Forecasting supplier lead time to optimize supply chain operations

Solve supply chain challenges with more-accurate, data-driven lead time forecasts

The COVID-19 pandemic disrupted consumer habits. Shortages forced people to try new brands, and particularly in the grocery sector, many people shopped less frequently but often bought more when they did. As a result, inventory started moving out of stores faster, straining both supply chains and financial models and causing gross margin problems.

At the same time, the cost of being out of stock has increased and replenishment problems can impact profitability and overall business success. Consumers are simply less tolerant of empty shelves when they have virtually instant access to pricing and product availability from an increasing number of competitors who can deliver services and products in multiple ways to meet their needs. In fact, 29% of consumers say out-of-stock items would drive them to shop at another brand.

The challenge for retailers is to consistently satisfy customers who want to find the quantity of merchandise they want both where they want and when they want. To successfully achieve their financial goals, retailers must strategically manage the inventory they carry at every point in the supply chain and make sure the replenishment process is always smooth and efficient.

Forecasting supplier lead time—predicting the amount of time it will take for a supplier to deliver a product or service after an order is placed—helps retailers plan their production schedules and manage inventory levels to effectively meet customer demand while minimizing excess inventory and its associated costs.

The lead time for a supplier depends on various factors, such as the distance of the supplier from the product’s destination, the complexity of the product, the availability of raw materials, production capacity, and transportation time, among others. Because of the number of variables, retailers need a data platform that provides them with centralized access to historical and real-time data from a range of enterprise systems, business records, and technical inputs, which can then be used to train machine learning models to forecast expected lead times based on purchase order transactions.

Derisk supply chain operations and improve inventory management with advanced analytics and machine learning

In this use case, we’ll demonstrate how Oracle Data Platform is built to help retailers use advanced analytics and forecasting methods (including statistical modeling, trend analysis, and historical data analysis) and machine learning to accurately estimate the expected delivery dates of goods. With this information, retailers can optimize inventory planning and effectively manage the impact of variables such as

• Lead times and transportation—including coordinating source availability, shipping schedules, travel times, and costs
• Diverse product portfolios—including the challenges of managing a wide range of products, availability, pack configurations, ordering terms, and costs across hundreds of suppliers
• Local market complexities—including demand patterns and influences such as seasonality and promotions
• Financial and physical constraints—including budgets, storage limitations, and desired turns
• Inventory pressure at fulfillment locations—including the financial impact of overstocks and markdowns, the pressure to consistently deliver excellent customer service, and the need to maintain availability to prevent lost sales and the erosion of customer loyalty

This image shows how Oracle Data Platform for retail can be used to forecast supplier lead times and optimize supply chain operations, helping retailers maintain market position while maximizing profitability. The platform includes the following five pillars:

1. 1. Data Sources, Discovery
2. 2. Ingest, Transform
3. 3. Persist, Curate, Create
4. 4. Analyze, Learn, Predict
5. 5. Measure, Act

The Data Sources, Discovery pillar includes three categories of data.

1. 1. Application data comes from Fusion Financials, Oracle E-Business Suite, SCM, EPM, and eSourcing.
2. 2. Business record data comprises inventory, SCM (control tower), supplier performance data, and supplier surveys.
3. 3. Technical input data comes from logs.

The Ingest, Transform pillar comprises three capabilities.

1. 1. Batch ingestion uses OCI Data Integration, Oracle Data Integrator, and DB tools.
2. 2. Bulk transfer uses OCI FastConnect, OCI Data Transfer, MFT, and OCI CLI.
3. 3. Change data capture uses OCI GoldenGate.

All three capabilities connect unidirectionally into the cloud storage capability within the Persist, Curate, Create pillar.

The Persist, Curate, Create pillar comprises four capabilities.

1. 1. The serving data store uses Oracle Autonomous Data Warehouse or Exadata Cloud Service.
2. 2. Cloud storage uses OCI Object Storage.
3. 3. Batch processing uses OCI Data Flow.
4. 4. Governance uses OCI Data Catalog.

These capabilities are connected within the pillar. Cloud storage is unidirectionally connected to the serving data store; it is also bidirectionally connected to batch processing.

One capability connects into the Analyze, Learn, Predict pillar: The serving data store connects to both the analytics and visualization capability and machine learning capability.

The Analyze, Learn, Predict pillar comprises three capabilities.

1. 1. Analytics and visualization uses Oracle Analytics Cloud, GraphStudio, and ISVs.
2. 2. Data Products, APIs uses OCI API Gateway and OCI Functions
3. 3. Machine learning uses OCI Data Science, Oracle ML, and Oracle ML Notebooks.

The Measure, Act pillar comprises three consumers: dashboard and reports, applications, and models.

Dashboards and reports comprise people and partners, supplier collaboration and data sharing, historical supplier performance, demand analytics, and stockouts and overstocks.

Applications comprises advanced inventory management, and demand planning.

Models comprises supplier operations./p>

The three central pillars—Ingest, Transform; Persist, Curate, Create; and Analyze, Learn, Predict—are supported by infrastructure, network, security, and IAM.

There are three main ways to inject data into an architecture to enable retailers to effectively forecast supplier lead time.

• To start, we need to understand our overall inventory position to ensure products aren’t overstocked or understocked. To do so, we use Oracle Cloud Infrastructure (OCI) GoldenGate to enable change data capture ingestion of near real-time warehouse inventory data from operational databases for all or a subset of product lines. We can then use this data to adjust prices to either move inventory or avoid a stockout.
• To accurately predict supplier performance, we also need to understand historical performance, trends, and patterns. This typically requires loading a large volume of transactional data (including ERP data, such as procurement, invoicing, supply chain, and logistics data) and other operational metrics and datasets (such as data on consumption, inventory, and hot swaps) from on-premises data stores using bulk transfer methods and services, such as OCI Data Transfer Service.
• We can now use batch ingestion to add datasets relevant to suppliers, such as orders placed with the supplier over a specific time period, including the date of the order, the quantity ordered, and the delivery date. These datasets often comprise large volumes of typically on-premises data, and in most cases, batch ingestion is sufficient and most efficient. For our supplier data, we’ll use Oracle Data Integrator to ingest the data on a daily cycle. This data is primarily sourced from operational transaction process systems and normally modeled in highly structured relational form. Examples of this data include purchase order transactions, including supplier details (for example, their name, ID, registration, and contact information), the origin and destination, agreed delivery date, actual delivery date, contract items and price, shipment method, and so on. Data on the supplier's performance, including their delivery reliability, the quality of their goods or services, and any delays or issues that have occurred in the past can also be ingested, although this data is typically less structured and may require a greater degree of processing.
• By calculating the lead time for each order previously placed with the supplier, we can calculate an average lead time and identify trends and variations. These trends and variations can be correlated with external factors that could impact the supplier's lead time, such as transportation delays, changes in the supplier's production capacity, environmental events (such as severe weather), or sociopolitical events (such as conflict or industrial action). Additional data can be used to monitor market trends and demand patterns to anticipate potential spikes in demand that could impact the supplier's lead time.

Data persistence and processing is built on three components.

• The ingested raw data from all sources is stored in cloud storage. We will use OCI Data Flow for the batch processing of this now persisted data, stock levels, geo-mapping data, and product reference data. The batch processing will reprocess the data and remove any duplicates, missing values, or outliers that could skew the analysis. These processed datasets are returned to cloud storage for onward persistence, curation, and analysis and ultimately for loading in optimized form to the serving data store in a format that can be easily analyzed.
• We have now created processed datasets that are ready to be persisted in optimized relational form for curation and query performance in the serving data store provided by Oracle Autonomous Data Warehouse. This will enable us to identify and return the products by price, demand profile, inventory level, and location.

The ability to analyze, learn, and predict is built on three technologies.

• Analytics and visualization services allow us to use statistical techniques, such as regression analysis and time series analysis, and machine learning algorithms to identify patterns and trends in the data. Using this analysis, we can then develop a forecast model that can accurately predict the supplier's lead time and continually validate the accuracy of the model by comparing the predicted lead times with actual lead times for a set of orders. The results of this validation will be used to refine the model and improve its accuracy. Our analytics and visualization services include the following capabilities:
  
  
  • Descriptive analytics describes current trends with histograms and charts and supports the development of pricing algorithms that use predefined rules to adjust prices based on specific criteria, such as sales performance, inventory levels, or competitor pricing. For example, a retailer may set a rule to decrease the price of a product by 10% if it has been in stock for more than 30 days and delay the purchase of new inventory or negotiate a price for a later delivery using lead time forecasts to determine the appropriate timing.
  • Predictive analytics predicts future events, identifies trends, and determines the probability of uncertain outcomes. With predictive analytics, retailers can use historical sales data to identify correlations between price and demand. They can then use this analysis to predict how changes in consumer behavior will affect demand and adjust inventory plans accordingly, using estimated lead times to help ensure they have enough stock on hand when they need it while minimizing excess inventory and its associated costs. Additionally, predictive analytics can provide price elasticity models, which use statistical models to measure how sensitive demand is to changes in price. Retailers can use this analysis to identify the optimal stock level points to maximize sales and profitability and time their inventory purchases accordingly.
  • Prescriptive analytics proposes suitable actions to support optimal decision-making and can be used for lead time forecasting to help minimize the costs associated with holding inventory and stockouts. By aligning procurement and production activities with supplier lead times, retailers can reduce excess inventory, carrying costs, and the expense of expedited shipping and better negotiate pricing and terms with suppliers based on accurate lead times.
• Alongside the use of advanced analytics, machine learning models are developed, trained, and deployed. These models use artificial intelligence to analyze large amounts of data and identify patterns and trends that can be used to optimize inventory purchases and stock levels. Retailers can use machine learning algorithms to predict customer behavior, identify when to purchase inventory and from which suppliers, and optimize prices across multiple products and markets.
• Our curated, tested, and high-quality data and models can have governance rules and policies applied and can be exposed as a “data product” (API) within a data mesh architecture for distribution across the retail organization.

Improve inventory management and customer satisfaction with a retail data platform

By accurately forecasting supplier lead time, retailers can better plan their inventory levels and production schedules to ensure they have the right products available in the right amounts to meet customer demand, even as it fluctuates based on seasonality, promotions, and other influences. As a result, they’re able to

• Identify when to purchase inventory and from which suppliers
• Minimize their inventory carrying costs by ordering the right quantities of products at the right time and avoid understocking, which can result in lost sales and dissatisfied customers
• Manage their cash flow by planning their purchases and payments to suppliers, helping them optimize their working capital and avoid cash flow shortages
• Build stronger relationships with their suppliers through better communication about lead times and other performance metrics, which can lead to improved performance, better pricing, and more-reliable delivery schedules