1st Presentation - Overview of Demand
In Demand, Demand Unit represents the unique product identifier (can be the same as item number on a SKU).
101001-01 represents Tomato Sauce SKU. The same Item component of a SKU can also be represented as a DmdUnit component of a DFU in Demand.
An accurate forecast is a blend of the following two components: Statistical Analysis and Market Intelligence.
A DFU is the entity for which a forecast is created. It represents an item selling in a market. The key components of a DFU are Demand Unit, Demand Group, Location, and Model.
Demand uses a variety of terms that are specific to the functionality of the product. One key term is Demand Forecasting Unit (DFU).
A DFU is the entity for which you are creating a forecast. Demand Unit, Demand Group, Location, and Model. A DFU represents an item selling in a market.
A DFU is the entity for which you are creating a forecast. In simpler terms, a DFU represents an item selling in a market.
A Stock Keeping Unit (SKU) is a store's or catalog's product and service identification code, often portrayed as a machine-readable bar code that helps the item to be tracked for inventory.
Demand includes a variety of tools available to determine how accurate the statistical forecast and market intelligence has been in the past.
Actual forecast accuracy performance metrics are another way to prioritize the workload of a demand planner to make any necessary adjustments to improve forecast accuracy.
Hierarchical forecasting is a good technique when history at lower levels is too sporadic or history is too short to give a good forecast.
Aggregating to mid-levels or upper levels can reveal more accurate trends and seasonal patterns to be used for forecasting.
Provides a benchmark for continuous improvement of the forecasting process
Allows forecasters to manage by exceptions
Demand usually runs a business-defined automated process to complete the closing period step in a batch process.
Always evaluate the forecast performance. This is an important step for improving the accuracy of the total business forecasting process.
Demand Classification will also automatically select parameters for the recommended algorithms.
An exception is a message or signal that indicates that a particular model, with the current parameter settings, may not be a good predictor of future demand.
The demand planning cycle is made up of several phases to create your forecast.
As you can see, the cycle is a blend of planner-driven and automated processes as they relate to the tasks of the demand planner integrated with the automated tasks of the software.
Key Capabilities are Comprehensive set of statistical & configurable business rules for forecasting Hierarchical, multichannel forecasting Automated history cleansing for lost sales & events
Automated algorithm selection and parameter setting Streamlined consensus forecasting Integrated promotional planning & analysis
Benefits of an Accurate Forecast are: Enhanced Customer Service, Lower Inventory Costs, Lower Transportation Costs, Reduced Reactivity, Enhanced Planning Processes, Lower Manufacturing Costs, and Maximized Profitability.
Benefits of an Accurate Forecast are: Enhanced Customer Service, Lower Inventory Costs, Lower Transportation Costs, Reduced Reactivity, Enhanced Planning Processes, Lower Manufacturing Costs, and Maximized Profitability.
Prior to Demand Post Date is historical data. From the Demand Post Date forward is forecast and planning data.
Common processes performed to close the period include updating the history, storing the historical data, storing forecast performance, updating Demand Post Date, running Calculate Model with the new history data, and so on.
You can review the historical patterns in customer demand, analyze the patterns with a statistical algorithm, and project those patterns into future as forecast.
Consider the following questions while preparing your history: What does the history represent? What is included in the history?
Market Intelligence can be anything that the statistical model cannot formulate, such as: Information about new technology products, Actions of your competition, Weather conditions near your plants and stores
Current economic factors, Retailer's data regarding customer buying habits, and The popularity due to celebrity endorsement
Demand provides seven different algorithms to meet the many unique demand patterns for products, industries, customers, locations, and so on. It is not always a one size fits all.
Demand Classification can be used to help planners determine the correct algorithm to use. Right Level to Forecast can aid in the design of any DFU hierarchies.
Massive data volumes make the process of analyzing history patterns and selecting an algorithm increasingly challenging.
Demand Classification helps to identify an appropriate statistical forecasting algorithm that matches the type of history patterns.
The forecast data is adjusted to include an event that has not happened before or has changed. Examples of forecast events are promotions, gain/loss of customers, new market opportunities, and so on.
Demand allows you to gather information from sales, marketing, production, suppliers, and other sources to adjust your forecast for more accuracy.
Demand forecasting is a way to calculate or estimate customer needs in advance to ensure that the right product is at the right place, at the right time, and in the right quantity.
Demand forecasting is a way to calculate or estimate customer needs in advance to ensure that the right product is at the right place, at the right time, and in the right quantity.
How much history is stored and used? What time buckets is history stored? What unit of measure does the history represent?
Demand helps you clean the history and prepare it to give you a more accurate forecast using several different options of functionality. When your history is good, your forecast will also be good.
A DFUTOSKU map is used to transfer the demand forecast information from the four DFU components to the two SKU components needed downstream in the planning process. Typical DFU values having four components Typical SKUs representing Item@Location
Demand includes different techniques for achieving an accurate forecast for DFUs such as Manage Hierarchical Forecast, Leverage Statistical Algorithms, Manage by Exception, Incorporate Event Information, and Track and Measure Performance.
Demand is an exception driven system which allows users to prioritize their workload and determine the specific DFUs to focus on and manage.
Demand is an exception driven system which allows users to prioritize their workload and determine the specific DFUs to focus on and manage.
Parameters allow control of the key time series modeling components: Level, Trend, Seasonality, and Smoothing.
Demand provides default parameter settings that can then be evaluated and tuned to improve forecast accuracy as needed.
The forecast table will now have the final forecast available to be transferred to the rest of the supply chain to drive the business.
Demand provides ways to select the appropriate model to publish from the forecast draft table. This is because Demand has the ability to have multiple models for a Demand Unit, Demand Group, and Location combination (DFUView). You can publish only one forecast.
Demand is a powerful tool that uses the sales history to create an accurate forecast of future demand. You can then use this forecast to drive planning for distribution, manufacturing, materials, and transportation.
Demand supports critical decisions in your day-to-day activities and enables you to determine and build optimal inventory levels throughout your support chain network.
Demand helps you answer multiple questions related to your forecast and drive efficient supply chain operations
Demand uses shared database and in-memory model across Demand and Supply and provides visibility throughout the supply chain.
This process sends the finalized forecast to other processes and applications within the supply chain. This is the forecast signal that the downstream supply chain will plan and execute against.
Demand will translate the Demand DFU forecast to a Fulfillment SKU forecast. This relationship is defined in the DFUtoSKUMap table.
The reconciliation model enables comparisons of statistically generated forecasts at multiple levels. It allocates time-phased forecasts across defined hierarchy levels.
Different demand patterns require different forecasting techniques. Using the wrong algorithm and forecasting technique can lead to a lower level of forecast accuracy.
How Demand helps you for accurate forecast?
Enables goal setting based upon characteristics of products you are forecasting
Multiple level DFU forecasting includes best statistical level and intelligence across different levels.
Example: At the lowest level, the Rate of Sale is forecasted, and at the highest-level clear seasonality is forecasted.
Demand Group is a dimension of the DFU, which is often used to describe a sales channel or major customer.
Examples: Key account/customer Brand Vendor type Channel of distribution, such as Web Sales, Retail, Institutions, etc.
Exceptions are identified through Dashboards and Workbenches, Exception Graphs, Report Pages
Exceptions are identified through Dashboards and Workbenches Exception Graphs Report Pages
Hierarchies exist in demand because, history at the most granular level does not have an accurate trend or seasonal pattern to generate a forecast whereas aggregating the history will provide Demand with a clear trend and seasonal pattern to generate a forecast
Hierarchies exist in demand because, history at the most granular level does not have an accurate trend or seasonal pattern to generate a forecast whereas aggregating the history will provide Demand with a clear trend and seasonal pattern to generate a forecast
The Model component of the DFU represents the type of history stream and algorithm used to generate the forecast.
History Stream identifies the type of history, such as point-of-sale or shipment or invoiced orders.
It is important to understand this cycle before moving forward, since this and other lessons will build upon it.
History is required to generate a statistical forecast with Demand. The history accumulated will help define the forecast created.
Demand stores and tracks forecast accuracy across multiple lags at different hierarchy levels.
It understands the individual forecast components that contribute to the accuracy figure, allowing planners to measure the accuracy of both the statistical forecast and other adjustments.
Market Intelligence What demand planners know manually added additional information
Market Intelligence What demand planners know manually added additional information
Planners need to focus their time on exceptions having the greatest impact on overall performance.
Planners need to focus their time on exceptions having the greatest impact on overall performance.
Priorities can be set by exception types and key product groupings. Filters can be used to alert demand planners to critical exceptions or changes which need to be prioritized and addressed first.
Priorities can be set by exception types and key product groupings. Filters can be used to alert demand planners to critical exceptions or changes which need to be prioritized and addressed first.
Demand provides many tools to manage your exceptions and errors. Errors can occur when a particular process is run.
Some errors can stop the processing of the particular DFU, while others are warnings or informational.
Statistical Analysis is What the data says and is derived by Demand algorithms
Statistical Analysis is What the data says and is derived by Demand algorithms
Classification should run after the history has been cleansed and before the forecasts have been created.
Statistical algorithms, also referred to as statistical models, are used to predict future demand based on the historical data and how the history behaves.
The "planner-driven" phases within Demand Planning Cycle are Evaluate and Fine-tune the Model, Manage Additional Information, Manage Exceptions, and Evaluate Forecast Performance
The "planner-driven" phases within Demand Planning Cycle are Evaluate and Fine-tune the Model, Manage Additional Information, Manage Exceptions, and Evaluate Forecast Performance
Types of forecast algorithm used for example Fourier, MLR, Lewandowski, Holt-Winters, Croston, etc.
The Location component represents a geographic region to which the forecast applies. Examples: Southern sales region Chicago area Europe Distribution center (DC)
A forecast draft table can be used as a work-in-progress table. The forecast table can be used to define the "final" forecast. If you are using the forecast draft table, the Publish Forecast process is required.
The Publish Forecast process would be run once the forecast has been reviewed and is considered complete and accurate, based on the information available at the present time.
The hierarchical multilevel forecasting process organizes your DFUs into a manageable structure. DFU can be any connection across these three dimensions. Each defined DFU level can generate its own statistical forecast based upon aggregated sales history.
The aggregation model aggregates history through the nodes in a hierarchy. It generates independent forecast at each level in the hierarchy. The hierarchies can be in other dimensions, such as Channel and Location.
The goal is not to simply hope that you have the right inventory in the right amount at the right time. You need to be as accurate as possible with the proper use of time, inventory space, and money. This requires forecasting the demand.
The best indicator we have on what the future demand will be is based on what has happened in the past. Historical demand data and patterns are used to predict the demand of future sales, shipments, orders, or other types of data that best suits your company's needs.
The demand planners are responsible for to Evaluate Forecast Performance, Evaluate and Fine-tune the Statistical Forecast, Apply Market Intelligence of the Market to the History and/or Forecast, and Manage Exceptions
The goal is to use the right blend of statistics and market intelligence to create the most accurate forecast possible.
This is where additional information is added to cleanse history and/or adjust forecast projections. This data, called Market Intelligence, can be anything that the statistical model cannot know or recognize regarding customer buying behavior.
The historical data is cleansed to remove one-time events to prevent predicting the event occurrence again in the future. Examples of historical events: Weather, promotions, product recalls, and so on.
A DFU is the entity for which a forecast is created. It represents an item selling in a market. The key components of a DFU are Demand Unit, Demand Group, Location, and Model.
The key components of a DFU are Demand Unit, Demand Group, Location, and Model.
Demand forecasting starts with certain assumptions based on historical data, experience, knowledge, and judgment.
These estimates are projected into the coming weeks, months, or years using one or more mathematical techniques.
Exceptions can also indicate when the model finds data out of a given tolerance. Exception Graphs and Flexible Editor (FE) pages within Demand are used to view these exceptions.
These exception alerts allow a planner to prioritize their workload and focus on the items that need attention and resolution as necessary.
The statistical forecasting process is used to generate or update forecasts for the statistical models. This mathematical calculation is created using a variety of software processes that handle data-intensive tasks.
These processes work with the data stored in tables in the database that provide the criteria for the calculations.
Transfer Forecast can also be used to send the forecast to other system such as Space Management, Workforce Management, or Enterprise Planning.
This closes one forecast period, so we can begin a new one. Where we are in the planning cycle is identified by Demand Post Date.
Monitoring the process and providing feedback to demand planners, stakeholders, and other members of the organization who have input to the forecast can ensure the best possible forecast.
This feedback helps minimize bias, ensure the maximum amount of actual demand is explained by the forecast to minimize noise, and enable continuous improvement of the business demand planning processes.
Depending on the algorithm, there are parameters that can be modified to fine-tune the model.
This, of course, depends on the forecast being generated. These parameters provide the criteria for the statistical forecasting process.
To create your forecast, Demand uses mathematical calculations to predict future demand based on historical data.
To create your forecast, Demand uses mathematical calculations to predict future demand based on historical data.
Supports different accuracy targets at hierarchy levels
Tracks performance by lag
Demand does not know all information about your business the way you demand planner know.
You can make specific adjustments that contribute to a more accurate forecast based on your knowledge about your business, suppliers, marketing, and other sources.