Using Periodic Audits To Prevent Catastrophic Project Failure

8VLQJ 3HULRGLF $XGLWV WR 3UHYHQW &DWDVWURSKLF 3URMHFW )DLOXUH
Paul Dorsey - pdorsey@dulcian.com
Dulcian, Inc.

In the early part of the 17th century, Sweden endeavored to construct the grandest warship ever
built. It would be larger, heavier, and carry more firepower than any ship ever built. It was
covered in beautiful carvings so, not only would it be capable of destroying any enemy vessel, it
would do so in grand style. The construction of this ship consumed 5% of the country’s GNP. In
1628, the VASA set out on its maiden voyage. It sank to the bottom, killing 50 people, after
sailing only 1000 meters from the dock in a light summer breeze. The irony is that this was not
an unexpected event. The project made every mistake common to IT projects (scope creep,
change in project lead, using new technology, failure to follow a careful design). The worst part
was that although the ship failed when it was tested, no one informed management (the king).
Not only did the project fail, but 50 innocent lives were lost. This cautionary tale exactly
parallels some massive IT project failures (although rarely do IT failures kill more than careers
and profits).
Year after year, we continue to build the IT equivalent of the Vasa. The main difference between
the doomed historical ship and “IT Vasas” is that when a ship sinks, construction stops; on an IT
project, we continue to pump millions of dollars into a system after it has already failed. We
cannot prevent poor decisions, but we can prevent people from ignoring them.

0DQDJLQJ 6XFFHVVIXO 3URMHFWV
Several years ago, I was contacted by a graduate student who was working on a thesis comparing
system development methodologies. The student wanted to create a contingency model
indicating which style of development was appropriate for which type of project. I was asked to
give my opinion about when each proposed development methodology should be employed. I
thought it was an excellent question; however, I don’t think the answer I gave satisfied the
student very much. I replied that the software development method employed had very little
effect on project success. I have seen successes and failures using every method. If the
development team used no formal process at all, failure was guaranteed, but it did not seem to
matter which development method was used.
What I told the student was that project management is like leading a bunch of children on a
wilderness hike. Periodically, you should climb to the top of a tall tree and see if you are still
going in the right direction, find out if there are obstacles in the way, and make sure that the path
you are on will get you to your final destination. Climb back down from the tree, gather all of the
children (most will have wandered off into the woods), and point them all in the right direction
prior to moving on.
As simple as this description of project management sounds, it is shocking how often managers
do not follow it. Project managers may go for months (if not years) without taking stock of the
real status of a project. As a result, the architecture of the system may be unsound and failure
may be inevitable, but this critical information will not be discovered in a timely manner.
Frequently, an unsound system foundation is not discovered until there is an attempt to put the
system into production. Of the many shortcuts that are taken in the name of expediency, there is
none more dangerous than failing to make a periodic assessment of the overall project status.

,QWHUQDWLRQDO -RXUQDO RQ *RYHUQPHQWDO )LQDQFLDO 0DQDJHPHQW

Auditing is the critical success factor in preventing massive failures. IT projects need to be
periodically examined to make sure they are on the right track. Failure to periodically assess
means that you could be building a system that will never work and be unaware of it.
These assessments can be internal, but IT professionals are notoriously poor at self-examination.
The most successful way to get developers to consistently test their own code is to force them to
write the tests before they write the code (the so-called “test-first” concept from the Agile
methodology). The official audit should not be done by anyone on the project team.
A project audit is a significant, substantial effort. It is not a review of one part of the system (a
security audit is not a project audit), nor should its scope be limited. The audit should answer the
question “Are we building a system that will meet user requirements?”

$XGLW RVWV
Useful IT project audits are both time consuming and expensive and will the delay the project.
They are intrusive, annoying and usually create political land mines. Assume that the audit will
consume 5-10% of the total cost of the system (excluding hardware). For systems that have been
stuck in analysis or development paralysis, the cost may be much lower. Therefore the actual
audit cost is likely to be around 5-10% of the current total productive work completed on the
project, which, unfortunately, may be far less than the total amount spent on the project.
Meaningful audits cannot take place without the participation of the project team. Team
members will need to walk the auditors through the system, prepare architectural overviews, and
educate the audit team.
Audits can also be very threatening to a project team. In the rare cases when they are conducted
at all, audits are usually only initiated when a project is known to be in trouble and management
is looking for a scapegoat, or wants to find a different solution.
Everyone connected to the project will declare an audit to be a waste of time and effort. Project
management will feel threatened and wonder why you do not trust them. The developers will
wonder why they need to help these outsiders when they should be coding.
Given the economic and political costs, it is little wonder that IT project audits are not as
common as they should be.

$XGLW %HQHILWV
An audit will tell you whether or not you are building a Vasa. Failing a $200 million project after
$20 million (rather than $300 million after cost overruns) is a huge benefit. Of course, the second
time you try to build the project may be no better than the first, but at least you will not be going
down a road that is doomed to failure.
There are hundreds of articles quoting industry project failure rates that vary from 20% to 80%.
These articles differ about what failure means, but regardless of the source of the statistics, there
are enough project failures to conclude that audits make good economic sense.
Periodic assessments can help prevent several common problems:
1. Ensuring that project milestones have actually been achieved. Assessments should be
made:
• when determining project scope


• when the data model is thought to be relatively stable
• after the basic user interface design is complete
• after each major architectural decision.
Such assessments are best made with the assistance of an outside auditor who is not
otherwise connected to the project. The act of defending a decision to a third party often
helps to focus the thinking of the team.

2. Averting unforeseen disasters. It is very easy to think that a project is going along smoothly
when many people already know that it is doomed. Asking developers and users involved in
the day-to-day project development process how things are going on a regular basis can help
reveal problems looming on the horizon.

3. Inappropriate resource allocation. At the outset of a project, it is not always possible to
accurately determine the number of hours or all of the resources required for each phase of
the project. By taking time to frequently assess project status, tasks that have been omitted
will be discovered and a reassessment of the outstanding issues will ensure that they are
being addressed.
Audits do more than simply prevent disasters. They can also help to identify missed
opportunities and find places to improve even a very successful project. An audit will help the
existing team to take stock of the project and provide them with opportunities view the project
from another perspective that they might otherwise never have seen.

7KH 3ROLWLFV RI WKH $XGLW
The political impact of a project audit must always be taken into consideration. If the project
development team or the project manager is resistant to an audit, it is time to worry. A competent
IT professional should be proud of his/her work and be happy to get another team’s perspective.
If they resist an audit, this usually means that they already know what the audit will find and that
the result will not be positive.
It is important to ensure that the auditor is truly independent and brought in only to audit. There
should be no chance that the auditor can work on the project after the audit. There should be no
conflict of interest. The auditors should neither be partnered with, nor have an adversarial
relationship with the development team. On one audit (where litigation was likely), I was not
even told what consulting organization the previous team came from, and I had to perform most
of the audit from existing documentation.
Who does the auditor work for? This is a very important question. On one audit, I was brought in
by the prime contractor who had a bad relationship with the sub-contractor responsible for all of
the IT work on the project. I was told to make the report to the client as general and positive as
possible. In the report for the prime contractor, I was told to list everything that was wrong with
the system requiring correction. The point is that I was hired by the prime contractor as their
agent. Therefore, only in my interactions with the prime contractor was I a totally independent
auditor.
An audit is never totally objective. Biases always exist among the members of the audit team.
They bring their own way of doing things and their own views of the world. However, if you are


building a Vasa that will never float, even a biased audit will reveal that valuable information.
An audit team with a different approach will help to bring new ideas to the project.

)LQGLQJ WKH 5LJKW $XGLWRU
This may be one of the more challenging aspects of performing a successful audit. IT projects
are so rarely audited that auditing is not a widely discussed or advertised service.
You may need to get a team to perform the audit whose members have never done a project
exactly like the one being audited. However, any reputable consulting firm should have some
experience of reviewing existing work as well as creating plans for assuming responsibility for
an existing project to go forward, which is effectively the same task as an audit.
You will not need a large audit team, but each member should be highly skilled. You will need to
make sure that each area of the audit has a technical lead who can evaluate the quality of the
system. Minimally, you will need the following resources:
1. Project Manager – This individual must have managed projects similar in size and scope to
the one being audited. If he/she has never managed anything larger than a $1 million dollar
project, that person is not a good choice to audit a $100 million dollar project.
2. Database Administrator (DBA) – A DBA is needed to audit the underlying database and
make sure that the system is set up to conform to industry standards. This resource can also
help evaluate the system backup and recovery procedures.
3. Application Architect – User Interface design and development is its own specialty. This area
is just as important as the database. In Object-Oriented (OO) or Service Oriented
Architecture (SOA)-driven systems where most of the logic is not coded in the database, this
individual may be more important than a good DBA. Note that SOA and UI architecture are
not two different areas that can be audited independently. The application audit must be done
as a whole.
4. Security Specialist – Assessing system security (particularly web-based systems) is usually
not within the skill set of most designers. You need to have a security specialist for this task.
Experience with high-security military, medical, or financial systems is useful in selecting
the appropriate specialist since security in those industries is vital.

When selecting the audit team, be sure to evaluate the actual individuals performing the audit
and not just their firm’s reputation. This is a very specialized operation that requires very high-
quality talent. You do not need an army of junior developers reviewing each Java class for
adherence to documentation and coding standards. An audit is not a total quality evaluation. It
should indicate the health of the project, not identify every undocumented class.

6WUXFWXULQJ WKH $XGLW
An audit should not be an isolated effort. In addition to substantial periodic audits, large projects
should have some level of independent continuous monitoring to make sure that the ongoing
management of the project is on track.
There are various sections within a system audit:
1. The first step for the audit team is to understand the system. The auditors should have a
thorough knowledge of the system prior to critiquing it. This initial phase of the audit follows


the same process as that of a new team taking over a project. The auditors formulate a project
plan, begin reviewing documentation, and walk through the system architecture and features.
2. The audit should include scalability and performance tests. IT development teams are
notorious for hiding system performance and scalability flaws.
3. Part of the review should also include direct interviews with users and observations of the
working system.

.QRZOHGJH 7UDQVIHU
This step allows the auditors to understand the entire system architecture as if they were taking
over system development. Declaring that the level of understanding of the auditors must be
sufficient to take over development of the system helps to clarify the detailed knowledge needed
by the auditors. Many auditors do not try to really understand a system prior to commenting on
it. This makes their audits superficial and necessarily flawed. The goal of this phase is to gain an
understanding of the complete system prior to working with any or all of the individual system
areas.
The following areas should be reviewed for the knowledge transfer portion of the audit:
1. System Overview – Review the existing Strategy Document. This will allow the auditors to
understand the project charter, roles, responsibilities, and core use cases to be supported.
2. Data Model Walkthrough – Review or create documentation describing each database table
and attribute.
3. Review/Identify Transaction Use Cases – Review the requirements documentation or
interview current personnel to identify how the use cases are supported by the current system
and to understand how users interact with the existing system.
4. Review User Interface Code Architecture – This will help the auditors understand how the
existing applications are built and maintained. Of particular interest is the amount of
information being moved from the client to the application server and from the application
server to the database.
5. Review Reporting Requirements/Architecture – Examine how the reporting requirements are
met in the existing system.
6. Review System Architecture – This includes the information transfer/synchronization
mechanism in distributed systems.
7. Review System Installation/Upgrade Mechanism. This requires installing the entire system
on the auditor’s computers.
8. Internal Control review – Identify known exposures and match them to existing internal
controls. This is a much broader review than simply examining what computer security
controls are in place and includes business exposures such as human data entry errors and
collusive fraud.
9. User Access review – Examine architecture controlling user names, passwords, roles,
privileges and how these are controlled in the system.
10. Review process for handling system bugs and enhancement requests – Review the existing
bug handling and configuration management processes.
11. Multi-Lingual capabilities (if applicable) – Review the current system's multi-lingual
architecture and capabilities.


12. Basic System Requirements – The number of users, bandwidth restrictions, transaction
volumes, performance requirements should be quantified and evaluated.
13. Process Flows – Determine the current system process flows and how they are implemented.
14. Custom Framework – Review the application development framework and documentation
for any custom frameworks.
15. Performance – Audit system performance to determine the number of users in the system,
number and types of transactions performed, and the effectiveness of existing performance
testing. Identify performance bottlenecks.
16. Standards – Review user interface coding standards and how these standards are
implemented and enforced.
17. Training – Review existing training methods and trainer qualifications.

,QIUDVWUXFWXUH $XGLW
In this phase of the audit, examine each of the following areas from a technical soundness
perspective. Compare existing system practices to current industry best practices and document
any discrepancies.
1. System and User Documentation
• Includes code commenting and training manuals

2. Data Model Audit – Examine the system model for the following:
• Violations of database normalization rules
• Appropriate level of abstraction
• Consistency of naming conventions
• Adequate documentation of the model
• Consistency and appropriateness of data types
• Flexibility to accommodate future changes to requirements

3. Database Review – Examine the following aspects of the database structure and design:
• Appropriate usage of server-side code, particularly for batch routines
• Quality/performance of SQL
• Database performance including creation of appropriate indexes

4. User Interface (UI) Architecture Review – Examine the following aspects of the user
interface code and design:
• Binding framework
• UI complex rule enforcement
• Resulting network traffic
• DML Operations
• State management
• Row caching
• Appropriate use of relevant architecture
• Code and documentation

5. Distributed System/ETL Audit
1. Complexity and maintainability of code


2. Quality of architecture

6. Security Audit
• User interface susceptibility to code injection and cross-scripting
• Control/exposure matrix review
• User privilege security review

7. Hardware/Software/Networking Review
• Review existing hardware/software and networking equipment and assess suitability
to meet system requirements

8. Backup/Recovery Procedures
• Review plans for system backup/recovery in the event of system failure or damage

9. Appropriateness of system upgrade mechanism

6VWHP $ELOLW WR 0HHW XUUHQW DQG )XWXUH 5HTXLUHPHQWV
Examine the current set of system requirements; identify all of the use cases, and review a
sample of the use cases for suitability, specifically:
• Compare documented requirements to the existing use cases and how they are handled.
• Interview users to assess satisfaction with the existing system.
• Determine whether or not the existing backup and recovery procedures are sufficient to meet
maximum downtime requirements.
• Assess flexibility of the system to meet new requirements.

)LQDQFLDO 5HYLHZ
Auditors should review how resources have been spent over the lifetime of the project, including
budgeted expenses versus actual expenditures for each year of the project.

$XGLWV RI RPPHUFLDO2II7KH6KHOI 276

6RIWZDUH 3URMHFWV YV XVWRP 6VWHP
'HYHORSPHQW 3URMHFWV
Given that COTS projects fail with the same regularity as custom development projects, the need
for an audit is just as great; however, the structure of a COTS audit is a bit different. If the
project uses mainstream COTS packages (e.g. Oracle or SAP), there is little need for a data
model audit because the strengths and weaknesses of those structures are well known.
Very serious attention should be given to any customization of the COTS:
• Were the customizations necessary?
• Were they appropriate?
• What will be the impact of the customizations on the ability to move to the next release
of the COTS?
Most COTS projects have enough customization that they need the same level of scrutiny as a
custom development project.


7KH $XGLW 5HSRUWV
A comprehensive audit should result in multiple reports. There is usually a high-level finding
report for top management to answer the big questions. This report should be written in terms
that non-technical professionals can understand. This high-level report should be no more than 2-
5 pages.
A detailed audit report should be written for the project manager. It should describe all major
system flaws in detail and discuss the architectural strengths and weaknesses of the system
design. The executive summary of the detailed report should be no more than 10-15 pages. A
complete listing of all observations, flaws, and suggestions might easily require 100 pages or
more.
The existing team should have access to the audit report and be given a chance to respond to it.
Existing team members should have participated in many parts of the audit and senior team
members should have been included in evaluative discussions of the system. The existing team
may not agree with the findings of the audit but, by the time the report is released, there should
be no surprises.
The existing team should be provided with an opportunity to write a response to the report to be
included with the final audit report.

$FWLQJ RQ WKH $XGLW
The most difficult part of the whole audit process is determining how to take corrective actions.
These actions may potentially result in discarding large portions of the existing system.
If an audit brings significant architectural weaknesses to light, the findings of the audit must be
validated by additional review (probably with a different audit team). Even if the existing team
agrees with the findings of the audit, major shifts in the direction of a large project should not be
undertaken lightly.
It might even be prudent to bring in another outsider to review the audit report to determine
whether or not its findings are legitimate. If the project being audited costs hundreds of millions
of dollars, some extra prudence is appropriate.
Once the audit report has been accepted, the basic sunk cost argument must be applied. The
amount of money already spent on the project is not relevant information. The decision about
how to proceed should be based on the lowest estimated cost to complete the project.
Technology may have advanced since the project’s inception. Therefore, starting over may be
the easiest way to proceed.
Sometimes calling something an “upgrade” rather than a “refactoring” or a “rewrite” is much
more palatable, especially to management. However, when some or all of an existing system is
not capable of meeting user requirements, the system must be modified whether the changes cost
$10 or $10 million.


DVH 6WXGLHV
Most Financial Management Information Systems (FMISs) are never audited. An informal poll
at the 2007 ICGFM conference indicated that only 30% of systems typically receive a review.
Frequently, when an audit team is brought in, it is only after it is clear that the project is failing
and the government is looking to better understand the reasons for the failure. Even when an
audit (or “review” as it is frequently called) is done, it is rarely completed to the level of depth
described in this paper. Some audits are best described as “journalistic.” In these cases, the audit
team comes in for a few weeks, interviews all of the stakeholders and then reports on what they
said.

Even though most audits are not done to the level described in this paper, they usually seem to
come to the right conclusions and provide significant value to the project. However, some audits
can reach erroneous conclusions. Journalistic audits will usually report what any stakeholder
says, which may or may not be true or accurate.

Governments may not act upon the conclusions of the audit report, even when these conclusions
are correct. Their representatives may not agree with the report, or may be unwilling to accept
that a project that has already consumed so many resources is a failure. Audit reports including
good news that suggest next steps seem to be more likely to be accepted.

Gathering case study information is difficult. In this article, I can only report the names of
projects and their reviews when the audit report was favorable. Struggling or failing projects are
not usually discussed publicly.

Projects facing massive cost overruns that deliver only a small percentage of the user
requirements still may be reported as successes in presentations and in publications. No
government is anxious to publicize wasting significant resources on a failed project.

DVH 6WXG (WKLRSLD
Ethiopia has been involved in a 12-year economic budget and finance reform led by a team from
Harvard University. As part of the project, a customized budget and finance software system was
built. When Harvard’s role on the project concluded, a detailed audit of the software was
performed. The only part of the audit process described in this article that was not done was the
evaluation of how well user requirements were met.

The conclusion of the audit was quite favorable. The system was well built, well documented,
and mostly conformed to industry standards. The entire system had been built for less than US
$2M (a comparatively very low cost for systems of this type). The network capabilities
throughout Ethiopia at the time required that hundreds of installations be deployed throughout
the country, since it was initially impossible to support a centralized environment. The long-term
goal of the government was to centralize when the infrastructure could support it. The audit
concluded that the existing system would probably not support a centralized deployment and
needed to be redesigned.

The government accepted the audit team’s recommendations and ultimately hired the audit team
company to manage the development of the redesign project. Although the audit had been quite


thorough, when the team took over, they were surprised by how little they had learned about the
system during the audit process. A few areas of the system were far less well designed than
originally believed which demonstrates that even a substantial audit may not uncover all the
problems with a systems development project.

DVH 6WXG 7DQ]DQLD
For their FMIS, the Tanzanian government has deployed the finance and accounting functions
(General Ledger, Cash Management, Accounts Payable, and Accounts Receivable) of the
EPICOR Enterprise Resource Planning (ERP) application.

The system was implemented by a local vendor with funding from various donors. The FMIS has
been implemented in all ministries, departments, and agencies, as well as 21 sub-treasuries
located in various regions of the country. A scaled-down version of the system has also been
implemented by two-thirds of Tanzania’s local government authorities.

A major audit is planned for the future so a preliminary review was conducted by a single
external consultant. Various aspects of the program were reviewed including:

• Business processes, both automated (EPICOR and non-EPICOR) and manual
• Software modules and functionality
• Peripheral applications supporting the FMIS
• Hardware and infrastructure
• User and technical support capabilities
• Knowledge transfer and training
• Vendor relations (including contracting)
• Disaster recovery and security procedures

The audit did not conclude that the project was a failure but did suggest that the system requires
some significant work before it can be declared a success. Some of the key recommendations
included:

1. Resolving contractual agreements with vendor
2. Acquiring additional EPICOR expertise
3. Resolving long standing technical problems with FMIS application
4. Improving reporting functionality
5. implementing a disaster recovery plan
6. Expanding user and technical support training
7. Revising the budget planning process

Although the Tanzanian government did recognize that the FMIS had not yet achieved all of its
objectives, the government disagreed with many of the consultant’s findings. Donors were
convinced by the consultant’s findings and recommendations and continue to request that the
government act on these recommendations.


DVH 6WXG $QRQPRXV *RYHUQPHQW )0,6
The government has implemented a number of public expenditure reforms which are expected to
lead to improvements in fiscal management and service delivery. An FMIS for the central
government and several local governments had been implemented.

Many believed that the system has been successful thus far and performs reliably and
consistently. Despite this, many questions have been raised by some donor staff, especially
within financial management sector with regard to system’s lack of performance and/or its
unsuitability within the present environment.

The donor initiated a review to document the system’s present performance as well as
suggestions for its future functional and practical improvements. Three team members spent two
weeks in country reviewing the system.

The goals of the audit were as follows:

1. Review the FMIS
2. Review the system reporting functionality
3. Observe how the system is utilized
4. Evaluate the technical infrastructure

The major focus of the audit was to assess the status of the deliverables and milestones. It was
concluded that the system is a total success. The main recommendation was to continue on the
current path to deploy to the rest of the country.

Later a second audit was done as there were continuing questions about the quality of the system.
Two auditors spent two weeks reviewing the system. The second audit was far less
complimentary citing significant cost overruns and functional deficiencies.

DVH 6WXG $QRQPRXV *RYHUQPHQW )0,6
In spite of spending millions of dollars and nearly 10 years trying to get an FMIS operational, the
project in question did not seem to be making material progress. Some modules were completed
and pilot tested, but the lack of networking infrastructure prevented the FMIS from being
deployed.

Two auditors were brought in for three weeks to evaluate the system. The audit report was
understandably critical. The report cited poor management, and an inadequate infrastructure to
support the project. Some module development made little progress due to lack of government
involvement.

It seems that there were two critical factors preventing project success:
1. Many commercial FMISs require significant bandwidth. They assume network capability
common in North America and Europe but less common in developing nations. This was
a known issue and should have been recognized prior to spending millions of dollars on
software development.


Using Periodic Audits To Prevent Catastrophic Project Failure

More Related Content

What's hot (20)

Viewers also liked (19)

Similar to Using Periodic Audits To Prevent Catastrophic Project Failure (20)

More from icgfmconference (20)

Recently uploaded (20)

Using Periodic Audits To Prevent Catastrophic Project Failure