Jaymalya (“Jay”) Deb is currently the Division Materials Manager in the Industrial Process Filtration Division at Parker Hannifin, a Fortune 250 company and global leader specializing in motion and control technologies, including filtration, hydraulics, pneumatics, electromechanical, and other related tools for the industrial, mobile, energy, and aerospace industries, where he is responsible for supervising the Materials teams in the division. Jay applies more than 20 years of leadership experience and technological expertise in supply chain and commodity management, engineering, IT, and manufacturing. 

Read also: Investing in Supply Chain Technology Will Give You a Competitive Edge

Jay previously served as an Oracle Consultant to GE Aviation, where he deployed his significant expertise in Oracle R12 to solution design by implementing the Advanced Supply Chain Planning module (ASCP) and configuring the Manufacturing Execution System (MES) module. Prior to that, he furthered his supply chain career, leading materials management, operations management, and supplier development for a US-based global provider of upstream equipment and technology to the oil and gas industry. 

In his current role, Jay now leverages his expertise in advanced supply chain technologies to design systems and processes that improve forecasting, production, demand planning, delivery, and profitability outcomes. Jay is recognized for designing innovative solutions across diverse functional groups that deliver measurable organizational improvements. Among his recent accomplishments, he developed complex statistical models for demand planning that utilized machine learning techniques to improve finished goods delivery lead times, and built an automated shop floor scheduling system which continuously optimizes the available capacity against released work orders to achieve positive on-time outcomes for the organization. 

Jay received a bachelor’s degree in Mechanical Engineering from R.V. College of Engineering in Bangalore, India. After relocating to the US two decades ago, he earned an M.S. in the Manufacturing Systems Engineering Program (MSEP) from the University of Wisconsin-Madison, and an M.B.A. in Management from the University of Texas at Austin-The McCombs School of Business. Jay is the author of “Managerial Perspective to Operational Excellence: Using Lean Ideas to Compete Against Low-Cost Countries,” a book that presents a case-based approach to Lean implementation and change management, including examples of Robotic Process Automation concepts used in manufacturing and analytical tools used to solve operations management problems. He has also written about his innovative machine learning approach to improve forecasting for intermittent demand industrial products in the oil and gas industry. 

We spoke with Jay about current challenges in demand planning and supply chain management, the impact of accelerating technologies, his approach to solution design, and his forecast for the industry. 

Ellen Warren: Jay, in your career, you have led various supply chain optimization initiatives across different industries, including oil and gas and industrial filtration. What do you see as the most pressing challenge in supply chain management today?

Jaymalya Deb: Supply chain resilience amid global disruptions has emerged as the most pressing challenge facing the industry today. The escalating impact of external shocks—geopolitical tensions, trade tariffs, climate-induced delays, and raw material shortages—ripples through the entire supply chain, creating widespread instability. Notably, sudden supplier disruptions, particularly within the oil and gas sector, have driven a 15% year-over-year increase in supply chain interruptions, as evidenced by recent data. In the oil and gas industrial market, where intermittent demand already poses significant planning challenges, these disruptions intensify the demand for robust contingency measures. Concurrently, the scarcity of reliable data forces manufacturers to maintain elevated inventory levels to meet stringent service requirements. This approach becomes increasingly untenable as customers, unable to anticipate their own needs, seek alternatives when stock is unavailable, thereby driving up holding costs and undermining customer loyalty.

EW: In your experience, how is technology, particularly machine learning, changing the way companies respond to these challenges? How have you personally adopted new technologies in designing enterprise solutions? 

JD: Technology, and particularly machine learning (ML), is revolutionizing how companies address supply chain challenges such as resilience amid global disruptions and data scarcity. ML enhances demand forecasting by analyzing vast, unstructured datasets—including historical demand patterns, real-time market signals, and external factors like geopolitical risks or climate data—enabling more accurate predictions even for intermittent demand scenarios, such as those in industrial oil and gas products. Advanced algorithms identify demand likelihood and optimize inventory levels, reducing overstocking while improving service levels. Additionally, ML-driven predictive analytics supports resilience by modeling supply chain risks, allowing companies to simulate disruption scenarios and develop agile contingency plans. Real-time insights from news sources highlight a growing adoption of these tools, with firms reporting up to 20% improvements in supply chain efficiency..

Personally, as a senior supply chain manager, I have embraced these technologies to design enterprise solutions tailored to oil and gas industrial products. I pioneered a hybrid forecasting framework that integrates advanced methods to refine demand predictions based on sparse data. This approach, informed by my field insights, has been implemented across multiple organizations, enhancing inventory optimization and reducing stockouts to the extent of 20-22%. I also incorporated real-time data feeds and scenario analysis tools to build resilient supply chains, enabling dynamic reordering and proactive factory adjustments. These innovations, validated through multiple years of data, reflect my commitment to leveraging cutting-edge technology to address pressing industry challenges. 

EW: You recently developed an innovative hybrid machine learning approach for forecasting intermittent demand in oil and gas. Can you tell us why traditional forecasting methods fail in these scenarios, and how your hybrid model solves this problem more effectively?

JD: Traditional forecasting methods such as Simple Moving Average, Exponential Smoothing, and ARIMA often fall short in scenarios involving intermittent demand in the oil and gas industrial products sector due to their inherent limitations. These methods excel with stable, continuous demand patterns, but struggle with the sporadic, zero-heavy series typical of industrial products. Specifically, they exhibit an upward bias following demand spikes, fail to adequately handle prolonged zero-demand periods, and lack the capability to incorporate complex covariates or recognize latent demand patterns, leading to inaccurate projections and excessive inventory.

My innovative hybrid ML approach addresses these shortcomings by integrating Croston’s method with advanced machine learning techniques, notably XGBoost. This framework separates demand size and inter-arrival intervals, as in Croston’s model, and enhances it with a classification model to predict demand likelihood, using historical data and external factors as features. If the predicted likelihood exceeds an optimized threshold, a regression model estimates demand magnitude. This dual approach leverages the statistical foundation of Croston while harnessing ML’s ability to process multidimensional data, resulting in forecasts that are both more precise and adaptable. Validated with multiple years of daily data, this model has demonstrated a 10-14% improvement in on-time delivery and a 20-22% reduction in stockouts, offering a more effective solution for the oil and gas industry’s intermittent demand challenges.

EW: In your current role, you led the development of a monthly forecast review process based on Sales and Operations Planning (S&OP). In this case, how does machine learning integrate into the S&OP process to improve forecast accuracy and collaboration across teams?

JD: In my roles in several organizations, I have led the development of a monthly forecast review process rooted in Sales and Operations Planning (S&OP), leveraging my experience to enhance alignment and decision-making. Machine learning (ML) integrates into the S&OP process by elevating forecast accuracy and fostering collaboration across teams. My hybrid forecasting model, analyzes historical sales data, market trends from web sources, and operational inputs to predict intermittent demand—such as in oil and gas industrial products—with double digit percentage point improvement in on-time delivery.

During the S&OP cycle, ML provides a data-driven baseline forecast, which Sales refines with customer insights, while Operations adjusts based on capacity constraints. This iterative input, facilitated by the real-time dashboards I implemented, reduces bias and enhances accuracy significantly in stockout prevention. Collaboration improves as ML-generated scenarios—e.g., demand spikes or supply risks—serve as a common language, enabling cross-functional alignment in review meetings. This integration, validated over months of implementation, ensures a proactive, unified approach to forecasting and resource planning.

EW: Your career and leadership role spans engineering, IT, and supply chain management. How do you bridge the gap between technical teams and operational teams when designing and implementing solutions that involve complex technologies like machine learning or ERP systems?

JD: My career has equipped me with a unique perspective to bridge the gap. This interdisciplinary experience allows me to act as a translator, aligning the technical expertise of data scientists and IT specialists with the practical needs of operational stakeholders.

I foster collaboration by initiating cross-functional  workshops where technical teams present simplified overviews of ML algorithms or ERP functionalities, while operational teams articulate real-world challenges, such as intermittent demand for industrial components related to oil and gas. For instance, when developing my hybrid ML forecasting model, I ensured focus on algorithm accuracy, while supply chain managers highlighted inventory constraints, creating a shared understanding. Regular feedback loops, including pilot testing with operational data, further refine these solutions, ensuring usability—my model’s sustained stockout reduction, stemmed from such iterative input.

EW: Leading cross-functional teams is a significant part of your role. How do you ensure that Sales, Engineering, and Operations teams are aligned with the advanced forecasting models you implement, and how do you get buy-in from stakeholders who may not be familiar with the technical side?

JD: Leading cross-functional teams is indeed a cornerstone of my role, drawing on my long experience across engineering, IT, and supply chain management. To ensure alignment between Sales, Engineering, and Operations teams with advanced forecasting models, I employ a structured yet collaborative strategy. I begin with cross-functional workshops, where each team’s perspective is integrated—Sales provides market insights, Engineering ensures technical feasibility, and Operations highlights execution challenges. For instance, when implementing my forecasting model for oil and gas industrial products, I facilitated sessions to align Sales’ demand signals with Operations’ inventory goals, using Engineering to validate the advanced ML based forecast model integration.

To secure buy-in from stakeholders unfamiliar with the technical aspects, I focus on translating complex concepts into tangible business benefits. I use visual aids, such as simplified flowcharts or dashboards showing the model’s percentage point improvement in on-time delivery, to demystify the technology. I also conduct targeted demonstrations, like pilot runs with real data, allowing stakeholders to see reduced stockouts (20-22% in my case) firsthand. Addressing concerns through one-on-one discussions and tying outcomes to their KPIs—e.g., Sales’ revenue stability or Operations’ cost savings—further builds trust. This approach, honed over years of leadership, ensures alignment and enthusiastic adoption across the board.

EW: In your experience, how can organizations best leverage data—whether it’s from ERP systems like Oracle EBS or external sources—to drive decision-making processes and improve supply chain performance?

JD: Organizations can best leverage data—whether sourced internally or externally—by adopting a strategic, integrated approach.

First, consolidating data from ERP systems provides a robust foundation. For instance, Oracle EBS offers real-time insights into inventory, procurement, and production, which I’ve utilized to optimize reorder points in oil and gas industrial products. Integrating this with external data—such as market trends or supplier performance metrics—enriches the dataset. My hybrid forecasting model, for example, incorporates such external inputs alongside internal ERP data, improving demand accuracy and reducing stockouts.

Second, organizations should invest in advanced analytics, including machine learning, to process this data. By training models on historical ERP data and real-time external signals, firms can predict disruptions or demand spikes, as I’ve demonstrated in my work. This requires clean, standardized data, achieved through governance frameworks I’ve implemented to ensure quality.

Finally, decision-making improves with accessible visualization tools. I’ve deployed dashboards that translate complex analyses into actionable insights, enabling cross-functional teams to align on strategies. This holistic data utilization, validated by my field experience, drives proactive adjustments, enhances resilience, and boosts overall supply chain efficiency.

EW: As an expert in both machine learning and supply chain management, what’s your vision for the future of demand forecasting? Do you foresee AI-driven solutions playing a larger role in autonomously managing supply chains, and if so, how can companies begin preparing for this shift?

JD: My vision for the future of demand forecasting, shaped by my expertise in machine learning and supply chain management, centers on a seamless integration of AI-driven solutions that transform the field. I anticipate demand forecasting evolving into a highly predictive, self-adaptive process, leveraging real-time data from diverse sources—such as IoT sensors, market sentiment, and ERP systems—to anticipate demand with unprecedented precision, even for intermittent patterns like those in oil and gas industrial products. My hybrid forecasting model, which already boosts on-time delivery by several percentage points, is a precursor to this shift, and I foresee AI expanding this capability by autonomously refining algorithms and incorporating external variables like geopolitical risks.

AI-driven solutions are poised to play a larger role in autonomously managing supply chains, potentially automating inventory adjustments, reorder decisions, and supplier coordination. This could reduce human intervention by up to 30%, based on emerging trends from web analyses, while enhancing resilience against disruptions. Companies can prepare by first investing in robust data infrastructure—clean, integrated datasets from Oracle EBS or similar systems—to fuel AI models. Next, they should upskill teams through targeted training, as I’ve done with cross-functional workshops, to bridge technical and operational gaps. Finally, piloting AI tools in controlled environments, like my stockout reduction trials, will build confidence and refine processes. This proactive approach will position organizations to harness AI’s full potential in supply chain management.

EW: You’ve authored a book on Lean manufacturing and operational excellence. How do the principles in your book align with or complement the data-driven, machine learning approaches you now apply in supply chain management? How can Lean principles and advanced analytics be harmonized for optimal supply chain performance?

JD: My book on Lean manufacturing and operational excellence, drawn from my years of industry experience, emphasizes waste elimination, continuous improvement, and value stream optimization—principles that align seamlessly with the data-driven, machine learning approaches I now apply in supply chain management. Lean’s focus on reducing excess inventory and enhancing process efficiency complements my hybrid forecasting model, which minimizes stockouts and overstocking by leveraging precise demand predictions. For instance, Lean’s just-in-time philosophy is enhanced by machine learning’s ability to anticipate intermittent demand in oil and gas industrial products, aligning production with actual needs.

Harmonizing Lean principles with advanced analytics offers a powerful framework for optimal supply chain performance. Lean provides the operational discipline—streamlining workflows and fostering a culture of Kaizen—while analytics, such as the hybrid forecasting model, delivers the predictive intelligence to support these efforts. Companies can integrate them by using data to identify waste (e.g., excess inventory flagged by ERP systems) and applying Lean tools like value stream mapping to address it, guided by real-time insights. I’ve implemented this synergy by combining Lean’s pull-based systems with ML-driven reorder points, achieving significant percentage improvement in on-time delivery. This dual approach maximizes efficiency, reduces costs, and builds resilience, creating a balanced, future-ready supply chain.