How AI is Changing Inventory Management
Updated: Jun 9, 2020
The title of this one contains two big words, each needing a 10,000-word piece to give it least justice. I will do it in less than 2,000 words. So, Here goes:
Artificial Intelligence is forecasted to impact the inventory management sector with a value of $215Bn, that’s according to McKinsey. Although, I think these statistics often seem extraneous to most business owners/executives, I will make you see how it could do so!
I will talk about A.I in the analytics theme, so considering no Robots, 3-D printers, Autonomous vehicles nor aliens. This is just pure data in, Algorithm out, dollars saved, prizes won and many smiley faces. Also, I will be taking the inventory levels aspects, so warehousing, sales, operations are not exactly focused upon here.
Let’s begin with two main principals of inventory management:
The cycle stock, one of the critical activities of inventory management. The cycle stock drains as customer orders and replenished cyclically when supplier deliver the goods.
What’s the dream of any supply chain manager you ask? Well, other than a beach house on a south Asian island, it is that the cycle stock amounts, exactly match the replenished stock, so they can always supply their customers, and don’t keep unused material. To be able to achieve perfect Cycle stock, it means that 100% of the demand is fulfilled and no inventory is kept.
Yup, this rarely happens. The stock matching I mean, Beach houses on the other hand are more common with supply chain managers who use A.I 😉. This brings us to the other principal, that is the Safety stock.
The safety stock offers protection against uncertainty in the demand or in the replenishment lead time.
Like everything in life, Safety Stock comes at a cost. Though, this cost Supply chain managers are willing to pay, to make sure they’re meeting the demand coming in. Usually this cost has 2 components, Inventory Withholding cost, which is the is the amount of capital bound in inventory.
The second component is Warehousing costs. Depending on the size and shape of the product a different fraction of personnel costs to handle the product, rent costs of the facility, utility costs, inventory shrinkage, insurance costs..etc is ensued.
Also, there is another dimension to the above, supply chain managers must keep in mind, and that’s Time to replenishment. Safety stock must cover for the replenishment lead time as well, that is, the time from ordering the material from the supplier, till receiving the material.
Let’s say if takes 9 days for the business to order and receive materials, manufacture the product, and transport it and the customer’s expectation is to receive an order within 6 days of placing it, then inventory needs to be kept to account for 9-6 = 3 days’ worth of demand plus demand and supply variability, and this is really uncertain, especially in Egypt and the region, where shipping schedules have a high degree of flexibility.
See what’s the need for the beach house now? This is only a fraction of what supply chain managers need to be attentive to on daily basis.
To manage adequate inventory levels, Supply chain managers need to account for two main uncertainties.
Demand Uncertainty, Examples of demand changes are inaccurate forecast or shifting customer orders:
Customers not buying the product due to its price
Customers found a new competitor
Customers are ordering too much
Supply Uncertainty, Examples on the supply-side variability in the lead time of the incoming products needed to manufacture the outgoing products
Bad weather
Shortage of Material
Prices of material
Long lead times
Now as one sees here, Supply chain management is, for the most part navigating uncertainty. This is where Artificial Intelligence comes to play.
We're now about ~600 words in! This is a gentle reminder to do a quick stretch every now and then even while sitting down. Our lemur friend here knows where it's at.
Predictive Analytics
An ideal Artificial Intelligence is be capable of dynamically managing item reorder/stock replenishment points and preferred stock levels based on lead times, historical data, seasonal-based sales demand, market macro data (such as CPI, inflation rates..etc), current stock levels, patterns in demand change, weather trends, social media data and number of days’ supply to stock.
Basically, an artificial intelligence can look (Be trained on) at all these data points, understand how they correlate and cause each other, and solve for Inventory optimization, to reduce working capital and its associated costs, while always meeting demand which leads to improved cash flow.
Instead of using fixed safety stocks, or primitive demand forecasting, A.I replenishment considers the predicted demand distribution and dynamically restocks, considering the nature of supply variance to fulfill a certain levels of efficient safety stock levels. This A.I can attain a balance of demand and supply, based on the capacity and planning parameters of the business, resulting in, cost efficiency with lowest risk not to meet customer demand.
This A.I is using Predictive analytics in demand planning, which analyzes hundreds to thousands of internal as well as external demand influencing variables (e.g., weather, trends from social networks, sensor data) to model complex relationships of all variables and develop precise demand plans. This greatly reduces forecasts uncertainty and variability, since these algorithms are trained on big volumes of data, and base their prediction on so many features, that they exceed human based forecasts by orders of magnitude.
Artificial Intelligence in Egypt can be particularly impactful, due to the greatly varying lead times from the suppliers, uncertainty of demand, and the continuously changing market dynamic. In one of Synapse Analytics A.I deployment cases in Egypt, we reduced the forecast error from a staggering 300% to a 15% error.
Not to mention, that these algorithms are dynamic. In the sense, that they evolve and learn as the business operates, further reducing their forecasting errors and can flexibly order merchandise, if it predicts that the material price would increase for example (Supply chain managers sometimes refer to this as hedging material).
Furthermore, an Artificial Intelligence can be integrated with MRPs (Production scheduling and material requirements planning software) to create a dynamic, short-term production schedule to fulfill the customer orders while balancing costs, such as inventory holding costs previously discussed.
Computer Vision
Moreover, one of the fast-growing fields in A.I is computer vision, by which algorithms can be trained on Video data, to provide meaningful insights. When it comes to inventory management, this can help with stock counting inside a warehouse, giving a more real-time and accurate view of the current stock levels, dynamically sending email alerts (or auto-replenishing, where an automated purchase order can be placed, avoiding stockouts) the inventory manager to restock, when stock goes below a certain level and a predicted surge in demand is identified.
A.I can also detect, in some cases, defects in inventory which can be replaced proactively, also reducing the risks of stock outs, and machine downtimes due to the lack of material. Although, this requires special warehouse organization, planning and high quality CCTV, by connecting A.Is to their network streams, these videos can be analyzed for inventory counting and defect identification with no extra hardware.
Not to mention, that using Computer Vision, A.I can detect if there are any unidentified personnel within the warehouses, or any type of hazardous fires/flares, alerting warehouse managers, and proactively interfering to avoid any inventory loss or damage.
Image Processing
Moreover, an alternative application of artificial intelligence in Egypt especially, is using A.I based OCRs, by which distributors can digitize invoices, by taking pictures of the invoice, which can add to the dimensionality of the data as well, adding invoice details, and location of the drop improving A.I recommendations.
In short, as Inventory reduction or optimization remains one of the key targets of companies and at the same time customers are demanding higher service levels and increased convenience for example, shorter lead times, A.I is able to make this happen more than ever. By leveraging all available data/market intelligence, improving the forecast quality significantly,and applying methods Artificial Intelligence, such as demand sensing to account for systematic changes/trends, or Computer vision to count inventory and check for quality, the service level can increase dramatically , lost sales will decrease significantly and overall better supply chain efficiency is achieved.
Artificial Intelligence in Egypt and the Middle East could have a wide impact on inventory management, due to the region’s supply chain challenges. Furthermore, many of the region businesses are using more of tribal experience in managing their supply chains, rather than using calculability, predictability and tech-based inventory management. For that matter, introducing A.I optimization capability in the Egyptian and regional supply chains can prove to be a great investment, towards creating more efficient and synchronized supply chains.
All that's mentioned in this piece, is a quick win for artificial intelligence, that can be deployed in the span of weeks or a few months, depending on the complexity and size of the business. The current time provides a unique opportunity to clear the ambiguity and better control the uncertainties of the supply chains.
Finally, I’ll leave you with the below distinction between an Ad-Hoc inventory and a smart and dynamic inventory management in the below table.
Founder & CEO of Synapse Analytics Ahmed Abaza on how AI is changing inventory management. Synapse is a data science and AI company based in Egypt. Synapse Analytics
Want to make your operations A.I. powered?