Finding the Supply Chain Sweet Spot

Figure 1 - Finding the Sweet Spot

In any supply chain, manufacturing operation, or field deployment of equipment, ‘time is money’ when it comes to detecting quality problems. Even small delays in detecting these issues can result in large costs, such as rework, scrap, product recall, product defect and associated brand damage, legal liability, and the list goes on.

Manufacturers have a large incentive to detect quality concerns before they become actual problems. I call this ‘Predictive Quality Management.’ It’s best to detect these issues on any incoming parts or assemblies before they are built into your product, but it’s critical to detect problems before shipping finished products to your customers.

Finding the Sweet Spot

Most of us have relied on Statistical Process Control (SPC) tools to detect when processes are out of control, thus portending bad quality. But, these tools have limitations. For subtle emerging problems, SPC provides a poor tradeoff between timeliness and statistical confidence. Namely, credible results are late (or require lots of information), and early results are fraught with false positives. The ‘sweet spot’ is early and accurate!

To get there, IBM harnessed Cumulative Sum (CUSUM) mathematics which provide highly sensitive and accurate change detection, even with smaller data sets. CUSUM applied to quality management is an innovation from IBM, with very large implications for industry. Our high-performance CUSUM detection engine can be applied to just about any data where it’s important to quickly and reliably understand if the underlying system is ‘out of control.’

Big Data & Big Value at IBM

We apply our Quality Early Warning System (QEWS) to a huge, cascading waterfall (think big: Niagara, Victoria, Iguacu…) of quality data relating to purchased assemblies, our own manufacturing shops, and on our products deployed in the field. When a QEWS alarm sounds in our upstream supply chain, for example, we quickly alert our supplier and work with them to understand the root cause and resolve the issue. This proactive detection and resolution keeps our total cost of quality as low as it can be, minimizes our warranty exposure and protects the reputation of our brand.

Imagine the Possibilities

Recently, we have begun working with a multi-billion dollar global automaker to apply QEWS to pressing business challenges around product warranty. We engaged in a friendly challenge with our client to see who could detect an emerging quality concern earlier based on the client’s detailed warranty data. We both analyzed many years of data starting from when this particular vehicle model was launched. And guess what.. QEWS won! We spotted the ominous trend over 3 years earlier than the client did, which translated to 170,000 vehicles earlier. These stark results even took us by surprise.

We are now working with this same client on a quality transformation initiative, where the proposed final vision includes performance data streaming from vehicles to a big data analytics environment where QEWS will help spot worrisome trends – early and accurately. One can easily imagine big benefits to all the players in this scenario; the vehicle owner (a safer driving experience), the service repair shop (additional services to customers, operational efficiencies), and the automaker (early detection of warranty exposure, vital vehicle performance information to improve vehicle component design, and brand protection).

I find it to be a great privilege, as well as a responsibility, to work at IBM on such impactful projects as QEWS. As the word gets out, and this new approach to quality is applied throughout the economy (electronics, automotive, pharmaceutical, consumer products, call center management, and beyond), I expect to be surprised time after time by the very powerful value proposition of early and accurate problem detection through the advanced analytics of QEWS.