Since the introduction of computer-based material planning solutions, planners and supplier chain managers have been scratching their heads trying to determine how to properly size inventory buffers and safety stock settings. Try a Google search today reveals a multitude of different theories and formulas on how this can be done. Yet, planners today are still unsure how to effectively solve this age-old problem How do you calculate safety stock settings? How do you know which of the hundreds of safety stock formulas your Google search returned is the right one? Will DDMRP logic be of value? Are AI/ML-based concepts relevant?
More than anything, planners and buyers want settings they can trust!
Since founding DD Tech in the fall of 2011, we’ve been helping clients improve inventory performance through the implementation of Demand Driven MRP (DDMRP) concepts. As the leading DDMRP solution in the market we gained tremendous experience in finding the DDMRP buffer settings that would improve our client’s results.
Critically, we also learned that the traditional, statistically calculated safety stock methods weren’t giving appropriate recommendations. The simulations we’ve performed for clients have demonstrated that items with the same coefficient of variation (std deviation of demand / average demand) require substantially different safety settings. While statistical methods of safety stock settings assume normal distribution of demand, we’ve observed that the shape of an item’s demand pattern also has a big influence on safety settings. In fact, the statistical assumption of a normal distribution of demand is quite flawed.
Our inspiration:
The music app Shazam has served as our inspiration for Auto Pilot. Shazam can listen to a song playing in a noisy restaurant and quickly identify the song name, artist and album. It does this by taking an analog sound wave out of the air, digitizing it and matching it to Shazam’s library of songs. Typically, Shazam completes this process in seconds regardless of what part of the song is playing.
In a similar manner, item demand patterns look like sound waves, each with their own unique shape and pattern. With this in mind, we thought “If Shazam can be incredibly effective in music recognition, why can’t we apply similar concepts to understanding demand patterns and item attributes in generating appropriate safety settings?”
The Auto Pilot initiative was born.
Customer Supply Chain Collaboration:
With a commitment to earning user trust in our safety recommendations we recognized that the development of Auto Pilot could not be a solo endeavor. We wanted to ensure the voice of our customers guided our efforts. In early 2023, we formed a Customer Advisory Board (CAB) of notable global companies to help guide and validate our efforts on a new module that we’ve named Auto Pilot. Our goal was simple, to develop safety stock settings at the item level that would enable improved inventory results and a substantial improvement in user confidence.
Our CAB included some of our largest global customers, including Aptiv (transportation technology), Greif (industrial packaging), and Coca Cola Beverages Africa (beverage) to provide diverse use cases to help shape the Auto Pilot feature. Their insights have been instrumental in fine-tuning the system for better value and increased confidence in letting Intuiflow configure and manage their supply chains.
Auto Pilot Safety Settings:
The focus of the Auto Pilot feature is to incorporate AI and machine learning in determining safety settings for either dynamic (DDMRP) or static (min/max) inventory buffers. The solution has been developed to provide visual proof to the user that the proposed settings are tested and valid.
As part of the development of Auto Pilot we also created Data Exception functionality to help customers ensure that the resulting item recommendations were free of data anomalies that would distort the safety settings.
With our customer’s support, we’ve pushed the boundaries of existing methodologies and dogmas. Through this holistic process, we developed a new paradigm for configuring, managing, and automating replenishment.
How Auto Pilot works:
In our commitment to making Intuiflow Auto Pilot as user-friendly and efficient as possible, we also wanted to ensure the solution aligned to the conventional practice of ABC classification of items with respective service level targets for each.
Auto Pilot introduces pulse thresholds[1] to categorize demand levels and resulting service level targets. For example, parts with annual demand pulses below ten are identified as buy or make to order, those between ten and twenty annual pulses aim for an 85% service level, between twenty and forty days of demand target a 90% service level, and anything above forty demand pulses targets a perfect 100% service level.
The magic lies in the strategic approach towards achieving these service levels. Auto Pilot generates six (6) solutions for each part based on service level targets: three (3) dynamic buffers and three (3) static buffers. Through simulations for each part, the system determines the optimal buffer sizes and calculates the correct min-max values to achieve the desired service levels with the lowest possible inventory.
[1] Count of days of demand over a selected period of time
Results:
Through our development efforts over the last several months we’ve continued to refine and improve the Auto Pilot algorithms and logic. We’ve also addressed a wide range of ‘edge cases’ that we uncovered along the way and refined the visual elements of the solution. We’ve recently started testing in client production environments gaining further verification of the expected benefits of the functionality. Most critically, we’ve gained essential feedback and insights from the members of our Customer Advisory Board.
Intuiflow Auto Pilot is not just a feature but a major advancement in solving one of the most basic aspects of material planning and inventory management. Our commitment has always been to deliver tangible and sustainable improvements to our client’s supply chain performance while delivering software technology that is trusted by our users.
On Wednesday, January 31st at 11 am EST, we’re hosting a webinar on the new Auto Pilot feature where we’ll dive deeper into the user benefits and key functionalities. Make sure to sign up to the Auto Pilot webinar to get a closer look at the software!
This blog was written by Patrick Willcox and edited by Bernard Milian.
