The “Hidden” Cost of Failure
The cost of failure in manufacturing is often generalised as an operating cost rather than measured at the micro level, hitting production yields and profitability. Without the actual cost analysis, this makes deploying process improvement initiatives harder to justify from a Return on Investment (RoI) standpoint and does not address core production issues head-on. Simply put: if you are not measuring the failures how do you know the cost impact and how to fix them.
There are many reasons for production failures, from machine and operator errors, to environmental, tolerance and design challenges. But a lack of visibility and cost association can lead to some unintended consequences. Measuring the RoI for technologies focused on process improvement, from automated inspection to AI powered machine learning, is a key decision driver, but relies on recording failures along with successes. Production yields alone do not tell the whole story. Process failures often result in a degree of waist which can be measured (components in vs. units out), but what about operator failures and interventions. These are far less likely to be reported and measured but have a material impact on the decision to deploy capital expenditure.
Process Improvement Relies on Feedback
In one such example, an OEM outsources production of their products, and on day one negotiates a unit price with the subcontractor, and they deliver to that price. However, unseen to the OEM, there are rising costs being absorbed by the contractor for production failures. “Thats the contractor's problem, not the OEMs”. However, during contract renegotiations, the price to manufacture goes up faster than the competition so the OEM decides to move production to another sub-contractor, and the cycle begins again. Process improvement relies on feedback from production to engineering, identifying production issues and designing them out. This often does not happen and costs go up. Whilst product design issues are not the only causes of production failures, they are a contributing factor along with poor component choice that can impact the overall ease of manufacturability.
Machine learning is not a new concept but with the recent advances in Artificial Intelligence (AI), the opportunities to deploy process improvement technology has never been greater. But the key to success is in the availability of high-quality data. Production lines are complex systems with many variables and for AI to work effectively, data is at the core. Production failures are frequently dealt with on the fly and without recording the event or cause. This results in patterns of failure being missed and consequently, no corrective action taken.
Human intervention is one of the least well measured variables in modern manufacturing, despite being one of the largest contributors to cost, quality variation, and production risk. The soft data problem is defined as the difficult to measure points in processes that have a human element. For example, even the most automated of processes will have some human oversight and timely interventions when things go wrong. Capturing these events for data analysis is challenging but with the right tools possible.
Understand The Human Intervention Impact
Poka-yoke, the Japanese term for mistake-proofing or error-prevention, is a good example of semi-automated processes that control and capture planned human activity for quality control, process improvement and data analysis. But as processes have become more automated, the need to measure unplanned human intervention is critical to understanding the total cost of failure.
AI powered machine learning is now an extremely powerful tool if deployed well but relies on vast amounts of data to detect patterns. By empowering operators and engineers to suggest improvements to their own work and systematically record human interventions, this will enhance the ability of machine learning technologies to diagnose issues and suggest solutions. These could be design modifications, better component choice, adjustments to scheduling and process equipment settings. Not all production processes can collect certain data sets, but the human can, if empowered, through the deployment of smart technologies.
Measuring the hidden cost of failure leads to better decision-making, along with compelling RoI calculations that will keep the finance director happy. By monetising failure and investing in technologies to reduce and eliminate error, manufacturers can become increasingly efficient, more competitive, and crucially demonstrate micro-level quality control through the collection and analysis of data.
Duncan Nicol 2026