The $58 Billion Wake-Up Call

McKinsey published an article earlier this month on automotive warranty costs: $58 billion globally in 2024, nearly double 2012 levels. For some OEMs, warranty spend now rivals their R&D budget.

But the good news is that this can be fixed.

The problem isn’t the data

Manufacturers’ biggest challenge isn’t the lack of data but that it’s fragmented and siloed across systems, making it difficult to create a unified view of your environment. 

While most OEMs rely on many different sources like QMS, MES, telematics, and warranty data, they are not viewing this data collectively. Fragmentation across systems, processes, and functions limits the ability to translate signals into decisions, and that's what's exacerbating the problem at hand.

Why does a single unified view of your operations matter? It goes back to the $58 billion problem. You can’t reduce warranty claims until you have your data in one place. As the article referenced “...early detection depends on correlating multiple signal types rather than relying on a single indicator.” 

But, most organizations aren't creating that unified data foundation, which is slowing down root cause investigation time. McKinsey reports a 7-week average from issue detection to root cause, plus 9-10 more weeks to deploy a fix. Four months of exposure per defect and during that time, more vehicles are going into the field with the same defect. The problem only gets worse. The customer experience suffers and your bottom line suffers even more.

But there is a solution to all of this. We’ve helped teams cut that cycle by 90% by simplifying starting with a unified data foundation.

AI isn’t the starting point

AI, AI, AI. It is definitely what everyone is talking about and it will be for the foreseeable future. But don't rush AI for the sake of doing it. McKinsey notes that 75% of OEMs are experimenting with AI-driven quality analytics and only 10% have satisfactory solutions. There are many reasons for that gap - existing mindsets, fragmented workflows, and various resource constraints. But the most common gap I see is organizations using AI before creating the data foundation underneath it. 

AI doesn't fix fragmented data. It amplifies it. A model trained on siloed, inconsistent inputs will produce unreliable outputs and waste cycles and valuable resources.

While AI initiatives should have a strong data foundation, it shouldn’t be a requirement for moving AI initiatives forward. Both efforts can work in parallel.

What to do next?

I’m going to keep it simple with three easy recommended next steps:

• Identify and quantify your biggest pain point. It may be a quality issue in the field or it may be equipment maintenance issue.  
• Figure out the data sources that you want to be part of our data foundation. You don’t need to do this manually; this is what your vendor should be able to do automatically for you.
• Get comfortable with AI by using out-of-the box functionality. I think Viaduct’s features like “Smart Groups” are an easy way to start. You can identify early anomaly detection with just a push of the button.

The $58 billion number doesn't have to keep climbing. What it takes is a unified data foundation and an urgency to address your problems before they grow any bigger.

That's why I created Viaduct. During my PhD at the Stanford AI Lab, I saw an opportunity to use the power of data and AI to solve the most challenging use cases in manufacturing. With our team of industry-trained data scientists, we have built the fastest and most accurate predictive models in the industry to help OEMs identify issues faster, prevent failures, and reduce costs. And now, as part of Sumitomo Rubber, we are extending that work across the globe.

I would love your feedback so please reach out.