“Good software, like wine, takes time.” This quote by Joel Spolsky eloquently describes how good software should be made - with love and patience. But in real world, with strict budgets, business goals to hit, bloated requirments and tight deadlines, things in software development follow more of Murphy's Law. There have been a few instances detailed below where big software project despite a lot of planning and preparation ultimately did bite the dust. The list includes not only important Government institutions, it also includes huge private companies which failed spectacularly in their ambitious software projects.
A project that started in 2006, as a means to create a new modernized payroll system for 80000 employees in Queensland turned into a hugely chaotic and confusing product, which caused failures to the tune of $1.2 billion over the subsequent 8 years. Several of the staff of the Queensland health department, which included nurses, doctors, janitors, etc. were either underpaid,overpaid or not paid at all.
There were early signs of warning when in 2007, then software vendor for the system, Talent2, said it would not pursue the project further after June 2008. Racing against a deadline led to a rushed tender and a rushed implementation which ultimately caused the failure. Queensland Health had 78000 employees on its payroll which included full time and part time employees. These would result in 24000 different combinations that could have been used to pay the employees. The practitioners, the Queensland government and IBM, failed to take this into account. This resulted in the system which was to go online in July 2008, went online in 2010, after a delay of an extra two years. According to IBM, the delay was due to the Queensland Health Department, constantly changing their requirements. IBM was constantly trying to freeze scope, whereas the health department was trying to keep it Open ended.
The key lesson learnt from this case is that the communication between the Client and the vendor about the product should be concise, clear and detailed before the software is made. The client has to be specific in its requirements and once finalized should not be allowed to change unless under unavoidable circumstances.
The Ariane 5 Missile Launcher was to carry out its maiden flight in 1996 from Kourou, French Guiana. However, 40 seconds after the launch the launcher drifted away from its flight part, slowly disintegrated into parts and exploded. The reason for such an unprecedented event? A software bug. The investigating agency reported that the launcher had drifted off course because of the pressure heaped on the boosters by three powerful nozzles. In other words, the launcher was making up for a wrong turn that it never made. The steering was handled by the on board computer, which based upon numbers given to it by the internal guidance system, thought that the rocket needed a direction change.
This guidance system had actually shut down 36 seconds after the launch. This was due to the fact that the system failed to convert one piece of data from a 64 bit format to a 16 bit format. There was a redundant system to provide backup but that too failed because it was using the same software as the former. Thus the developers while designing the code for the system had failed to take into the account the fact that Ariane 5 was a much faster rocket than Ariane 4 resulting in the failure of the system.
So project management in software development is not only about completing the software, but also testing it extensively before marking it as done, especially when software is running some critical systems.
Sainsbury, is one of the largest retailer store in the UK. In 2000, it earmarked a four year IT expansion project with Accenture. The objective was to make its 4 large depots fully automated as well as build an Electronic Point Of Sale (EPOS) system. Another objective involved installing the barcode-based fulfilment system to streamline operations and improve efficiency. This system was to be incorporated at the Waltham Point distribution center in Essex in 2003.
However, it was found out that the system frequently ran into several ‘barcode-reading errors’. Moreover, it was found that the new system was unable to track stock properly which resulted in Sainsbury recruiting 3000 shelf-stackers in 2004. The total cost over the four years for Sainsbury was to the tune of 3 billion pounds. In addition, writing off IT assets where were obsolete in 2004 cost the retailer another 140 million Pounds. As was the trend, the whole blame was put on Accenture. However, the issue was that just like in the first case, there were issues right from the rollout of the system. Sainsbury couldn’t communicate its requirements convincingly and the project always faced difficulties due to it and thus never got off the ground into production.
Foxmeyer, in 1995, with total sales of 5 billion US$ was the fourth largest wholesaler of drugs in the United States. Anderson Consulting was hired to integrate the SAP system and the warehouse automation system purchased by Foxmeyer. The cost of the whole implementation of the project was worth around $35 million. However, it was bankrupt by 1996 and sold to a competitor for a paltry $80 million.
One of the reasons for the debacle, was that Foxmeyer set up an unusually aggressive timeline for the project, that of 18 months. The warehouse employees, felt their jobs threatened by the project and thus refused to grant their support. The warehouse which was the first to be fully automated was destroyed by the workers in retaliation. Conclusively, the new system was not as good enough as the one which it substituted. To add insult to injury, Foxmeyer allegedly said that SAP and Anderson used the project as a training exercise for its junior employees. All of this resulted in various lawsuits and allegations leading to a complete collapse of the project.
In 1997, SHL-SystemHouse and Electronic Data System came together to work on the Canadian National Firearm Registration System. It was thought that the cost would be only $2 million as the $119 million cost in implementing the system would be recovered by the $117 million that would be gained in licensing fees.
However, more than 1000 change order were demanded from the gun lobby and other politically motivated groups. These changes involved interfacing with 50 other agencies. This interaction wasn’t part of the initial contract and thus the money for it had to come out of the government’s pocket. These resulted in the cost rising up to $688 million by 2001. The maintenance cost meanwhile were going at the rate of $75 million per year. An audit conducted for the project in 2002 revealed that by 2004, the project could end up costing more than a $1 billion, with just $140 million recovered as revenue. The lesson to be learnt here is to finalize the project scope as soon as possible before the demands for change get out of order.
Building software is hard and completing mission critical software projects under tight deadlines and bloated requirements is an invitation for failure. And often adding more manpower to the software projects end up delaying the projects further. And as Frederick P. Brooks Jr. right puts it in Mythical Man Month “For the human makers of things, the incompletenesses and inconsistencies of our ideas become clear only during implementation.”