Multidimensional Perceptual Map for Project Prioritization and Selection - 2014 update

Multidimensional Perceptual Map for
Project Prioritization and Selection
Jack Zheng
Vijay Vaishnavi
2014 Update
Originally Presented at AMCIS 2009
Citation:
• Zheng, Guangzhi and Vaishnavi, Vijay (2011) "A Multidimensional Perceptual Map Approach to Project
Prioritization and Selection," AIS Transactions on Human-Computer Interaction (3) 2, pp. 82-103,
https://www.researchgate.net/publication/264935893_A_Multidimensional_Perceptual_Map_Approach_to_Project_Prioriti
zation_and_Selection
http://www.slideshare.net/jgzheng/multidimensional-perceptual-map

Multidimensional Perceptual Map:
First Impression
2

Introduction
 Projects are commonly prioritized using a scoring approach
 evaluated according to predefined categories, which are then
aggregated into one or two priority numbers.
 Decision is based on the understanding of multiple aspects
(dimensions) of projects, such as project size (in terms of budget,
time, or people), risk level, expected return, business goal,
strategic impact, etc.
 Aggregated scores may only offer a limited view of project
importance. This often leads decision makers to ignore the
possible differences masked by the aggregation.
 This research presents a visual exploration approach based
on multidimensional perceptual maps which is generated by
self organizing maps.
 It incorporates human intuition in the process and maintains
the multidimensionality of project data as a decision basis for
project prioritization and selection.
3

Common Prioritization Methods
 Follow an indexing or scoring approach which
evaluates projects in a set of predefined categories
with an option of providing simple quadrant diagrams.
 The project priority is commonly represented by one
aggregated number (score) based on a weighted
summation of scores in each criteria.
 Some other methods use two numerical indicators
instead of one
 the additional indicator adds one more dimension of
information and enrich the meaning of projects.
 projects are readily plotted on a two-dimensional diagram
based on two indicators; in doing so, users can easily see
project distributions and overall portfolio composition.
4

Problems when dealing with
multidimensional projects data
 The final decision relies on simple calculated numbers.
 Multiple attributes may be used as inputs and contribute to the calculation
process, but at the end, these attributes are transformed into one or two
indicators for interpretation simplicity. Such simplicity does not always
satisfy business need.
 These calculated final scores may only offer a limited view of the project
importance. An aggregated score tends to homogenize many projects,
hiding useful and relevant information that may effectively distinguish
them (Wang et al. 2003). That often leads decision makers to ignore the
possible differences that get masked by the aggregation, and may result
in decisions that are not well justified.
 Visualization is a good mechanism to comprehend portfolio
composition intuitively. Challenges:
 many visualization diagrams are more confirmatory (for reporting
purposes) than exploratory, where they are mere static reflections of
results after the decision making process has been completed; they are
not well integrated into the decision making process itself.
 focused on techniques of generating the visualization, less focused on
the use of visualization
5

2D Perceptual Map
 Traditional perceptual maps are created using scatter charts or quadrant
diagrams, which are based on two dimensions (X and Y axes). Then data
items are plotted on the plane based on their values for the two attributes.
 These perceptual maps are commonly use for marketing segmentation or
project portfolio management.
6
Cooper, R., Edgett, S., and Klwinschmidt,
E. "Portfolio Management for New
Product Development: Results of an
Industry Practices Study," R&D
Management (31:4) 2001, pp 361-380.
Quadrant or matrix diagrams are
fundamentally constructed based on
only two dimensions. Trying to fit
high dimensional information into
these low dimensional models often
leaves out the richness of project
information, and leads to a narrower
understanding of project distribution.

System Design Overview
 A system that provides assistance in viewing, understanding and analyzing
projects and project portfolios directly based on multiple dimensions of
project data in the complete decision process.
 An intuitive visual exploration approach based on multidimensional
perceptual maps (MdPM)
 addresses the weaknesses of traditional scoring/ranking approaches and
visualization approaches, while keeping their simplicity and interpretability
 reveals the values of underlying attributes and makes them transparent in the
process of viewing, understanding, and analyzing projects and portfolios
 utilizes proper interactive visualizations to effectively and intuitively handle
multidimensional information for the information seeking process.
 involving human strength and supporting managerial intuition (Kuo 1998)
 Two visual elements
7
Profile Chart A profile chart is a visualization of an object based on the values of its
multiple attributes (dimensions) selected to represent the object; such a
visualization forms a representative shape pattern that can offer a unique
visual impression of the object.
Multidimensional
Perceptual Map
A high level overview visualization that shows the distribution and relative
positioning of all objects based on multiple attributes. It is the basis to map
analysis targets (products, projects, people, etc.).

8
Sample Data and Dimensions
Six dimensions
Scores for each
dimension (for one
sample project)

Profile Chart
 A profile chart is a visualization of an object based on the values of its multiple attributes
(dimensions) selected to represent the object; such a visualization forms a representative
shape pattern that can offer a unique visual impression of the object.
 Profile charts are able to present complete multidimensional “profiles,” avoiding the
reduction of multiple dimensions to a single “number,” and providing a strong and
memorable impression that is easy for users to remember and compare.
 Examples: candle-stick charts (used in stock trading technical analysis), Star and Petal,
Parallel Coordinates, radar (or star, spider) diagrams, or can be created using various
types of basic charts such as bar charts, line graphs, area graphs.
9

10
Profile Chart (Details on Demand)

Multidimensional Perceptual Map
 Two major features:
 The map is composed of smaller areas (cells), which are
characterized by a vector of values that represent multiple
attributes (dimensions).
 The vector of each area may be directly visualized on the map.
 The map can be divided into areas at different granularity levels to
meet various exploration needs.
 The positioning of data items in the map is determined by its
calculated measure (usually Euclidean distance) again each
cell.
 The multidimensional perceptual map does not rely on
the definition of any fixed axes.
 An unsupervised clustering technique called Self-
Organizing Map (SOM) (Kohonen 2001) is used to
generate such maps.
11

Multidimensional Perceptual Map
Various map layout
12
Each cell is represented
by a vector (rather than
coordinates) for its
properties

Cells
View
14
An unsupervised
clustering technique
called Self-
Organizing Map
(SOM) is used to
generate such maps.
Each cell is represented by a
vector which represents the
characteristics of a map cell.
Each vector is visualized using
the profile chart, which is
embedded directly in the cell. A
Cell Profile View displays profile
charts of all cells. In such a view,
the changing trend/pattern of all
cells can be directly observed on
the map.

Another Four-Cluster Setting
17

Prioritization Based on Clustering
19

1. Map Generation
 In this sub-process, the goal is to define and create a multidimensional
perceptual map for visual exploration. The essential step is to apply SOM
algorithm and further customize the results by visual exploration.
21
The first step is to prepare data
for map construction and
analysis. The most important
data are the attributes of
projects selected for a
particular task. Once the
attributes are determined, a
data table is prepared based
on all values of these attributes
for every project. Depending on
the value domain of each
attribute, weighting and scaling
may be applied.
In the second step, the map is defined using
the same attributes that describe the
projects. The most important map setting at
this step is the map size, defined as number
of cells (number of rows by number of
columns). The finest granularity of map
regions is determined by the map size. The
size of the map is not predetermined or
suggested by the tool, but rather to be
explored and tested out by users. Based on
a certain map size, the map is generated
using a computer algorithm such as SOM.
The verification step can be done initially or
may be conducted later in the “Visual
Exploration” process if any abnormality is
discovered.

2. Visual Exploration In this sub-process, the key steps are
user interactions with the visualization
(visual exploration actions). The process
and actions are designed in accordance
with the visual information seeking
mantra: Overview, Zoom & Filter, Details-
on-Demand (Shneiderman, 1996).
22
Users utilize three map views to quickly
focus on certain parts of the map and
narrow down candidate projects for final
comparison. They may focus on specific
regions and projects that are of interest
(“Zoom”). Users can switch between fine
grained Cells Profile View (“Drill Down”)
and any coarse grained Region Profile
View (“Drill Up”). A set of target project
groups can be defined and highlighted
on the map (“Filter”) (Figure 9).
Once candidate projects are selected,
users can go further to compare
individual projects head-to-head using
the profile chart comparison tool
(“Details-on-Demand”) (Figure 10).
The profile charts give clear
justifications for analyses and
decisions.
The overview is used to comprehend the whole map and customize it.
The system will transform SOM results into the Cells Profile View (Figure
6) and give users a general sense of the map (“Overview”).
A major action in this sub-process is setting different granularity levels of
map regions (Region Profile View). The system supports the definition of
multiple region sets (multiple ways and levels to divide a map). A user
defines these regions by directly observing, comparing, and contrasting
cell patterns. Each region’s profile is created by the system on the fly and
presented to users though profile charts (see Figure 8).
If the map is not satisfactory, then users have three options: 1) try a
different set of SOM parameters and re-create the map; 2) increase the
size of the map (to decrease the granularity level) such that the resulting
vector change trends are smoother; or 3) directly change selected cell
vectors in the SOM result (bypassing the algorithm).This kind of direct
human intervention is an example of applying intuition and sometimes is
very effective.
Once the map is deemed to be satisfactory, it can be saved and reused
later for analysis consistency. Last, selected projects can be plotted on
the map (Item Projection View).

Conceptual Model
23
In general, the designed
approach is a computer
system driven visual
information seeking
process

Understanding the Approach
An exploratory approach, rather than a
confirmatory one
It helps to quickly understand the big picture,
discover potential patterns, narrow down areas
of focus, and come up with hypotheses
intuitively.
The system is complementary to other
approaches and systems, not a
replacement
May need to use together with other kinds of
tools
24

Applicable Areas
Project portfolio management, project
prioritization
Marketing research
Product evaluation
Performance management
Portfolio management
25

More Infomration
 Zheng, Guangzhi and Vaishnavi, Vijay (2011) "A
Multidimensional Perceptual Map Approach to
Project Prioritization and Selection," AIS
Transactions on Human-Computer Interaction (3)
2, pp. 82-103
 https://www.researchgate.net/publication/264935893_
A_Multidimensional_Perceptual_Map_Approach_to_Pr
oject_Prioritization_and_Selection
 http://aisel.aisnet.org/thci/vol3/iss2/3/
 Zheng, Guangzhi, "A Multidimensional and Visual
Exploration Approach to Project Portfolio
Management." Dissertation, Georgia State
University, 2009.
 http://scholarworks.gsu.edu/cis_diss/34
26

Multidimensional Perceptual Map for Project Prioritization and Selection - 2014 update

In this document

More Related Content

What's hot

Viewers also liked

Similar to Multidimensional Perceptual Map for Project Prioritization and Selection - 2014 update

Recently uploaded

Multidimensional Perceptual Map for Project Prioritization and Selection - 2014 update