Let’s look at a classic supply chain challenge and illustrate the value of implementing an APA solution.
Company A: The Classic IT System
This company implements a very standard process. An analyst inputs the expected shipping time information when the part is “born” into the system. The initial estimate placed into the system can have high variation based on the different logic being used by each analyst who inputs these numbers. A more risk-averse analyst could potentially place more time, while a more aggressive analyst might use lower shipping time numbers to optimize inventory. See the potential for discrepancies? After creating over hundreds of thousands of parts, these timing estimates become even more problematic as seasonality, shipping issues, and the global pandemic skew the shipping times and create large disconnects to the estimates that were placed into the system.
While analysts can work to manually update the fields, it becomes nearly impossible to keep up, and bad estimates result in stock shortage, premium shipping fees to keep production moving, and significant cost and downtime. While this can be corrected with rigorous training and process improvement, it doesn’t eliminate the real challenge.
Company B: The Analytic Approach with Model Management
The traditional analytic approach would be to build a model to estimate the shipping time for each part, and instead of an analyst placing an estimate into the system, the estimate would be automatically created by the model. This significantly reduces the initial variations, but how does the system deal with the shifting underlying data? For companies that have strong discipline with model implementation, they would likely monitor the model periodically to understand if the data has shifted to the point where new models need to be created. For most companies, model management happens on an annual, quarterly, or even monthly basis. However, when significant events like a pandemic hit, this can be too long to wait. Again, even more problematic for most of these implementations is that the system is then adjusted for new estimates placed into the system but doesn’t automatically update all the prior estimates being used.
Company C: The Modern Analytic Process Automation Approach
Model monitoring may quickly become an outdated concept as modern analytic tools allow continuous adjustments to the model based on the shifting incoming data. Instead of monitoring the model, the system automatically adjusts. In this example, the current shipping times are monitored by the system, and as the times shift, the estimated shipping times in the system are all adjusted. This approach can have real-time adjustments occur as the shockwave of a global event occurs.
It’s amazing that so many companies are still struggling to move toward a modern-day approach, as the technology to make this happen is already with us and can be implemented by domain experts. The shift from an old IT-system approach (Company A) to the modern-day Analytic Process Automation approach (Company C) is now being accomplished in mere weeks, saving companies millions of dollars in both inventory and premium shipping. Of course, this approach also leverages monitoring of the models, but instead of determining when to create a new model, the goal is to alert when new models are significantly changing so that a human can review and approve the change.