BRIDGING UNCERTAINTIES GAPS IN SOFTWARE
DEVELOPMENT PROJECTS
Deniss Kumlander
Department of Informatics, Tallinn University of Technology, Raja St.15, 12618 Tallinn, Estonia
Keywords: Software engineering, uncertainties.
Abstract: The uncertainties is a well known factor affecting the final result of nearly any software project nowadays.
Their negative impact is either a misfit between customers’ expectations and released software or extra
efforts that software vendors have to invest into the development process. The paper presents some novel
approaches to uncertainties handling including an ambassador driven communication, discussion groups and
varying length internal cycles with software demonstration meetings.
1 INTRODUCTION
The ultimate goal of software engineering process is
to provide customers with tools that will help them
to automate their activities or achieve desired goals.
The modern software development faces new
challenges as customers demand much higher
quality of the released software, shorter development
cycle and increased flexibility of defining
requirements. The flexibility requirement and
quickly changing business environment produces a
lot of uncertainties for software implementation as it
becomes very hard to match expectations and the
released software after several months of
development. This huge pressure on software
vendors produces a relatively high level of software
projects fails. Researches show that up to 27% of all
projects fail because customers are not satisfied with
the delivered software (Bennatan and Emam, 2005)
and a lot of other projects fail since those do not fit
into budgets. Sometimes it happens since those
projects are having difficulties with meeting
customers’ requirements during final stages and are
rebuilding the software again and again. A situation
with the actual percentage of functionality in use
doesn’t look much better: just 20% of functionality
in average is used “often” or “always” and 16%
“sometimes”. The remaining 64% is either never
used or used just occasionally (Khan, 2004).
Therefore the uncertainties management becomes
very important in order to ensure software
engineering projects success (Kumlander, 2006a).
2 UNCERTAINTIES IN
SOFTWARE PROJECTS
Uncertainties in modern software development
projects are well-known elements although the name
could vary depending on whether the cause of
problems or consequences are discussed. Those can
be produced by external world – business
environment, wrong initial propositions of
customers formulating requirements, or by the
internal project environment – requirements that are
faulty, unclear, corrupted during implementation by
missing important details and so forth. Under
uncertainties in this paper we mean the lack of
certainty about all kind of requirements to software
to be developed by its release date and software
properties (like technology, included functionality
and so forth). Notice that the software in result target
expectation arising by the end date of the project
rather than existing at the moment. The uncertainty
is not the same as the risk since here we deal with a
lack of information (certainty). Of course the
negative uncertainties’ impact is usually measured
and treated as a risk, but it will be just one particular
case. Therefore the “risk” term is not sufficient to
describe the “uncertainty” term and therefore should
not replace it.
There are two major approaches to the
uncertainties handling in software development
projects nowadays. The first approach is to ignore
possible problems and uncertainties until those
occur. This approach is based on a belief that the
240
Kumlander D. (2008).
BRIDGING UNCERTAINTIES GAPS IN SOFTWARE DEVELOPMENT PROJECTS.
In Proceedings of the Tenth International Conference on Enterprise Information Systems - ISAS, pages 240-245
DOI: 10.5220/0001693702400245
Copyright
c
SciTePress
problem can disappear by itself. Actually there are
different uncertainties types and one massive class
of uncertainties are low occurrence probability
uncertainties (that the certainty is different to what is
expected or known at the moment), which are
triggered by strange or add-hoc thoughts that
customers or designers could suddenly come up
with. In this case the reactive approach is working
well as people use to discard many requests they
raised earlier. Unfortunately the reactive approach is
not universal. Sometimes it is too late to react on a
problem when it finally occurs. Uncertainties are
potential problems by our definition as in the future
the software could have to meet requirements that
are not initially included into the design. Therefore,
if a certain (well formulated) requirement appears,
let say in the end project phase, then it could demand
a full rebuild of the architecture and it is surely an
expensive thing to do so late – expensive from time,
resources, marketing and so forth points of view. All
this means that the reactive approach is efficient to
decrease the number of issues the software team has
to deal with, but is expensive if any potential issue
will turn into reality.
The second approach is a “pro-active” approach.
Here any uncertainty is addressed right from the
start (after an uncertainty is identified). There are
different sub-approaches as for example risk
management best practises and open issues
monitoring methods (Sumner, 2000; Karolak, 1997;
Kumar 2002).
Comparing the proactive approach to the reactive
one, we could say that the proactive one surely
demands much more resources and is not very
efficient dealing with a large number of
uncertainties. At the same time it is capable to
decrease the overall risk a lot, so the project is
efficiently controlled since all potential problems are
addressed as much as possible minimising the
potential loss in the project.
All this means that uncertainties should be
divided by priorities and potential impacts. The
uncertainties with low occurrence probability and
low potential loss are not directly addressed by pro-
active methods and the paper is going to follow this
best practise. The remaining uncertainties should be
addressed actively and the paper will propose a set
of actions specific for software engineering to deal
with those uncertainties.
3 BRIDGING UNCERTAINTIES
GAPS
First of all, an efficient uncertainties gaps bridging
requires close monitoring and dealing with all
average and high impact uncertainties. Secondly
those should be dealt actively in order to achieve the
required efficiency. Finally uncertainties should be
addressed using different methods rather than using
a common approach. Uncertainties are like people –
they have different properties, nature etc. General
methods for addressing all kinds of risks that project
management is used to use from projects to projects
is shown itself either as not very efficient since it is
usually quite passive or as not applicable at all for
the uncertainties management in software projects.
The common methods been applied usually give
some “good” feeling (confidence) to project
management that risks are under control hiding the
real status. From our point of view, it is one major
reason why so many software projects fail nowadays
– those are monitored using general project
management approaches (notice again that
uncertainties is something else than risks, so risk
handling methodologies are not applicable here) and
as a result uncovered uncertainties turns into
problems close to the project release date and as a
result the project either requires much more
resources than initially was planned or its
functionality is cut down or the functionality is
implemented “somehow” in order to release the
project in the agreed time period.
Underneath, different types of uncertainties will
be reviewed proposing corresponding actions for
each one in order to minimise uncertainties impact
to the software engineering project result. Under
minimising the impact the following is meant. Let’s
review a moment in the future when an uncertainty
will turn into a certainty, i.e. a certain requirement to
the developed software (its functionality, used
hardware, performance etc.) will become formalised.
The local uncertainty managing task is to minimise
the difference between software that will be
developed at that moment and the new formalised
requirement. The closer we are to the new target the
better we managed the uncertainty. At the same time
the spent efforts to the moment of eliminating the
uncertainty should not exceed the benefit. Therefore
the global (and true) uncertainties gaps impacts
minimising task is to minimise the total efforts to be
applied to achieve the final goal: the software
product that meets all expectations after it is
released.
BRIDGING UNCERTAINTIES GAPS IN SOFTWARE DEVELOPMENT PROJECTS
241
Therefore the process of bridging gaps should
always make some trade-offs between current extra
activities (spent resources) to handle uncertainties
and activities avoided in the future (how much we
will have to spent after an uncertainty is cleared up).
Notice that the potential impact could include extra
hours we need to invest into engineering sometimes
including over-hours in order to meet deadlines, a
potential loss of contracts if the desired functionality
cannot be developed in time, postponed releases and
negative impact on the software vendor imago or
penalties.
The first uncertainty to be discussed here is an
unclear specification or a specification missing some
important elements. The process of requirements
gathering is not discussed here as there are well
known approaches (Somerville and Jane, 2005;
Somerville and Jane, 1997), but those do not ensure
that the final specification can be produced
immediately. Quite often some additions are
expected in near future. The easiest way to address
this problem is to implement as broad system scope
(architectural, technological and functional) as
possible to make the system quite universal. At the
same time the system scope should stay as narrow as
possible to minimise resources spent on the system
implementation, to make the final system fast and
user friendly. Therefore there is always a trade-off to
be made between universality and applicability of
the system, which requires:
A skilled person to make the good trade-off
instead of the bad one (the system should be
efficient and broad instead of narrow and
slow);
Enough information including probabilities and
varieties that uncertainties produce.
Concluding the previous list, we need first of all
information on uncertainties flowing through the
whole system as early awareness allows minimising
negative impacts. Thereafter enough competencies is
required to evaluate uncertainties – guessing the
likely result after the uncertainty is turned into a
specification or what the uncertainty could
mean/require. Many software companies isolate the
development team from talking to customers
deriving obvious advantages from that and
forgetting about the lost part. The less designers,
developers etc. know about customers the more
artificial system they produce in the end of ends.
Therefore we do propose organising customers
review meetings, which can be done by consultants
or requirements collectors, in order to keep key
project persons (those who are making decisions,
design the system etc.) informed on use cases. This
will make information on uncertainties meaningful.
Notice again that uncertainties are not always
potential problems. Those allow to see the future
system from more points of view and the broader
scope doesn’t always mean a slower or less user
friendly system. Besides uncertainties could
highlight other potential ways to use the system, so
it will already meet some future (so far
unformulated) requirements and markets.
The next uncertainty to be discussed is more
general than the previous one. Consider a situation
when the testing team is waiting for development to
generate test cases and development is not able to
come up with a technical solution. Consider also the
previously described case of incomplete
specifications, incomplete design etc. All this can be
generalised by the following definition: there can be
an uncertainty of functionality at any software
development step produced by a previous step as its
production is put on hold since the previous team is
waiting for some information. In most cases the
missing information is not more than 30% of the full
package. The main issue here is that other teams
cannot effectively do their work. So, they either stop
their activities or do something basing on their
“believes” facing a risk to re-work that part. Notice
that typically, despites such slips in deliveries, the
general project deadline is not shifted since it is
agreed with customers or top management. As there
nothing the dependent team can do about the
problem, the solution for this case is based on
addressing the “believes” word from the problem
description above. If the dependent team has to act
then they should do it basing on all currently
available information. Therefore it is useful either to
provide periodically information further down the
project cycle or grant them access to the previous
step team documentation base. The common outputs
(documents) infrastructure that is visible for all
teams’ members can greatly assist solving the issue.
Notice that the lack of information is not the only
case. The other one is too much information – for
example plans are changing each day, each small
probability of changes in future is immediately
submitted down to the development cycle producing
quite a lot of chaos in plans and the implementation
process. Therefore a balance is needed to achieve the
global optimum minimising all kinds of extra work
handling uncertainties.
The next uncertainties class a matching between
the implemented software and stable requirements,
considering a case when developers do interpret
requirements incorrectly. It is an art of
ICEIS 2008 - International Conference on Enterprise Information Systems
242
communication and management to ensure
implementation of exactly what was asked. The
typically problem appearing here are communication
gaps. Formally, a communication gap is a problem
in the communication process that makes the
transferred information to be either lost or deformed.
So requirements, which are flowing through design
and thereafter (in a form of design) through
development, have a lot of chances to be
misinterpreted leading to incorrect software release.
There are a lot of reasons why it could happen. The
corruption of information can occur because of
inequality in knowledge, experience, background etc
of the involved persons (senders, receivers, and
messengers). It can be produced by impossibility to
provide full information communicating by phones
(loss of visual information) (Ludlow and Panton
1995; Kumlander, 2006b; Hadelich at all., 2004),
slow or bad lines including internet communication
forcing to compact messages. The most common
scenario of this case is a distributed organisation
with branches forced to communicate over long
(extra long) distances (Cramton and Weber, 2003;
Kumlander, 2006c). It is quite a typical situation
nowadays as there are much more distributed
organisations than it looks like at the first glance.
Sometimes companies become distributed by their
own wish since:
The development process will be cheaper.
There is a misfit of a skilled personnel
location(s) and product markets. Unlike the
previous case the cost of development is not
necessarily decrease, but company gets much
more skilled employees.
Sometimes companies become distributed because
of external reasons:
After buying other companies located in other
geographical regions;
Company branches have to work together
although it wasn’t planned so initially;
Globalization of operations, i.e. a need to
establish groups in other regions.
All this results in decoupling the development team
into offices that are managed remotely. The problem
is usually even deeper since information submitters
rarely verify the information transfer process results
and therefore the corruption stays invisible until the
release is reviewed, i.e. until the very late phase.
This type of uncertainties can be solved varying
the iterations’ length. The modern software
development, like for example the “extreme
programming” and the “agility with SCRUM”, do
release software in a set of concurrent steps – the
software functionality to be released is divided into
parts and each part development goes through the
full set of development work cycle steps and is
called an iteration (Boehm , 1988). The central idea
here is to produce a possibility to verify intermediate
versions with customers or management and fix
possible problems during the next iteration, i.e. as
soon as possible. So the iterative software
development is a possibility to remove uncertainties,
and the shortened iterations cycle will sufficiently
decrease the overall uncertainty for all project teams.
Unfortunately we cannot just propose using the
shortened cycle since it still should be as long as it
needs to be efficient from the development point of
view. Besides, too often releases (demos) makes
customers unhappy since are demanding too much
their time, which is of course valuable. Therefore the
shortened development iterations are proposed to be
internal and probably exclude some substeps of the
full iteration. Under the internal cycle we mean an
iteration that will end up with an internal demo to
business analysts, management, testing team and so
forth in order to verify the done part and ensure its
correctness. Notice that this demo is done not only
to persons that do verify the result, but also to others
sufficiently increase their awareness and solving
problems of uncertainties described above (see first
two classes of uncertainties). The demo usually
takes just an hour, so it will not demand too much
time and resources from others. Under excluding
some steps, we mean that it doesn’t necessarily
require the full testing or documentation in order to
achieve the goal (verified iteration) with minimal
resources. The internal demo can be done during
weekly/daily meetings, if any is established as
normally such project meetings involve all required
attendees. Moreover it can be done in a web
environment in case participants are located far
away from each other.
The varying iterations length leaves enough
space for the next extra rule: the more important part
is developed (from its impact on the next phases) or
the more misunderstandings we faced recently the
shorter the cycle should be. It is vital to verify
results acting in an unstable software development
environment.
Notice that possible changes of expectations
during software project implementation require both
earlier mentioned approaches to ensure the match
between expectations and software in future: we
need to verify the current software (developed so
far) and have a good architecture, which is broad
enough to feet re-designs into. Moreover the
expectations migration can be produced by
BRIDGING UNCERTAINTIES GAPS IN SOFTWARE DEVELOPMENT PROJECTS
243
intermediate releases, but it will be a controlled
migration which isn’t raised suddenly. So the
expectations’ change is not an uncertainty impact
any longer, but is a cooperative evolution.
A very important approach to be proposed next is
an ambassador driven communication. The idea of
ambassadors is similar to the political one – the
ambassador is an official person accredited to solve
a team problem either in another team or in another
branch. A division on teams always means that
certain communication barriers arise between teams
within the same project. Sometimes you need
somebody, who can speed up some processes by
walking into a person office and asking to do
something the team believes is important. People are
normally quite slow to do something that is
requested by emails or phones and do much quicker
after they are asked (pushed) personally. The same
can be stated for a special person within the
customer company who can help to find
requirements in a reasonable time frame. Another
typical barrier that was mentioned earlier is a
physical distance between teams. For example a
manager locating in the head office looses a lot of
information sources like informal one, a possibility
to walk around and see what people are doing and
having troubles with and so forth. All these troubles
can be solved by ambassadors and ambassador’s
duties could take 10% or less work time of a person
speeding up the process and decreasing the project
uncertainty sufficiently.
One more interesting approach to decrease the
overall uncertainty in the project is promoting
discussion groups. The main idea here is to let
people collaborate and post problems, highlight
potential disadvantages of the current solution,
discuss possible impacts of future requirements and
propose solutions. The following uncertainties to be
targeted here:
Mis-implemented functionality. It is a way to
verify and control how the information is
understood by other people. The incorrect
information interpretation will likely produce
faulty questions or will lead to lack of
understanding during conversations;
Incomplete outputs that are not posted– early
awareness of all team members on possible
future problems and solutions is ensured
during discussions;
Uncertainties on methods to be used to
implement the software.
Notice that too big groups are not advised to have as
it will produce too much information moving around
and sometimes certain anarchy. Therefore local and
global groups can be created – local groups within
each team and a global one that should include key
persons and experts. This will produce certain
hierarchy of discussions. So, all questions from less
experienced members (which are probably their
wrong vision or misunderstanding) can be addressed
without requiring experts’ attention and good ideas
can be promoted to the highest level. The discussion
groups do also promote the collaboration and
cooperation (Rauterberg and Strohm, 1992;
Forsgren, 2006) between different groups producing
a synergy in result.
Another uncertainties type is technological
uncertainties. Sometimes it is specified what should
be done, but it is not clear whether or how it can be
achieved. The technology lack can arise at any stage
including development, testing (for example how
new functionality can be tested in order to ensure
required performance, extensibility etc), logistic (the
system installation procedures and supply channels)
and so forth. These uncertainties are typically
covered by pilot projects and we don’t see any better
solution here. The only remark we can do – the pilot
project has to be carefully organised, for example it
requires information on desired and acceptable
outputs, correctly defined and results should be
verified.
Finally uncertainties of missed information needs
to be addressed actively by corresponding iterations
and functions’ implementation project planning. The
information about uncertainties should not be just
kept somewhere. Instead, it should be directly
reflected in the project plan. It is obvious that
features we don’t know anything about should be
planned to later phases. Unfortunately it is less
obvious for many managers that features which are
specified only partly should also be planned to later
stages. Sometimes the incorrect planning leads to a
situation when the next step team starts to press out
outputs from previous teams although they are not
able to provide any, and lead to internal conflicts. It
is important to not mix this situation with
technological uncertainties or uncertainties on
wishes. The last one can be solved using pilot
projects to demonstrate the proposed software to
customers in order to formalise desired functionality,
interface design etc. – in the same way as
technological uncertainties are solved by attempts to
achieve the desired technological goals. Those have
to be planned as soon as possible to decrease the
overall project uncertainty.
ICEIS 2008 - International Conference on Enterprise Information Systems
244
4 CONCLUSIONS
The uncertainties is a well known factor affecting
the final result of nearly any software project
especially in the modern quickly changing world.
The article aim is to propose some new methods to
minimise the negative impacts that uncertainties
have on software development process. The
following uncertainties types arising in software
development projects were reviewed in the article:
Unclear / incomplete specification
(requirement);
Unstable customers/management’s
opinion/vision on how the final system should
work / look like etc.
Inability to predict the software project final
output (how the system looks like, works etc.)
due possible requirements transformation
during implementation since requirements are
incorrectly interpreted, details are lost etc.
Unclear effect of the current requirements on
later stages: technology (can the required be
achieved), amount of required work (testing,
development, education) etc.
Although uncertainty risk management is not
something new and there are several methods
targeted to solve those problems (Sumner, 2000;
Karolak, 1997; Kumar 2002), the number of failed
projects because of extra costs or mismatches of
customers’ expectations and released software
produced by uncertainties is quite high. The paper
has proposed the following methods in addition to
existing to coupe with uncertainties in order to
bridge the earlier mentioned gap between reality and
requirements:
Promote information on uncertainties to flow
freely through the whole system and
customers review meetings in order to give
enough knowledge to key persons to deal with
provided uncertainties information;
Shortened iterations cycle (varying length
cycles) with an internal software
demonstration meeting to verify it and make
others aware of what is done, what are current
problems and what is in development;
Ambassador driven communication;
Discussion groups;
Pilot projects (is not a new method);
Careful project planning to start pilot projects
as soon as possible and uncertain
functionalities (that cannot be finalised
[developed, specified etc] right now)
development as late as necessary.
REFERENCES
Bennatan, E. N., Emam, K.E., 2005. Software project
success and failure, Cutter Consortium,
http://www.cutter.com/press/050824.html
Khan, A. A., 2004. Tale of two methodologies for web
development: heavyweight vs agile, Postgraduate
Minor Research Project, 619-690.
Ludlow, R., Panton, F., 1995. The Essence of Effective
Communication, Prentice Hall.
Kumlander, D., 2006a. Software design by uncertain
requirements, Proceedings of the IASTED
International Conference on Software Engineering,
224-2296.
Somerville, I., Jane, R., 2005. An empirical study of
industrial requirements engineering process
assessment and improvement, ACM Transactions on
Software Engineering and Methodology, 14(1), pp. 85-
117.
Kumlander, D., 2006b. Bridging gaps between
requirements, expectations and delivered software in
information systems development, WSEAS
Transactions on Computers, 5(12), 2933-2939.
Boehm, B.W., 1988. A spiral model of software
development and enhancement, Computer, 21(5), 61-
72.
Rauterberg, M., Strohm, O., 1992. Work organisation and
software development, Annual Review of Automatic
Programming, 16, 121-128.
Somerville, I., Sawyer, P., 1997. Requirements
Engineering – A good Practice Guide, Wiley.
Kumlander, D., 2006c. Providing a correct software
design in an environment with some set of restrictions
in a communication between product managers and
designers, Advances in Information systems
development: bridging the gap between academia and
industry, Springer, 181-192.
Forsgren, O., 2006. Churchmanian co-design – basic ideas
and application examples, Advances in Information
systems development: bridging the gap between
academia and industry, Springer, 35-46.
Cramton, C.D., Weber, S.S., 2003. Relationships among
geographic dispersion, team processes, and
effectiveness in software development work teams,
Journal of Business Research, 58(6), 758-765.
Hadelich, K., Branigan, H., Pickering, M., Crocker, M.,
2004. Alignment in Dialogue: Effects of Visual versus
Verbal-feedback, Proceedings of the 8th Workshop on
the Semantics and Pragmatics of Dialogue,
Catalog'04, Barcelona, Spain, 35-40.
Sumner, M., 2000. Risk factors in enterprise-wide/ERP
projects, Journal of Information Technology, 15(4),
317–327.
Karolak, D., 1997. Software Engineering Risk
Management, IEEE Computer Society.
Kumar, R., 2002. Managing risks in IT projects: an
options perspective, Information and Management, 40,
63–74.
BRIDGING UNCERTAINTIES GAPS IN SOFTWARE DEVELOPMENT PROJECTS
245