The Latest Insights into the Costs of Bad Code

The cost of poor software quality in the US has grown to at least $2.41 trillion as estimated by the Consortium for Information and Software Quality.

software testing

The cost of poor software quality in the US has grown to at least $2.41 trillion as estimated by the Consortium for Information and Software Quality in 2022 and will continue to raise in 2024. The cost of bad code is a known problem that creates a disparity between the software delivery team, and the clients where the failure costs lead to defective systems, affecting the client’s reputation and project performance.

Multiple Arenas of Software Development

Custom Software Development is an area of Software development offering specialized services through software for varying domains. In a parallel approach driving the no-code capabilities, the worldwide market for low-code development technologies is projected to total $26.9 billion in 2023. This estimate will increase in 2024 due to the growing number of enterprise-wide hyper-automation and composable business initiatives accelerating the adoption of low-code technologies through 2026. With the advent of generative AI, there has been a massive amount of investment by the founding organizations to capture the market and spend it on driving software development through an assistive approach. With such different software development approaches as those followed in 2024, realizing the system to function as per the customer’s needs as standalone systems requires a meticulous design and validation approach to prevent any kind of failure, and investments in testing the real-time system are no longer an option.

A project always considers a separate estimate involved in creating high quality software for a new project or an ongoing project. Operational maintenance costs are derived based on the cost of quality estimated during the project planning stage. Also, an approach to automating the software development model using Model-Based Development, new age software development for AI applications, and low-code applications all contribute our focus to newer methods of software validation. Instead of organizations investing in opportunity costing activities there are much more reliable approaches in eradicating the poor software code. 

Software Failures: A Case Review

The latest insight into software failure comes from the well known autonomous driving system failure. On May 21, 2023 a self-driving Jaguar I-Pace car was traveling on a low-speed Toland Street in San Francisco with an autonomous specialist in one of Waymo’s autonomous vehicles that ran on a dog because of an unoperated safety fail system, a poorly framed fail safe system, and bad software. The system detected the dog that was crossing an unconventional path, but the autonomous system did not activate emergency braking, though it identified the object and tracked its path.

A similar car crash cost the life of an individual in 2018 when the Uber ATG automated driving system detected the pedestrian 5.6 seconds before the impact but failed to identify the object crossing the road as a pedestrian.

In both of these cases, the cost incurred was not only life but also raised a big question mark on public trust in the autonomous driving system. The former accident was because of an unusual path traveled by the dog, but the latter was due to the system’s inaccuracy in activating the emergency system.

Cruise, General Motors’ subsidiary for autonomous vehicles, has issued a voluntary recall for 300 of its driverless cars in January 2023 after a bus crash involving a Cruise autonomous vehicle and a San Francisco Municipal Transit Authority bus. This is again a software malfunction where the system prediction is inaccurate. Before deploying a system to the public, a rigorous safety plan, occupant safety, and backup measures to mitigate any kind of risk are to be identified and addressed. This preventive cost is not mature for autonomous systems, where the regulations vary with geography according to the system condition and environment.

The Minnechaug Regional High School is a well-known example of a system failure where a connected system operating the lighting system of the school had almost 7,000 lights glowing for 2 years. There was not a direct system on/off of the lighting system. The issue was not clear until it was identified that the software was integrated into other school systems and could not be easily replaced. The cost of changing the system design and software update was around $80,000, apart from the electricity usage cost.

Optus, an Australian telecom provider, left 10 million Australians and 400,000 businesses without phone or internet for up to 12 hours in November 2023. The company accepted its failure to design a fail-safe mechanism to address such a large outage critical to serving customer needs. The impact cost the resignation of the Optus CEO, where the outage failed to address traffic delays, shut down the payment system, and over the head of 228 emergency calls.

A recent incident when the generative AI chatbot Gemini from Google was sued for personifying a country’s leader as fascist. The trigger was aggravated by the response of the chatbot to a question on misgendering Caitlyn Jenner. The AI system needs an ethical approach, and human biases are not to be coded into the training algorithm. In the past, there were several allegations, misinformation, and penalties laid for a faulty response by the chatbots. Though this is not the only failure instance, all these contribute to society’s willingness to adopt the new AI system-controlled connected world and AI assistance. 

Plan for a Better Strategy

When generative AI goes mainstream, there is a greater probability of hallucinations being noticed. Similarly, a well connected system fails to go smoothly with mechanical components. The cautionary tales explain our strategic alignment with the customer’s focus and their safety as priority in building any system. Especially in this technological landscape where outsourcing and cost-saving measures dominate other factors, a transparent approach to building the system and to the associated third parties clarifies their responsibility and liability in case of failure.

We see the below ways as the most likely ones to start to get control over the poor software quality problem.

  • Use of emerging technology-based software quality standards, related measurements, and tools like ISO 21434, road vehicle cybersecurity, ISO 21448, SOTIF (Safety of the Intended Functionality), UNECE WP.29, vehicle cybersecurity, and Automotive SPICE.
  • Analyze and assess the quality of all third-party or open source software components to be included in any system. Monitor them closely in operation. Apply patches in a timely fashion.
  • Avoid DevOps and CI/CD models that do not include continuous quality engineering best practices and tools.
  • Integrate continuous Technical Debt remediation into your SDLC
  • Invest in the professionalism and knowledge of your software engineers.
  • Consider having your developers certified for knowledge of the critical code and architectural weaknesses in ISO/IEC 5055 when OMG makes its “Dependable Developer’ certification test available in late 2023 or 2024.

Building a fail-proof system is not possible, but an accredited standard and transparent approach to mitigate the risks of software failure and poor code is possible.

Kirthikka Devi Venkataram

Kirthikka is a Product professional, served B2B segments of multiple industry verticals like automotive, wireless telecommunication services, AI start up and deep tech signal processing research incubator. She holds a Master of Engineering specialized in Applied Electronics securing Anna University’s V rank and as top 5% in the overall batch. She has been recognized for her significant contribution in projecting before prospects about the Business Unit’s technology and service during her early career. She has coordinated and led professionals of diverse experience and culture in her Executive Development Program adding value to the group. She is a Lead to Women Who Code Data Science community.  A busy mother and passionate about travelling and cooking.


Kirthikka Devi Venkataram
Kirthikka is a Product professional, served B2B segments of multiple industry verticals like automotive, wireless telecommunication services, AI start up and deep tech signal processing research incubator. She holds a Master of Engineering specialized in Applied Electronics securing Anna University’s V rank and as top 5% in the overall batch. She has been recognized for her significant contribution in projecting before prospects about the Business Unit’s technology and service during her early career. She has coordinated and led professionals of diverse experience and culture in her Executive Development Program adding value to the group. She is a Lead to Women Who Code Data Science community.  A busy mother and passionate about travelling and cooking.