component to improve the analysis when the (U, V) representation quality is too low. In practice the 2D GAIA analysis is reliable when the quality level is above or close to 70%. Each company is represented by a point in the GAIA plane. Its position is related to its evaluations on the set of criteria in such a way that companies with similar profiles will be closer to each other. Company 1 and 2 are very close to each other. They are quite similar companies. The Company 6 is also very different from the other companies.

Each criterion is represented by an axis drawn from the center of the GAIA plane. The orientation of these axes is important as they indicate how closely the criteria are related to each other:

• Criteria expressing similar preferences have axes that are close to each other.
• Conflicting criteria have axes that are pointing in opposite directions.

It is thus possible to identify groups of criteria expressing similar preferences and to better understand the conflicts that have to be solved in order to make a decision.

The Walking Weights window allows you to change the weights of the criteria and see the impact on the analysis. The window is split into two parts: The upper part is a bar chart showing the Complete Ranking. The lower part is a bar chart showing the weights of the criteria.

A more precise and thorough weight stability analysis can be done using the Visual Stability Intervals. The horizontal axis is the weight of the criterion from 0% to 100% and the vertical axis is the PROMETHEE net flow.

By defining different data sets, we can use different weights to compare the results of many scenarios as a comprehensive analysis. The Balance of Power window is similar to the Walking Weights but allows for changing the weights of the scenarios.

Clustering Companies for better evaluation and monitoring portfolio

K-means clustering is a type of unsupervised learning, which is used when you have unlabeled data (i.e., data without defined categories or groups). The goal of this algorithm is to find groups in the data, with the number of groups represented by the variable K. The algorithm works iteratively to assign each data point to one of K groups based on the features that are provided. Data points are clustered based on feature similarity. The results of the K-means clustering algorithm are:

• The centroids of the K clusters, which can be used to label new data
• Labels for the training data (each data point is assigned to a single cluster)

After clustering 20 different companies in 4 cluster, we can see most of the companies have similar characteristics in terms of performance evaluation factors.

 Dashboard

A data dashboard is a visual display of the most important information needed to achieve one or more objectives, with the data consolidated and arranged on a single screen so the information can be monitored at a glance.

BI tools like PowerBI come with their own tools to pull and view data from a SQL database. These same tools are used aggregate SQL database data with other data sources to build metrics and KPIs, and to visualize and assemble the metrics and KPIs to create dashboards.

Future Directions

This is just a brief report of the project for Performance Evaluation and Portfolio Planning for investment holding company.

Investment holding company is a company that earns income from the payment of dividends, rent or interest. These companies do not produce goods or offer services itself, and instead acts as a holding company by owning shares of other companies.

In the following steps of this project, I intend to develop the use of quantitative methods and tools in performance evaluation and portfolio planning for investment holding companies.

By sharing this post here, I look forward to having your feedback and suggestion any other methods that help me to use and improve the ability of evaluating, monitoring and managing the portfolio of investment holding company which are not listed in Stock Markets.

 Moreover, I am interested in collaborating on similar projects for companies.