The Advantage of AI in Optimizing Multi-Tier Supply Chains

The Advantage of AI in Optimizing Multi-Tier Supply Chains - Ceres Technology

Mapping Your Supplier’s Network May Be a Thing of The Past

By Carlos Amaral, Executive Vice President & CIO, Ceres Technology


In my earlier article covering Nostradamus’s low code / no code approach, I mentioned that Nostradamus could predict multi-tier supply chain risk without having to map the entire supply chain network. There are several mapping products available today, and these products offer a good visualization of a supplier’s suppliers’ connections, but the main issue with this process is the time it takes to map every supplier. Regarding a mapping project:

  1. Tier-1s are easy to map because the client purchases products from them, but this is not the end. In our multi-tier world, the tier-1s have their own suppliers. The complexity increases exponentially, because you will be adding multiple regions, countries, raw materials, etc.
  2. Tier-2s are not as easy to map because someone needs to ask for information from each Tier-1. Further complicating this is that you can only map Tier-2s once the data from Tier-1 is received, and another issue is that tier-1s normally do not want to share information about their own suppliers.
  3. Tier-3s are more difficult because of the degree of separation from the company performing the research and the mapping. In addition, with each degree of separation, there are typically fewer respondents (unwillingness to share data grows with each subsequent tier).
  4. Tier-4s and above are almost always extremely difficult to map.

A picture is worth a thousand words. To help you easily understand the points mentioned above, here is a supply chain network graphic that illustrates the limitations of mapping projects. Let’s use the following example of a single-product company:

The above example shows the complexity of a multi-tier supply chain for a single product.

Figure 1: The above example shows the complexity of a multi-tier supply chain for a single product.

The above example is straightforward because I used a company with a single product. Imagine your organization with tens, hundreds, or even thousands of products and suppliers (which may be your case). In such a complex scenario, the above picture turns into a “spaghetti” nightmare. Implementing such a mapping project would be quite time-consuming – even if you could complete one supplier per week.

There are a few questions to consider asking before embarking on a mapping journey:

  1. How long will this entire process take?
  2. How quickly can you begin mapping your entire supply chain?
  3. How often do you refresh each supplier’s data?
  4. What is the cost to implement this?
  5. And the final key question is: How fast will your investment start to pay off?

The Advantage and Power of AI

Disruptions in any one tier can cascade, affecting the entire supply chain. Ceres Nostradamus is an AI-based platform, and it uses over 25,000 real-time global data sets to predict Tier-1 and Tier-2 risk and understand complex relationships.  For example, Nostradamus identified the price of rice in Malaysia as a leading indicator of delays for latex gloves sourced by a Fortune 100 company. The price of rice influences the Malaysian economy, prices, wages, and ultimately, shipments originating from there. Nostradamus identifies overlooked factors that otherwise would remain hidden. By now, you must be asking, how is this possible? This is a fair question, and I will explain below using an example that almost everyone will recognize and understand.

The Hasbro Battleship®¹ Game

Did you play the game Battleship game as a kid? I did, and I loved it! It was and remains a great game of strategy and calculated assumptions.  As the game starts, the players place the pieces on the board,


and then each player takes a turn to try to “sink” the other’s pieces through guesses about the location of the opponent’s ships. Isn’t it exciting when you hit a boat? From there, you systematically start figuring out where the ship might be positioned.

Ceres Nostradamus acts similarly. We built our AI platform to perform several pre-processing tasks to understand each client’s transaction lifecycle.  During the training stage, Nostradamus learns transaction duration (e.g., 10, 15, 20, 30, 45+ days and how the client defines a delay). Furthermore, Nostradamus learns where the Tier-1 and Tier-2 suppliers are located.  With this information, Nostradamus then correlates where the raw materials/commodities and intermediate products are located around the globe, followed by selecting those data sets (amongst our 25,000+ external data sets) most correlated with transaction behavior.

Similar to the Battleship game, this process is equivalent to iteratively taking one targeted shot after another until all ship locations have been resolved and the targets sunk. As the AI system processes this information, it learns the correlations of the different data sets and how each influences the forecasting of delays/risks.  The picture below shows how supply-chain complexity goes far beyond problems between supplier tiers. Additionally, there are too many factors affecting the multi-tier supply chain for a non-AI solution that does not process integrated information in real-time or near real-time to effectively inform supply-chain risk-mitigation decisions.

Our Solution to This Complex Issue

Nostradamus was developed to solve this nightmare. As with any AI-based software, we created our own proprietary processes and models to accomplish this.  I cannot discuss them in depth because too many companies are trying to emulate our solution.  What I can share is that we need very few reference points from the client (similar to Hasbro’s Battleship) to target the problem. At the same time, Nostradamus can simultaneously review multiple data points and identify correlations amongst the different data sets to identify the features important for delivering accurate forecasts of multi-tier supply chain risks. What differentiates Nostradamus as a one-of-a-kind solution is the massive amount of data we have collected in a few short years (it was launched in October 2021). Nostradamus started with a small number of external data sets (~7,500) and today, it is over 25,000. The picture below demonstrates the complexity and interconnectivity among the data that cause delays and disruptions.

Diagram showing complexity beyond the supplier-to-supplier-to-supplier network - ceres technologies, inc

Figure 3: Complexity beyond the supplier-to-supplier-to-supplier network.

Unlike Battleship, Nostradamus can plot multiple features simultaneously, thus making it much more effective and faster than human analysis (especially ingesting the 25,000+ external data sets!). In addition, if a company has data sets unique to it or its industry, these can be easily incorporated into Nostradamus.  We made this process very simple and easy. You can just drag and drop, and Nostradamus will then do the rest.

The Nostradamus “Golden Nuggets”

The graphic below illustrates Nostradamus finding the “gold nuggets” such as identifying that the price of rice in Malaysia is one direct cause of delays for products manufactured there.  The price of rice, however, is but one of several influencing factors: there are normally many others. Nostradamus learns which features (i.e., the factors) affect delays and simultaneously removes unnecessary noise in the data (less important data sets).  The beauty of this process is that it does not require an army of analysts to review the numerous problems that impact the global supply chain at any one time.  It would be difficult for a human to identify a change in the price of rice and then integrate other data points to predict a potential delay quickly and reliably.  Such correlations and predictions are reliably made and managed in a short amount of time using our AI process and model. The following is a real-world example from one of our clients.

Figure 4: nostradamus learns the correlations among external data sets.

Figure 4: Nostradamus learns the correlations among external data sets.

In Summary

We have demonstrated the power of Nostradamus to several industry analysts and their feedback thus far is very positive.  They agreed that Nostradamus is unique in this market.  Also, our initial customer’s executive sponsor told us that he had previously hired the top 3 consulting companies to predict supply-chain disruptions, and these companies could only do 3% better than the model developed by the executive’s internal team.  We delivered 8x better results with a precision of 82% (mitigating false positives) and a recall of 52%.

Figure 5: the advantage of ai - nostradamus.

Figure 5: The Advantage of AI – Nostradamus.

I hope I was able to explain how Nostradamus identifies and captures the complexities of multi-tier suppliers without having to waste time waiting for data-collection forms to be returned from each supplier and mapping them.  Please let me know your thoughts.

In any case, are you intrigued? I would be. My recommendation is for you to try out Nostradamus. We offer many options (ranging from free tryouts to paid proof-of-concept projects) for you to learn about the power of Nostradamus and what it can do for your company.  Please contact our sales team.



[i] The Battleship game is a mark of Hasbro, Inc.