EVALUATION OF ENTERPRISE TRAINING PROGRAMS USING
BUSINESS PROCESS MANAGEMENT
Fod´e Tour´e and Esma A¨ımeur
D´epartement d’Informatique et de Recherche Op´erationnelle, Universit´e de Montr´eal, Montr´eal, Canada
Keywords:
Business Process Management, Business Intelligence, e-Learning, Evaluation, Return on Investment,
Optimization, Business Activity Monitoring, Machine Learning Algorithms.
Abstract:
The investment in human capital, by means of training delivered in enterprise, became an important constituent
of enterprise competitiveness strategy. Consequently business managers require from their human resources
managers, training departments, or even of consultants working in the field of training, the proofs of training
investment yield in terms of tangible and intangible profits.
To evaluate training in enterprise, two models predominate, namely the model of Kirkpatrick and that of
Phillips. In this paper, we propose an approach of training project evaluation, based on business process
management. It is an approach which fills the gaps raised in the literature and ensures an alignment between
training activities and business needs.
1 INTRODUCTION AND
PROBLEMATIC
Individual and collective skills are the most important
assets for organizations, and determine their produc-
tivity, competitiveness and ability to adapt and to be
proactive when uncertain. Training is a key strategy
for generating skills in people. This is why investment
in training is high; the American Society for Training
& Development (ASTD) estimated this investment to
126 billion USD in 2007 (Paradise, 2007).
Many organizations assess whether learners liked
a course or acquired new knowledge, but few have
cracked the code on how to determine learning Re-
turn On Investment (ROI). The most commonly used
metrics for evaluating training programs are those de-
rived from the work of Donald L. Kirkpatrick (Kirk-
patrick, 1994) and Phillips (Phillips, 1996). Table 1
shows the measures of course evaluation reported in
the 2008 Benchmarking Study conducted by Corpo-
rate University Xchange.
There are several models in the literature. Some
of these models allow calculating the return on the in-
vested capital, and could help organizations to make
better educated decisions regarding workforce train-
ing. However, because of the difficulties bound to
the use of these models, human resource departments
cannot estimate, in a concrete way, the impact of the
training on the economic and social growth of their
Table 1: Course Evaluation Methods by Level, in (Rozwell,
2009).
Course Evaluation Method Percentage of Courses
Evaluated Using
this Method
Level 1: opinion of the course
and instructor
75%
Level 2: knowledge acquisition 47%
Level 3: behavior change 20%
Level 4: business impact 12%
Level 5: return on investment 6%
enterprise.
A study, led by ASTD and Institute for Corpo-
rate Productivity’s (i4cp), shows that few organiza-
tions feel they have mastered the learning evaluation,
and many admit to face ongoing challenges (Patel,
2010). Besides, methodological problems are also
highlighted by respondents, in particular for evalu-
ation levels 3, 4 and 5 (the level 5 corresponds to
Phillips’ model) and isolation of training effects in the
results.
In this paper, we propose a training project man-
agement approach based on business process manage-
ment: going from concept to optimization, via the
evaluation of the financial and non financial yield.
In the remainder of this paper, we shall present,
in section 2, the two basic models for evaluating the
training in enterprise, the advantages and criticism of
505
Touré F. and Aïmeur E..
EVALUATION OF ENTERPRISE TRAINING PROGRAMS USING BUSINESS PROCESS MANAGEMENT.
DOI: 10.5220/0003837305050510
In Proceedings of the 8th International Conference on Web Information Systems and Technologies (WEBIST-2012), pages 505-510
ISBN: 978-989-8565-08-2
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
these models. Section 3 will be dedicated to the pre-
sentation of our model and a final conclusion is pre-
sented in section. 4.
2 THE KIRKPATRICK/PHILLIPS
MODEL
The concept of yield covers a rather wide spectrum,
going from effect perception to the return on invest-
ment calculation. These two dimensions join the
distinction between ”financial yield” to describe the
measure or the calculation of what the training brings
to the organization on financial plan and ”training re-
sults” to describe the impact or the effects which are
not of financial nature.
Kirkpatricks model began in 1959, with a series
of four articles on the evaluation of training programs
in the journal ”Training and Development”. These
four articles defined the four levels of evaluation that
would later have a significant influence on corporate
practices.
The four levels of Kirkpatrick’s evaluation model
essentially measure (Kirkpatrick and Kirkpatrick,
2006; Kirkpatrick, 1994):
Level 1 - Students Reaction
How did the trainees react after the training? Did they
appreciate this one? Are they satisfied? What they
thought and felt about the training.
Level 2 - Learning
What they learnt after the training? What knowledge,
skills and/or attitudes (knowledge, know-how, and social
skills) have been acquired? Have educational objectives
been achieved? The resulting increase in knowledge or
capability. It is about the educational evaluation.
Level 3 - Behavior
Do the trainees use what they learned in training at their
workstations? What new professional behaviors have
been adopted? Extent of behavior and capability im-
provement and implementation/application.
Level 4 - Results
What is the impact of the training on the results of the
company? Example: decrease of the rate of absenteeism,
occupational accidents, growth of turnover, the produc-
tivity, customer satisfaction, etc. The effects on the busi-
ness or environment resulting from the trainee’s perfor-
mance.
Although the four-level model of Kirkpatrick is
widely recognized and accepted, and although a sig-
nificant number of evaluation methods find their base
there, many have argued that this method does not
provide the data required by managers today, which
Phillips has to overcome.
According to Phillips, the calculation of the yield
of the training is made by means of a process by
stages which supplies a plan detailed for the plan-
ning, the collection and the data analysis, which in-
cludes the calculation of ROI (Phillips, 1996; Phillips
and Stone, 2002; Phillips and Phillips, 2003). The
process begins with the evaluation planning: where
objectives are developed and decisions are taken on
the way the data will be collected, treated, and ana-
lyzed. The data collection is made according to train-
ing evaluation levels (level 1: reactions/satisfaction;
level 2: learning; level 3: transfer of the learning and
the level 4: the organizational results). Finally, at the
level of the data analysis, we have the crucial stages
for the analysis of ROI:
Isolate the effects of the training from other fac-
tors of influence (use of one or several methods
to separate the influence of the training project of
the other factors which influence the measure of
the organizational results),
Convert the data concerning the organizational
impacts into money values for developing an an-
nual value of the project,
Profits and costs are combined in the ROI calcu-
lation,
The intangible profits are identified by this pro-
cess (they are included in this category only after
having tried to convert them in money values).
In conclusion, training is a key strategy for staff
development and for achieving organizational objec-
tives. Organisations and public authorities invest
large amounts of resources in training, but rarely have
the data to show the results of that investment. Only a
few organisations evaluate training in depth due to the
difficulty involved and the lack of valid instruments
and viable models (Pineda, 2010). The entire notion
of the Kirkpatrick/Phillips model may not truly mea-
sure the impact of the Learning Function on the orga-
nization, even under the most optimistic scenarios. It
measures only the possible impact of isolated training
events (Mumma and Thatcher, 2009).
Thus, to try to bring a solution to the enterprise
needs, we present in the following section an ap-
proach of training yield evaluation, based on the busi-
ness process management.
WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies
506
3 A MODEL OF TRAINING
EVALUATION BASED ON
BUSINESS PROCESS
MANAGEMENT
Business Process Management (BPM) represents a
strategy of managing and improving business perfor-
mance by continuously optimizing business processes
in a closed-loop cycle of modeling, execution, and
measurement. A global study by Gartner confirmed
the significance of BPM with the top issue for CIOs
identified for the sixth year in a row being the im-
provement of business processes (Gartner, 2010).
Given the success registered by the BPM solutions
in the management of enterprise processes, why not
use this approach to manage, efficiently, the training
activities in enterprise? An affirmative answer to this
question supposes that we have to consider these ac-
tivities as being business processes.
Indeed, to design and realize a training project
supposes getting through various stages: going from
the formulation of a request up to the implementation
of new skills. The reality in most enterprises is that
they need to figure out how to make their spending
for training have a greater impact on corporate perfor-
mance. When training needs are viewed with a criti-
cal eye, many organizations will find that they simply
do not have enough money to train every employee
equitably. So they need to focus their training expen-
diture on the roles that are most essential for business
success and that return the most value to the organiza-
tion. That’s why, for the management of the training
projects in enterprise, we propose an approach based
on five stages, as shown in Figure 1.
Figure 1: Steps of management of a training program in
enterprise.
In the following subsections we present the stages
of our approach. To facilitate the understanding, we
use the following sample scenario:
In an enterprise of software development, the -
nancial manager notes an increasing of penalties
owed to the delivery delay. After analyzing the situa-
tion , he remarks that the delays are bound to projects
which integrate the programming in Xforms. Hence,
the enterprise decides to offer an accelerated train-
ing to its employees. The training cost is estimated
at 72000$. The training is offered in the afternoons
(therefore, employees work half-time). How to insure
a simple and effective management of this project?
3.1 Stage 1: Study of Training Project
The first stage of our approach consists of analyzing
the demand for training and associating it with ele-
ments of performance of the enterprise. It is translated
by a certain number of actions such as: the conversa-
tions of exploration of the demand, the definition of a
plan of change, needs analysis, definitions of the ob-
jectives, the definition and the choice of performance
indicators.
In the scenario above, it is important to isolate,
first of all, indicators associated to the problem: cost
of delay in delivery, number of software deliveredlate,
cumulative time of delay....
Necessary to take into account factors which can
have the same effect. These correspond to the fol-
lowing indicators:staff turnover rate, employee’s ab-
senteeism rate,number of absence per employee, rea-
son of absence, cost of rotation, cost engendered by
the absenteeism, cost of absence per employee, job
satisfaction degree, personal initiative degree, staff
productivity, collaboration level between employees
within the enterprise, collaboration level per em-
ployee....
Having identified factors bound directly or indi-
rectly to the problem, it is necessary to calculate the
real cost of training for the ROI calculation. For our
scenario, we must add the losses incurred by the en-
terprise during the training period, the cost of time de-
voted to the identification, and needs analysis (com-
bined time of employee, supervisor. ..). Finally, it
is necessary to define the objectives of the training
and to link them to enterprise business needs. In our
scenario, it is to decrease the ”cumulative delay time
”that is linked to the turnover of the enterprise by the
cost of delay penalties. In other words, it is a question
of insuring an alignment between the objectivesof the
training and the business needs.
3.2 Stage 2: Modeling and Validation
To model a business process, we use graphic objects
EVALUATIONOFENTERPRISETRAININGPROGRAMSUSINGBUSINESSPROCESSMANAGEMENT
507
developed by Workflow Management Coalition
(WFMC, 1999).
A business process model can contain two types of
structural conflict: deadlock and lack of synchroniza-
tion (Sadiq and Orlowska, 1999; van der Aalst et al.,
2002; Lin et al., 2002; Sadiq et al., 2004). In order to
verify or to assure the correctness of a process model,
we use an algorithm based on reduction-based al-
gorithms and graph-traversal algorithm (Tour´e et al.,
2008).
For the management of a training project, there are
at least two process models: the process model bound
to the training planning and the process model bound
to the stages of evaluation of the training yield (in-
cluding the stages of evaluation presented in the sec-
tion 2). In this stage, we define indicators allowing
evaluating the training. These indicators allow react-
ing in real time to push aside any situation which can
lead to the failure of the training (non-achievement of
objectives).
The training planning is a graphic representation
of training progress stages. Figure 2 shows a possi-
ble process model, corresponding to our example (the
software development enterprise).
Figure 2: A possible process model for training planning.
We associate to this graph (Figure 2), the actors
of each stage, the temporal aspect and the perfor-
mance indicators linked to the training conduct. As
indicators, we can quote: average emotional state per
learner, average emotional state per training session,
general emotional state per training, satisfaction as
for the training program organization, satisfaction as
for the contents, satisfaction towards the trainer, rele-
vance of the perception, the utility and capacity of the
training to reach its objectives, note by examination,
average score of learning,.. ..
The training evaluation planning is a representa-
tion of information collecting stages, during and after
the training (Figure 3). With this graph, we must de-
fine the collecting means, the date, the objectives, the
actors and corresponding indicators. We also define
indicators allowing estimating the achievement of the
objectives of the training in enterprise. These indi-
cators are related to employee’s life in the company
after the training. We can add for example: increase
of innovation degree of an employee, increase of in-
novation degree in the enterprise, improvement of the
quality of the product, climate at work, number of
committee meetings, customer loyalty, earnings per
employee, ROI. .. .
Figure 3: A possible process model for training evaluation
planning.
After the modeling, we must validate the process
models by taking into account the structure, actions,
data flows and the temporal aspect.
3.3 Stage 3: Configuration and
Execution
A Business Process Management System (BPMS) is
an integrated collection of software technologies that
enables the control and management of business pro-
cesses. Compared with other model oriented develop-
ment tools, such as integrated service environments
and integrated development environments, a BPMS
emphasizes business user involvement in the entire
process improvement life cycle. As a discipline, BPM
is about coordination, rather than control (via au-
tomation) over resources. Beyond task automation, a
BPMS coordinates human interactions and informa-
tion flows in support of work tasks. People, infor-
mation, systems and, increasingly, business policies
are treated as equally important resources that affect
the desired work outcome. This comprehensive ap-
proach to resources also distinguishes a BPMS from
other emerging model-driven application infrastruc-
ture (Gartner, 2009).
This stage is dedicated to the evaluation before
and during the execution, the initialization of indica-
tors by their current values in the enterprise before the
execution of the project of training. The choice of in-
dicators depends on the type of training and especially
the target objectives of the enterprise.
The execution corresponds to the operational
phase where the solution of BPM is implemented. It
is in this stage that the evaluation of levels 1, 2 and
sometimes level 3 of the model of Kirkpatrick (dur-
ing execution some of indicators will already be under
observation) is performed.
3.4 Stage 4: Monitoring
This stage consists in controlling the progress of the
processes. A control based on precise indicators and
relevant in order to have dashboards allowing mak-
ing quickly the good decisions. The dashboard of
the training has to cover two big dimensions: the ef-
ficiency and the efficacy. The training process said
to be efficient if it gives the maximum of results by
WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies
508
consuming the minimum of resources and said to be
effective if it gives the expected results.
The dashboard of the efficiency of the training
will be composed of indicators of consumption of re-
sources and of activities output allowing measuring
the efficiency of each of the three stages of the pro-
cess, as well as the general efficiency of the train-
ing project. The following indicators allow building
the dashboard of the efficiency of a training program:
time dedicated to the identification and to the needs
analysis (combined time of the employee, his supe-
rior and the training manager), perceived usefulness
of the training/time dedicated, the gap enters what the
employee masters and what he has to master, the ad-
equate level of training to reduce or cancel the gap
(beginner, intermediate, advanced), mode of training
(external, intern, coaching, e-learning, tutoring, etc.),
time to design and the elaboration of the program, etc.
These indicators can be analyzed by sex, seniority, so-
cial status, type of training, or operational unity (ser-
vice, department, store, etc.). The dashboard of the
effectiveness focuses either on the effectiveness of a
training, or on the global effectiveness of the training
system. Its structure includes the model of training
evaluation and contains more indicators than the effi-
ciency dashboard.
This stage allows us to calculate the tangible and
intangible training benefits (without additional costs)
by using indicators values.
3.5 Stage 5: Optimization
In this stage of our approach, we use machine learning
algorithms (example, logistic regression, neural net-
works or support vector machines) to classify train-
ing activities according to defined criteria (example,
financial yield) and to do simulations to increase the
efficiency and efficacy of training activities. For this,
we realize a pretreatment on the indicator values to
have a data set for a supervised learning algorithm,
unsupervised or semi-supervised.
When the training evaluation process is com-
pleted, the enterprise training programs will be clas-
sified in two categories: profitable and unprofitable.
Hence, we will have a dataset D
n
that can be used in
training of a machine learning algorithm.
D
n
= {Z
1
, Z
2
, . . . , Z
n
}
i {1, 2, . . . , n}, Z
i
= (x
(i)
, y
(i)
) with x
(i)
R
d
and
y
(i)
) {0, 1}
Each Z
i
is associated to a particular training program
in the enterprise. The x
(i)
are the indicators (see 3.1
and 3.2) related to the training, y
(i)
represents the
training class (profitable or unprofitable), n is the
number of completed training program and d is the
number of indicators.
It is obvious that to use this data set with a machine
learning algorithm, it is necessary to make a pretreat-
ment to standardize or normalize the inputs x
(i)
.
The purpose of the classification is to be able to
predict the achievement or none achievement of the
training objectives by observing only the indicators
behavior. Furthermore, we must be able to determine
the indicators which have more weight in the realiza-
tion of training objectives. That’s why we may use
a parametric machine learning algorithm like logistic
regression, neural networks or the support vector ma-
chines.
The optimization consists of a simulation allow-
ing guiding the training process towards objectives
achievement. To do this we may use semi-supervised
learning. Semi-supervised learning is of great inter-
est in machine learning and data mining because it
can use readily available unlabeled data to improve
supervised learning tasks when the labeled data are
scarce or expensive. Semi-supervised learning also
shows potential as a quantitative tool to understand
human category learning, where most of the input is
self-evidently unlabeled (Zhu and Goldberg, 2009).
To materialize our approach, our objective is to
provide a tool (Figure 4) to help the training projects
management in enterprise with the alignment between
the training activities and business needs, modeling
and validating of the training processes, execution and
supervision of training projects, calculation of tangi-
ble and intangible profits of training, classification of
trainings (in two levels: enterprise - employee), train-
ings optimization (in two levels: employee - enter-
prise).
Figure 4: Architecture of our enterprise training processes
management system.
4 CONCLUSIONS
Business Process Management (BPM) was once de-
EVALUATIONOFENTERPRISETRAININGPROGRAMSUSINGBUSINESSPROCESSMANAGEMENT
509
fined in terms of tools and technologies, it has re-
cently emerged as a discipline encompassing a broad
spectrum of organizational practices. As a result,
the skillsets for BPM endeavors of today’s organiza-
tions have gone beyond the automation of processes
to encompass a wide variety of strategic and techni-
cal skills (Antonucci, 2010). The advantages obtained
through our approach can be seen from two angles. In
the domain of business process management, we add
a new category of business process and extend BPMS
by adding the validation pre-execution (through our
tool).
Concerning the evaluation of enterprise training,
we propose a complete approach of training project
management facilitating decision-making and the cal-
culation of the tangible and intangible profits. With
regard to the existing models, we add a level of di-
agnostic (classification and optimization) allowing to
understand the dysfunctions related to the attainment
or not attainment of training objectives. Our approach
ensures the training activities alignment with business
needs and allows the ROI calculation without addi-
tional investment.
Concerning the problems raised in the literature,
we reduce the bias and additional costs bound to train-
ing yield calculation. Indeed, from the beginning, we
associate the effects expected by the training with cer-
tain indicators that it already uses in the current man-
agement of the enterprise. When financial yield eval-
uation is required, it will be thus able, without addi-
tional costs, to provide data on the quantitative indi-
cators which will show the evolution of productivity
and quality and will be able to translate them into eco-
nomic value.
REFERENCES
Antonucci, Y. L. (2010). Business process management
curriculum. In Bernus, P., Blazewics, J., Schmidt,
G., Shaw, M., Brocke, J., and Rosemann, M., editors,
Handbook on Business Process Management 2, Inter-
national Handbooks on Information Systems, pages
423–442. Springer Berlin Heidelberg. 10.1007/978-
3-642-01982-1 20.
Gartner (2009). Magic quadrant for business process man-
agement suites. Gartner Research.
Gartner (2010). Leading in times of transition. The 2010
CIO Agenda. Stamford, CA, USA.
Kirkpatrick, D. L. (1994). Evaluating training programs
: the four levels / Donald L. Kirkpatrick. Berrett-
Koehler ; Publishers Group West [distributor], San
Francisco : Emeryville, CA :, 1st ed. edition.
Kirkpatrick, D. L. and Kirkpatrick, J. D. (2006). Evalu-
ating Training Programs: The Four Levels (3rd Edi-
tion). Berrett-Koehler Publishers, 3rd edition.
Lin, H., Zhao, Z., Li, H., and Chen, Z. (2002). A novel
graph reduction algorithm to identify structural con-
flicts. In HICSS, page 289.
Mumma, S. and Thatcher, C. (2009). The learning profit
chain ”connecting learning investments to financial
performance”. Corporate University Xchange.
Paradise, A. (2007). State of the industry: Astd’s annual re-
view of trends in workplace learning and performance.
Alexandria, VA: ASTD.
Patel, L. (2010). Overcoming barriers and valuing eval-
uation. Learning Circuits - ASTD’s Source for E-
Learning. American Society for Training & Devel-
opment.
Phillips, J. and Stone, R. (2002). How to Measure Training
Results: A Practical Guide to Tracking the Six Key
Indicators. McGraw-Hill, hardcover edition.
Phillips, J. J. (1996). Roi: The search for best practices.
Training & Development, 50(2):42–47.
Phillips, J. J. and Phillips, P. P. (2003). Using action plans
to measure roi. Performance Improvement, 4:24–33.
Pineda, P. (2010). Evaluation of training in organisations: a
proposal for an integrated model. Journal of European
Industrial Training, 34 Iss: 7:673–693.
Rozwell, C. (2009). Forget roi, measure time to compe-
tency to calculate learning value. Gartner Research,
(G00169917).
Sadiq, S., Orlowska, M., Sadiq, W., and Foulger, C.
(2004). Data flow and validation in workflow mod-
elling. In ADC ’04: Proceedings of the 15th Aus-
tralasian database conference, pages 207–214, Dar-
linghurst, Australia, Australia. Australian Computer
Society, Inc.
Sadiq, W. and Orlowska, M. E. (1999). Applying graph re-
duction techniques for identifying structural conflicts
in process models. In CAiSE, pages 195–209.
Tour´e, F., Ba¨ına, K., and Benali, K. (2008). An efficient
algorithm for workflow graph structural verification.
In Meersman, R. and Tari, Z., editors, OTM Confer-
ences (1), volume 5331 of Lecture Notes in Computer
Science, pages 392–408. Springer.
van der Aalst, W. M. P., Hirnschall, A., and Verbeek, H. M.
W. E. (2002). An alternative way to analyze workflow
graphs. In CAiSE, pages 535–552.
WFMC (1999). Workflow management coalition interface
1: Process definition interchange process model.
Zhu, X. and Goldberg, A. B. (2009). Introduction to Semi-
Supervised Learning. Morgan & Claypool Publishers.
WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies
510