PraxIS Nov. 2006

06-11 Contents: Human Error, Information Quality, Software testing, Eusprig

ISSN 1649-2374 This issue online at   [Previous] [Index]  [Next]

Systems Modelling Ltd.: Managing reality in Information Systems - strategies for success  


1) Human Error
     We don't make those kind of mistakes here
     Book review: 'Human Error', James Reason
2) Information Quality
     IRM UK Data Management and Information Quality conference
3) Software Testing
     Testing Accessibility and Performance
4) Spreadsheets
     Updates on the Eusprig web site
5) Off Topic
     eVoting cartoon
18 Web links in this newsletter
About this newsletter and Archives
Subscribe and Unsubscribe information


Welcome to PraxIS

Information professionals get worked up about a lot of things because we care about quality. In this issue, I highlight some areas where life and death are at stake.

Patrick O'Beirne

_______________________________________________________ _______________________________________________________

1)  Human error

In the articles below, I mention problems of data quality. The most serious repercussions are in medical and healthcare applications. These are life-and-death issues with real grief. A recent case in Ireland was that of a surgeon who did not check the condition of a patient before operating. A person who went in for an appendix operation had a sample mixed up with another patient and having their entire stomach removed. The 21-year old patient told the court that after being told by a surgeon that it was very unusual to see stomach cancer in such a young person, he had asked before the operation whether there might be a mistake. He said the surgeon told him: 'We don't make those kind of mistakes here'.

Recent studies have found that for the USA alone, "of the 304,702 deaths that occurred among patients who developed one or more patient safety incidents, 250,246 were potentially preventable". This story is covered in: 

'Human Error', J. Reason, Cambridge University Press

I have just finished reading this book by James Reason, a professor of psychology at the University of Manchester. The first and last parts discuss the nature of human error, provide accounts of the genesis of such systems disasters as Three Mile Island, Bhopal, Challenger, Chernobyl, etc. The central section is a somewhat detailed (and for me rather heavily academic but nonetheless interesting) discussion of the nature of human perception and how our responses can lead us into error.

He provides a classification of errors such as illusions, skill-based slips, biases, rule-based mistakes, and knowledge-based mistakes. He describes how the nature of the error can be related to the behaviour of our nervous system. A key point is that humans are 'satisficing' creatures; 'good enough' is the aim, to prevent overload of our perceptive systems. He describes how complexity forces us to fall back on explanations that have a higher frequency of occurring in the past. An important chapter for computer system designers is "A general view of accident causation in complex systems". His chapter on 'Potential measures for error reduction' has several sets of design principles, some of which I summarise here:

  1. Consistent mapping between designer's model and user's model
  2. Simplify tasks to reduce cognitive load
  3. Make visible both what is possible and what the consequences are
  4. Use natural mappings between system state and user intentions
  5. Constraints guide the user to the next action
  6. Assume that errors will occur and design for their occurrence and recovery
  7. Standardise actions, outcomes, layouts, displays, etc
  8. Training should support active exploration for self-knowledge and include error training


2) Information Quality

DMIQ Conference notes

The IRM UK Data Management and Information Quality conference was held in London on Oct 30-Nov 2. My own presentation was on "Minimizing risks in IQ Spreadsheets". Here are some jottings from the other sessions I attended:

Keynote Nov 1 Larry English

Larry's topic was "Moving from Data Quality to Information Quality to Business Quality". He wants delegates to move from thinking about data in databases to information about the business and in turn to business quality intelligence.

He presented an Information Quality maturity model the point is that people usually have to fulfil the conditions for earlier stages before proceeding to the next.

  1. Uncertainty why do we have problems?
  2. Awakening we need to have problems
  3. Enlightenment Identify & resolve problems
  4. Wisdom Specify quality requirements; quality, defect prevention
  5. Certainty We know why we do NOT have Data Quality problems

An objection is sometimes made at level 1 "Whatever the quality is, it's OK, we're making good profits" ; can you make too much profit? (My observation is that if you do it means you're vulnerable to leaner competitors on price.  Management need to feel the pain of bad quality to do anything about it. He points out that you don't pay an invoice more than once; why keep paying for creating data redundancy?

The first steps, by analogy with the traditional quality management world, is to save information scrap and rework, and improve employee satisfaction. He deprecates the usual quality definition of "fitness for purpose" on the grounds that a physical item usually has a well-defined purpose, while information is applicable to many uses.

He regards the Software CMM as defective because it misses the customer aspect. His view is that software is not a product, it's part of a process where the information is the product,

He recommends short term improvement initiatives to derive value but points out that a danger of first-step success is that management may award themselves the savings as bonuses and do not institutionalise improvement.

Four quality absolutes:

  • Conform to requirements
  • Prevent defects, not rework them
  • The target is zero defects
  • Know the cost of non-conformity
  • Graeme Simsion

    I attended two sessions from Graeme. One on "450 data modelers can't be wrong" concluded that the term 'data modelling' is used not as much for simply capturing requirements as for Logical Database Design. A comment on research methods is that if you ask a question on definition, you get the response people think should be given for them to look good; if you ask for a description of what they do, you may get closer to reality.

    In his workshop on consulting skills, these ideas were given among many others:

    Setting expectations: "What does a good outcome look like?"

    In the "Fixed price" vs "Time & Materials" classic debate, he introduced useful shades of interpretation, as an example of negotiating techniques .

    Involve people who may say "Why wasn't I asked?"

    Regular programme meetings are not cosy chats; they end with signing off by the customer that they are happy; that gives a checkpoint that you can go back to.

    If you encounter difficult people, consider in what circumstances you would behave like this.

    He was the only speaker to explicitly allow time in his schedule for filling in evaluation forms, so I did mine then for the whole conference!

    Daragh O'Brien, Fergal Crehan: Poor IQ Can get you sued

    Two members of the Irish Computer Society presented at this event: myself and Daragh. Fergal made the essential points for legal liability for information quality from the point of view of common law. He covered libel, negligence, and the extent of injury or loss. The English case Hedley, Byrne & Co. Ltd. v Heller & Partners established that a person suffering damages due to a negligent misstatement made by another can recover such damages from that person. Daragh drew similarities between the legal approach to identifying liability for negligence and key best practices in IQ Management. He also related some stories of hospital errors, including a heartbreaking one about a premature caesarean operation that lost the baby, even though the mother told the hospital they had the dates wrong.

    Peter Hinsson IT & Business alignment

    Peter commented on the HBR May 2003 article "IT doesn't matter" by Nicholas Carr. The article argued from the commoditisation of IT products to the commoditisation of IT services. Hence the somewhat nervous defence from IT managers. Here are some web links where this has been discussed:

    Books referenced by Larry English describes how to measure information quality, the costs of poor quality, select IQ tools, cleanse data, improve processes, and build in quality controls. Len Silverston provides design templates for financial services, insurance, retail, healthcare, universities, and telecom, in a ready-to-use electronic format, allowing companies and individuals to develop the databases they need at a fraction of the cost and time it would take to build them from scratch.


    3) Software Testing

    SoftTest Ireland held another members' evening on 7 November.

    Darren Pickering of Elemental Creative UK spoke of his experiences in testing the usability and accessibility of web sites. He showed a most telling video of a (sighted) person trying to accomplish three tasks on the website of the Blind Centre of Northern Ireland. She was unable to do them, which was a revelation to the site owner, who of course knew where everything was, and thought it was obvious. He described the limitations of Bobby, AAA, and WAI (Web Accessibility Initiative) certificates.

    One reminder for me is that the web logs on my site that show the search terms that people use is in fact a list of navigation failures. They were unable to find what they wanted from the page layout and menu and had to resort to search. I know myself that if if I don't see what I want immediately on the first page, I'll use a search. If the site doesn't provide one, I'll use Google with the site: qualifier. It gives me results from more quickly than Microsoft's own graphics-heavy search.

    John Murphy of the Performance Engineering Laboratory in University College Dublin spoke on performance testing. He stated that most undergraduates leave third level education with no concept of software performance and so it is easy to see why this ends up as a problem in most enterprise computing systems. Most testers and IT professionals do not deal with performance except in the bad way, there is a performance problem and they have to fix it. There is little planning and preparing for bad performance, even though it happens time and time again.

    Most people think that they know about performance but in reality they know very little. One company he met assured him that "our CTO is very performance aware", however they were not using any performance tools or methods. Faith seems to keep them going rather than science. His spin-off company Crovan is trying to make "performance" the same as "security", part of the project all the way though the life cycle, not just a bolt on.


    4) Spreadsheets

    Don Price's paper analysing the benefits to the UK taxman of checking tax return spreadsheets has been uploaded to

    We can infer from his paper that a significant proportion (ie 15%) of internationally based, professionally qualified corporate officers are signing off on materially incorrect VAT & tax returns. Or more cynically "Incorrect ; but not in error, if by error you mean a difference between intent and result!".

    On the Eusprig yahoogroup, Peter Sestoft recently posted this comment on patents:

    "As a side effect of some other work, I've compiled a list of 233 US patents and patent applications concerning spreadsheets, lightly commented. Indeed, quite a few patents and applications can be described as odd. "

    See appendix A of Technical Report ITU-TR-2006-91 at 

    Bill Seddon posted a link to a video of Mike Acuri one of the members of the Office 12 team talking about new features in Excel 2007. Among them a new way to specify formulas that it is hoped will help to reduce errors. It's long (57mins) 

    Eusprig is preparing for our July 11-13 2007 conference in Cambridge - sharpen your pencils!

    Spreadsheet Check and Control: 47 best practices to detect and prevent errors Available worldwide from Amazon.  Our special offer - free shipping to EU.



    Simply send your comments to FEEDBACK (at) SYSMOD (dot) COM

    Thank you! Patrick O'Beirne, Editor

    _______________________________________________________ _______________________________________________________

    5) Off Topic

    Diebold electronic voting machines  provides me with this cartoon and a YouTube video "how to break an electronic voting machine":



    Copyright 2006 Systems Modelling Limited, . Reproduction allowed provided the newsletter is copied in its entirety and with this copyright notice.

    We appreciate any feedback or suggestions for improvement. If you have received this newsletter from anybody else, we urge you to sign up for your personal copy by sending a blank email to   EuroIS-subscribe (at) yahoogroups (dot) com - it's free!

    For those who would like to do more than receive the monthly newsletter, the EuroIS list makes it easy for you to discuss issues raised, to share experiences with the rest of the group, and to contribute files to a common user community pool independent of the web site. I will be moderating posts to the EuroIS list, to screen out inappropriate material.

    Patrick O'Beirne, Editor
    "Praxis" means model or example, from the Greek verb "to do". The name is chosen to reflect our focus on practical solutions to IS problems, avoiding hype. If you like acronyms, think of it as "Patrick's reports and analysis across Information Systems".
    Please tell a friend about this newsletter.
    We especially appreciate a link to from your web site!
    To read previous issues of this newsletter please visit our web site at

    This newsletter is prepared in good faith and the information has been taken from observation and other sources believed to be reliable. Systems Modelling Ltd. (SML) does not represent expressly or by implication the accuracy, truthfulness or reliability of any information provided. It is a condition of use that users accept that SML has no liability for any errors, inaccuracies or omissions. The information is not intended to constitute legal or professional advice. You should consult a professional at Systems Modelling Ltd. directly for advice that is specifically tailored to your particular circumstances.
    We guarantee not to sell, trade or give your e-mail address to anyone.
    To subscribe to this Newsletter send an email to
    EuroIS-subscribe (at) yahoogroups (dot) com
    To unsubscribe from this Newsletter send an email to
    EuroIS-unsubscribe (at) yahoogroups (dot) com
    EuroIS is the distribution list server of the PraxIS newsletter. It also offers a moderated discussion list for readers and a free shared storage area for user-contributed files. The archives of this group are on YahooGroups website