We are breaking down Agentic AI’s evolution for supply chain planning use cases into three main buckets:

1. Enhancing and sustaining data quality
2. Model explainability
3. Autonomous capabilities / Planner co-pilot

Enhancing and sustaining data quality

Going with the first principle, data cleansing and completeness is one of agentic AI’s most promising and foundational use cases. Without reasonable data, the subsequent steps are sub-optimal.

Data cleansing

Agentic AI can autonomously clean and prepare input data. Agents can detect and fix inconsistencies, errors, and outliers in sales, price, and inventory data.

• Format standardization: Ensure consistent data formats across different sources and systems.
• Deduplication: Identify and remove duplicate records to prevent data inflation and distortion.
• Missing data handling: Address missing values using statistical methods or machine learning algorithms to fill gaps.
• Data validation: Implement and enforce validation rules to ensure data integrity and consistency.

Anomaly detection

Agentic AI can identify unusual patterns or outliers in sales history, price, and inventory data through various techniques.

Statistical methods: Use probability distributions to model expected behaviour and flag significant deviations.

• Machine learning algorithms: Unsupervised learning techniques can detect patterns and anomalies without labelled data.
• Local Outlier Factor (LOF): This algorithm examines the local density of data points to identify outliers with lower density than their neighbours.
• K-Nearest Neighbors (kNN): Use kNN to classify data points into normal and abnormal ranges, working well for small and large datasets.

TrueGradient’s home page depicting an agent at work

Model explainability

Agentic AI will play a crucial in model explainability for demand forecasting and inventory optimization results. This is particularly important as AI systems become more complex and autonomous in their decision-making processes. Figure-1 is TrueGradient’s home page depicting an agent for this use-case.

Explainability in demand forecasting

Agents will improve the transparency of demand forecasting models by providing detailed breakdowns of factors influencing predictions, such as historical sales data, market trends, and external variables. It will generate natural language explanations for forecast results and translate complex statistical data into understandable insights. It will help visualize decision trees or feature importancegraphs to illustrate how different variables contribute to the final forecast.

Inventory optimization explainability

Agents will enhance explainability for inventory optimization by offering clear rationales for suggested stock levels, considering lead times, demand variability, and storage costs. It will demonstrate the impact of various scenarios on inventory decisions through interactive simulations. It will provide audit trails of decision-making processes, allowing stakeholders to trace the logic behind specific optimization recommendations.

Note: The non-deterministic nature of AI systems can make it difficult to provide simple explanations for decisions. We will discuss the mitigation approach in our next article.

Autonomous capabilities / Planner co-pilot

Agentic AI will become an assistant to the planner aka planner co-pilot.

Automated reordering

Agents will create and execute reorders autonomously based on system alerts and predefined thresholds. These agents will continuously monitor inventory levels in real-time and automatically place orders for new stock when levels drop below-specified points.

Intelligent inventory movement

Agents will analyze inventory levels across multiple locations and suggest optimal inventory movements to address stock imbalances. For example, if one store is experiencing an out-of-stock situation for a particular item, the agent can recommend transferring inventory from another location with excess stock. This proactive approach will help maintain consistent product availability across the network.

Dynamic pricing adjustments

Agents will be able to autonomously adjust pricing strategies in response to inventory levels and demand fluctuations. For instance, if certain items are overstocked, the agent might suggest temporary price reductions to accelerate sales and prevent excess inventory.

In our next article, we will go deeper into these use cases and discuss their practical challenges and mitigation approaches.

At TrueGradient, we are innovating to empower the supply chain planning community. If you want to learn more about LLM / Gen-AI / Agentic AI — feel free to contact us!

Concern #1: My data is all over the place!

Data resides in different systems — ERP, BI tool, Excel sheets, Marketplaces, etc.

Solution: Ready-made connectors (APIs) and auto-data model creation can pull and stitch data together to build an underlying data model for AI use cases. The source data schema is not a limitation.This technology is more straightforward to implement now than before! The effort is in days not weeks or months!

Concern #2: My data quality is poor!

Imputation techniques can cleanse and normalize data sets. Outliers can be auto corrected using anomaly detection models. A typical example is out-of-stock situations in shipment history, it can be imputed using order fill-rate information. Planners can review outlier / anomalies detected by the model and approve the corrected values suggested by the model.

Concern #3: I don’t have this data!

Don’t get bogged down with external data sets, e.g. weather, GDP, macros, etc. They may not be statistically significant (relevant) for your business or use case in the first place. Start with fundamentals. Most essential data sets exist within the four walls of your organization. Yes, it needs to be organized and captured at the right level of granularity. Auto-data models help stitch the data from various sources. External data relevant to your business (statistically significant) can be collected and ingested after all internal data is leveraged.

At TrueGradient, we help our customers address challenges associated with data readiness by leveraging modern technologies and fundamental data cleansing techniques.

Learn more?

Schedule a demo.

https://truegradient.ai/product

Enter the Dynamic Duo

This is where our superheroes swoop in: Demand Forecasting and Inventory Optimization. But here’s the kicker — they’re not solo acts. Nope, these two work best when they’re tag-teaming your supply chain challenges.

Demand Forecasting: The Guide

Let’s bust a myth right off the bat — demand forecasting isn’t a crystal ball. In the real world, demand forecasting errors are unavoidable. Here’s the real deal for Mid-market CPG companies:

1. Erratic Retail Orders: Companies often face wildly unpredictable orders from retailers. One week it’s crickets, the next it’s a tsunami of demand. This erratic pattern makes perfect forecasting nearly impossible.

2. Forecasting Error is Inevitable: Due to the nature of retail orders, evolving consumer preferences, competition, long tail, new products, promotions, and market volatility, there will always be a degree of forecasting error. Yes, advanced AI/ML techniques (including probabilistic models and Neural Nets) have made significant progress, however NO AI model in the world can deliver zero forecast error. Accepting this is the first step to a more realistic approach.

3. A Starting Point, Not the Finish Line: Despite its imperfections, statistical demand forecasting provides a crucial starting point.

So, if demand forecasting isn’t perfect, why bother? Because it sets the stage for our next superhero…

Inventory Optimization: The Agile Responder

This is where inventory optimization truly shines. It’s all about having the right response to the inevitable forecasting errors and demand volatility. For companies, this means:

1. Dynamic Safety Stock: Calculating and adjusting safety stock levels based on demand variability and service level targets.

2. Flexible Reorder Points: Implementing dynamic reorder points that adapt to changing demand patterns and lead times.

3. Risk Mitigation: Balancing the risks of stockouts against the costs of excess inventory.

4. Rapid Response: Quickly adjusting inventory levels in response to unexpected demand spikes or drops.

Better Together: The Power of Integration

Now, here’s where the magic happens. When you combine demand forecasting with inventory optimization, you’re creating a system that’s both proactive and reactive:

1. Statistical forecast goes as input to Collaborative / Consensus Planning.

2. Planner reviews and enriches the forecast for high touch items or alert-based enrichment.

3. Consensus forecast goes into the inventory optimization model.

4. The optimization model calculates appropriate safety stock levels and reorder points, considering the inherent forecast uncertainty.

5. When those erratic retailer orders come in (because let’s face it, they will), your system is primed to respond quickly and efficiently.

🌟 Are you a CPG industry supply chain professional? Reach out to discover how TrueGradient can help you with demand and inventory planning.

Click for demo today or email us at info@truegradient.ai !

Guide to driving pre-season and in-season planning for e-commerce

Pre-season and in-season planning are critical processes in the fashion industry, enabling supply chain professionals to effectively manage inventory, forecast demand, and respond to market trends for their business.

Pre-Season Planning

Pre-season planning typically occurs six to twelve months before a new selling season. It includes-

• Merchandise Financial Planning
• Assortment Planning
• Size/Pack Optimization

Merchandise Financial Planning

Merchandise Financial Planning (MFP) is the strategic process of budgeting and allocating financial resources for purchasing inventory, managing assortments, and optimizing sales to achieve business objectives in the retail industry. It involves anticipating demand through forecasting and planning inventory needs to meet the forecasted demand while aligning with the company’s financial goals.

The goal of MFP is to optimize inventory investments, enhance sales forecasting accuracy, and improve overall profitability by aligning financial resources with consumer demand and merchandising strategies.

Assortment Planning

Assortment planning is the strategic process of selecting the optimal mix of products across different locations, channels, and seasons and determining which specific products, product categories, styles, sizes, colors etc. to include in the assortment based on factors like profitability, brand positioning, shelf space constraints etc.

The key objectives are to maximize sales, profitability, inventory productivity, and customer satisfaction by offering the right product assortment tailored to each market segment.

Size/Pack Optimization

Size or pack optimization, also known as prepack optimization, is the process of determining the optimal pack sizes or configurations in which products should be shipped from distribution centers to retail stores.

The key objective is to minimize total supply chain costs related to handling and inventory and improve inventory productivity with right pack sizes.

In-Season Planning

In-season planning focuses on managing and adjusting plans during the current selling season. It includes-

• Initial Allocation
• Replenishment
• Exit Strategy / Markdown

Initial Allocation

Initial allocation in fashion retail refers to the process of pre-allocating planned quantities of merchandise to different locations/channels based on anticipated demand, local preferences, and trends, before the selling season begins.

The goal is to ensure the right products are available in the right locations to meet localized customer demand from the start of the season.

Replenishment

Replenishment in-season for fashion products refers to the process of restocking inventory during the selling season based on actual sales data and demand trends.It involves monitoring real-time sales data and inventory levels during the season to identify fast/slow selling items and locations.Based on this analysis, additional replenishment orders are placed with suppliers to restock high-demand products at specific stores/channels.

The goal is to maximize sales by ensuring high-selling items remain in stock, while minimizing overstock of slow movers.

Effective in-season replenishment can improve inventory productivity, reduce markdowns, and boost revenue and margins for retailers.

Exit Strategy / Markdown

Markdowns are price reductions implemented by fashion retailers to clear slow-moving or aged seasonal inventory before it becomes obsolete.

The aim is to boost sales velocity, maximize revenue over the product lifecycle, and maintain healthy inventory turnover. Markdowns can be applied in-season through progressive discounts or at the end of a season.

🌟 Are you a fashion industry supply chain professional? Reach out to discover how TrueGradient can help you create pre-season planning strategies.

Click for demo today or email us at info@truegradient.ai !

In the complex landscape of consumer-packaged goods (CPG), aligning supply with marketing initiatives, particularly during promotions, presents a significant challenge. The unpredictable nature of promotional success can lead to either surplus inventory or stockouts, disrupting the supply chain. However, innovative technological solutions are emerging to tackle these challenges, offering unprecedented accuracy and adaptability in demand forecasting and promotion optimization.

Promotional Forecasting

Advancements in AI/ML have revolutionized promotional forecasting by enabling companies to simulate various strategies and predict their impact on demand. These systems analyze vast datasets, including historical sales data, consumer behavior, and external factors like market trends and competitor activities, providing insights into the potential success of promotional campaigns. This allows businesses to optimize their promotional strategies, ensuring they allocate resources efficiently and maintain adequate stock levels to meet increased demand during promotions.

An Optimal promotion balances the sell-thru and margin goals. Advanced algorithms can have multi-objective function, the output (Promotion price or Markdown price) aims to maximize revenue and maximize profit.

Dynamic Pricing and Demand Shaping

The capability to implement dynamic pricing strategies is another innovative solution that helps align supply with marketing initiatives. By analyzing real-time data, these systems can adjust prices based on demand fluctuations, competitor pricing, and market conditions, maximizing revenue and minimizing the risk of overstocking or stockouts. Additionally, they assist in demand shaping by simulating the effects of different promotional scenarios, such as bundling products or offering targeted discounts, to influence consumer purchasing behavior.

Enhanced Supply Chain Synchronization

Enhance supply chain synchronization by integrating demand forecasting with supply chain operations. This integration allows businesses to consider factors such as lead times, transportation constraints, and supplier reliability when planning inventory levels. By providing a more synchronized approach, these tools ensure that supply chain operations are closely aligned with predicted demand patterns, reducing the likelihood of disruptions during promotional periods.

Overcoming Implementation Challenges

Despite the clear benefits, implementing these advanced systems in supply chain and promotional planning comes with challenges. Businesses must navigate the complexity of integrating new technologies with existing infrastructure. However, those who successfully adopt these solutions often see significant improvements in cost efficiency, inventory management, and customer satisfaction.

In conclusion, these advanced technologies offer a sophisticated approach to solving the promotion puzzle in CPG supply chains. By leveraging advanced insights, companies can better align their supply with marketing initiatives, ultimately enhancing their competitive edge in a dynamic market landscape.

TrueGradient AI was founded to address time-to-value gaps in Supply Chain Planning. We are committed to democratizing AI and making advanced techniques like Deep Learning, auto-ML, and Stochastic Optimization accessible and easy to adopt for all supply chain professionals and organizations.

If you’d like to learn more, please visit our website and sign up for a demo! https://truegradient.ai/product

Here are key takeaways from a fascinating research paper by Ankur Verma. The research introduces an innovative approach to optimizing promotional offers in retail using artificial intelligence techniques.

The paper addresses the need for personalized marketing. Personalized offers can help retailers balance promotional spending and profits by targeting customers with suitable discounts at the right time.

The challenge

Traditional methods of managing promotions manually are complex and often inefficient. With increasing online shopping and market competition, retailers need innovative ways to balance promotional spending with sales and profits.

The solution

The solution involves using a non-linear machine learning model called a Temporal Convolutional Network (TCN), a type of sequence-based neural network, to predict how likely a customer is to purchase an item at a given time. The time window is decided based on the nature of the business and the customer’s purchase pattern. This prediction helps set optimal discount levels for each customer and item combination.

Predict which customers are likely to buy specific items.

Determine how different discount levels affect purchase likelihood.

Suggest optimal discount offers for each customer-item combination.

How does it work?

Optimization for profit: By understanding how different discount levels affect purchase probabilities, the model estimates the “offer-elasticity,” which is how sensitive a customer’s purchase decision is to changes in discount levels. This information is used to optimize discounts to maximize revenue and customer retention.

Data utilization: The model uses historical transaction data to make predictions. It considers factors like past purchase behavior, customer demographics, and item characteristics to tailor offers effectively.

Modeling methodology: The model treats each consumer-item pair as an individual object, using bi-weekly time series data to predict purchase probabilities. Features are generated from transactional data, including temporal and static attributes of consumers and items. The model architecture combines entity embeddings with TCN, using dilated convolutions to handle sequential data without information leakage from the future to the past. It uses F1-maximization to determine the suitable probability cut-off for each customer basis the probability distribution.Techniques like RMSProp, Adam optimizers, and Stochastic Weight Averaging (SWA) are used for model optimization.

Business impact: Implementing this AI-driven approach can significantly enhance a retailer’s bottom line by ensuring that promotions are practical and efficient, leading to increased sales and improved customer loyalty.

In summary, the paper highlights the potential of leveraging AI to refine promotional strategies, which would ultimately benefit the business by optimizing promotional spending and enhancing customer engagement.

The complete paper is available here: https://arxiv.org/abs/2010.08130

TrueGradient AI was founded to address time-to-value gaps in Supply Chain Planning. We are committed to democratizing AI and making advanced techniques like Deep Learning, auto-ML, and Stochastic Optimization accessible and easy to adopt for all supply chain professionals and organizations.

If you’d like to learn more, please visit our website and sign up for a demo! https://truegradient.ai/product

CPG companies face a significant challenge in achieving accurate forecasts for Amazon’s purchase orders, with industry leaders often operating at less than 50% accuracy. This issue is exacerbated by Amazon’s unique marketplace dynamics, where products are identified by Amazon Standard Identification Numbers (ASINs), distinct from the UPCs used by CPG companies. Furthermore, Amazon’s competitive marketplace means multiple sellers may offer the same ASIN, with only one seller featured in the buy box at any given time, determined by factors like price, stock availability, and seller reputation.

To enhance demand forecast accuracy for Amazon, the top 3 levers are:

1. Consider Requested Delivery Date (RDD) for Demand Forecasting

Unlike traditional shipment date-based forecasting, which may not align with customer demand timelines, focusing on requested delivery dates provides a clearer picture of actual demand. This approach considers the time dimension of demand forecasting, essential for maintaining high order fill rates and customer satisfaction.

2. Cleanse (Impute) Order History to Address Amazon’s Reordering System

Amazon’s automated reordering system triggers additional orders when CPG companies fail to fulfill initial orders promptly. This can lead to inflated demand forecasts if not correctly managed. To mitigate this:

Identify reorder instances by analyzing order quantities that deviate from historical patterns.

Treat these reorders as outliers in forecasting models and use statistical techniques to adjust forecasts accordingly.

Implement imputation strategies leveraging machine learning models to smooth out the impact of these outliers and improve forecast accuracy.

3. Leverage Rich Data Sets Provided by Amazon

Amazon offers a wealth of data that can significantly enhance forecasting capabilities:

Glance Views and Unique Visitors: These metrics provide insights into product visibility and customer interest, which are crucial for understanding demand drivers.

Customer Reviews and Rankings: Weekly updates on reviews and rankings help gauge customer sentiment and product performance relative to competitors.

Buy Box Metrics: Lost Buy Box (LBB) and Rep OOS (out of stock) indicate pricing competitiveness and stock availability dynamics, influencing demand forecasts.

Sellable and Unsellable Units: Detailed inventory data aids in adjusting forecasts based on actual stock availability.

CPG companies can use artificial intelligence and machine learning to process Amazon's complex data sets and generate more accurate forecasts.

By focusing on these strategies, CPG companies can mitigate the challenges posed by Amazon’s marketplace dynamics and achieve higher forecast accuracy, ultimately improving operational efficiency and customer satisfaction. These efforts enhance supply chain management and strengthen competitive positioning in the evolving e-commerce landscape.

Partner with TrueGradient to unlock your full potential.

Click for a demo today or email us at info@truegradient.ai

https://truegradient.ai/product

Let’s examine the top 5 challenges reported to us by the planning community in the Consumer Packaged Goods (CPG) industry.

1. Cracking the Amazon Code: Demand Forecasting

The exponential growth of e-commerce, particularly through Amazon, has introduced complex forecasting challenges. Traditional methods are proving less effective in the face of dynamic online sales, which are influenced by factors such as algorithmic pricing and consumer reviews.

2. Crystal Ball Needed: New Product Forecasting

Predicting demand for new products remains a significant hurdle due to the absence of historical data and uncertain market reception. This challenge is further intensified by the accelerating pace of product innovation, requiring more agile and adaptive forecasting methodologies.

3. Keeping Shelves Stocked: The Fill-rate Challenge

Ensuring high fill rates is crucial to prevent retailer penalties and mitigate lost sales opportunities. Recent supply chain disruptions and demand volatility have exacerbated this challenge, necessitating more robust inventory management strategies.

4. The Promotion Puzzle: Aligning Supply with Marketing Initiatives

Supply chain planning leaders face significant challenges when it comes to aligning inventory with promotional activities. While marketing departments invest substantially in promotional campaigns, accurately forecasting their impact on demand remains a complex task for supply chain teams. The unpredictable nature of promotional success creates a ripple effect throughout the supply chain. Overestimating demand can lead to excess inventory and increased holding costs, while underestimating can result in stockout and lost sales opportunities.

5. The Goldilocks Dilemma: Just-Right Inventory Levels

Striking the optimal balance between meeting demand and minimizing working capital tied up in inventory remains a persistent challenge. Planners are continually refining their inventory management practices to achieve this equilibrium.

To address these challenges comprehensively, CPG companies are implementing several strategic initiatives:

1. Digital Transformation: Investing in advanced technologies for end-to-end visibility and real-time data analysis.
2. Agile Planning Processes: Adopting more flexible planning methodologies to swiftly adapt to market changes and disruptions.
3. Artificial Intelligence: Leveraging deep learning and sophisticated analytics to gain deeper insights into consumer behaviour, market trends, forecasting and supply chain optimization.
4. Enhanced Supplier Collaboration: Strengthening supplier relationships and implementing collaborative planning processes to boost supply chain resilience.

In subsequent articles, we (TrueGradient AI team) will delve deeper into solution approaches from Planning and AI perspectives to tackle these pressing challenges.

Learn more?

Click for demo today or email us at info@truegradient.ai

https://truegradient.ai/product

🌟 Ready to Unlock Your Potential? Reach out to discover how TrueGradient can solve your supply chain pain points.

Click for demo today or email us at info@truegradient.ai !

🌟 Ready to Unlock Your Potential? Reach out to discover how TrueGradient can solve your supply chain pain points.

Click for demo today or email us at info@truegradient.ai !