BSCM

Basic of Supply Chain Management

Session 2: Demand Management

Purpose

Supply chain management reacts to customer demand. Through marketing and customer relationship management (CRM), supply chain management also influences and shapes demand. Through demand planning, supply chain management recognizes demand through forecasting and customer order management.

供应链管理对客户需求做出反应。供应链管理通过营销和客户关系管理(CRM)来影响和塑造需求。通过需求规划，供应链管理通过预测和客户订单管理来识别需求。

This session reviews the importance of these business processes and addresses the following topics related to demand planning:

这节课回顾了这些业务流程的重要性，并讨论了以下与需求规划相关的主题:

• Characteristics of demand -- including types, sources, and patterns
  需求的特征——包括类型、来源和模式。

• Forecasting -- including purposes and uses, principles, data collection, techniques, and tracking and measurement
  预测——包括目的和用途、原则、数据收集、技术以及跟踪和度量。

Learning Objectives 学习目标

• Demand Management Processes 需求管理流程

  • Describe the significance of marketing management and CRM.
    描述营销管理和CRM的意义。

  • Explain the role and objectives of demand planning, such as forecasting and customer order management.
    解释需求计划的角色和目标，例如预测和客户订单管理。

• Characteristics of Demand 需求的特征

  • Differentiate independent from dependent demand.
    区分独立的需求和依赖的需求。

  • Identity at least five sources of independent demand.
    识别至少有五个独立需求来源。

  • Recognize at least four demand patterns.
    识别至少四种需求模式。

• Basic Forecasting Concepts 预测的基本概念

  • Describe three planning levels that are supported by demand forecasts.
    描述需求预测支持的三个计划等级。

  • Explain four major principles of forecasting and three principles of data collection and preparation.
    说明预测的四项主要原则和数据收集和准备的三项原则。

  • Differentiate quantitative from qualitative forecasting techniques.
    区分定量和定性预测技术。

• Estimate Demand 估计需求

  • Calculate and explain the logic of and exponential smoothing forecast.
    计算和解释指数平滑预测的逻辑。

  • Explain the logic behind the calculation of a seasonal forecast.
    解释季节预测背后的逻辑。

  • Calculate and explain the use of the mean absolute deviation (MAD) to measure and evaluate forecast error.
    计算和解释平均绝对偏差(MAD)用于测量和评估预测误差。

Demand Management Processes 需求管理流程

Customer demand drives the supply chain. Customers have an inherent demand for products and services. Producers of goods and services react to that demand. Through a process called demand management, producers also influence and shape demand through marketing and sales activities; they recognize demand through forecasting and customer orders. In this session, we address three important business processes that relate to demand management:

客户需求驱动供应链。客户对产品和服务有内在的需求。商品和服务的生产者对这种需求作出反应。通过一个叫做需求管理的过程，生产者也通过营销和销售活动影响和塑造需求;他们通过预测和客户订单来确认需求。在本节中，我们将讨论与需求管理相关的三个重要业务流程:

• marketing management 营销管理
• CRM
• demand planning, such as forecasting and customer orders
  需求计划，如预测和客户订单

In the end, proper demand management is strategically important because it facilitates the planning and use of resources throughout the supply chain.
最后，适当的需求管理具有战略重要性，因为它促进了整个供应链资源的规划和使用。

What other factors might influence demand?
还有哪些因素可能影响需求?

Marketing Management

Marketing Mix 营销组合

The marketing management process relies on a set of tools called the marketing mix, or the four Ps, which are summarized below. The objective is to leverage each P in order to
(1) create products and services with order winning and qualifying characteristics consistent with strategy,
(2) price them,
(3) promote them, and
(4) place them in the right distribution channels.

营销管理过程依赖于一组叫做“营销组合”的工具，或称为“四个P”，总结如下。目标是利用每个P
(1)Product:创造符合策略的产品和服务，具有赢得订单和合格的特征;
(2)Price:给它们定价,
(3)Promotion:促销它们,
(4)Place:把它们放在合适的分销渠道。

For the purposes of this course, this brief marketing mix description provides an awareness of the decisions marketing faces in influencing and shaping demand.

对于本课程的目的而言，这个简短的营销组合描述提供了市场营销在影响和塑造需求方面所面临的决策。

Order Winners and Qualifiers 订单赢家和资格要素

The characteristics of products and services are a key element of the marketing mix. An important role of marketing management is to develop and implement marketing plans for products and services with order winning or qualifying characteristics.

产品和服务的特性是营销组合的关键因素。市场营销管理的一个重要作用是制定和实施具有赢得订单或符合条件的产品和服务的营销计划。

• Order qualifying characteristics must be exhibited by a firm's products and services in order for the firm to be a viable competitor in the marketplace.
  为了使公司在市场上成为一个有竞争力的竞争者，公司的产品和服务必须展示出符合要求的特点。

• Order winning characteristics cause customers to choose a product over its competitors' products.
  赢得订单的特点使客户选择产品而不是竞争对手的产品。

• Examples of order winning and order qualifying characteristics are quality, speed, dependability, flexibility, and cost. These are sometimes referred to as performance objectives.
  赢得订单和订单资格要素的例子有质量、速度、可靠性、灵活性和成本。这些有时被称为绩效目标。

A useful comparison of the difference between and order winner and an order qualifier is provided in a CPIM reference source, Operations Strategy, Slack et al., 2003. The authors cite fast-food retailer Burger King (BK) as an example. BK views the following as order qualifiers that do not significantly differentiate their products and services from those of competitors:

CPIM参考资料来源《Operations Strategy》（Slack等人2003年）提供了一个关于订单赢家和订单资格要素之间差异的有用比较。作者以快餐零售商汉堡王(BK)为例。BK认为以下是订单资格要素，它们的产品和服务与竞争对手的产品和服务没有显著区别:

• Speed of service -- Key competitors are about as fast in both drive-through service and counter service.
  服务速度——在免下车服务和柜台服务方面，主要竞争对手的速度都差不多。

• Dependability -- Reliability of BK's full range of product and service availability and its ordering experience consistency also are on par with the industry.
  可靠性——BK的所有产品和服务的可靠性及其订购体验一致性也与行业相当。

• Flexibility -- BK is similar to competitors in its ability to meet changing sales volume and mix, or product option, requirements during the day.
  灵活性——BK与竞争对手类似，能够在白天满足不断变化的销量和产品组合或产品选择需求。

• Cost -- BK's prices are not the cheapest but are not much different from their competitors' prices.
  成本——BK的价格不是最便宜的，但与竞争对手的价格相差不大。

Voice of the Customer 客户的声音

Determining order qualifying and winning characteristics relies to a great extent on the design of products to meet customer expectations: the voice of the customer. Later in this course, we will review a formal method developed for use in quality management and lean manufacturing for determining the voice of the customer during product and process design. This method, quality function deployment, will be addressed in Session 9.

在很大程度上，决定订单排定和获胜的特性取决于产品的设计，以满足客户的期望:客户的声音。在本课程的后面，我们将回顾在质量管理和精益生产中开发的一种正式方法，在产品和过程设计中确定客户的声音。这个方法，质量功能部署，将在第9课中讨论。

Customer Relationship Management 客户关系管理

CRM -- The collection and analysis of information designed for sales and marketing decision support to understand and support existing and potential customer needs. It includes account management, catalog and order entry, payment processing, credits and adjustments, and other function.

-- APICS Dictionary

CRM——为销售和市场决策支持的信息收集与分析而设计，以理解和支持现有的和潜在的客户需求。它包括账户管理、目录和订单录入、支付处理、积分和调整等功能。

The quality of a company's interaction with its customers can become a strategic asset and differentiator. It provides an advantage in attracting new customers and creating loyalty among existing ones. Important interactions with customers occur in a number of areas, as shown in the following examples.

公司与客户互动的质量可以成为战略资产和差异化优势。它在吸引新客户和在现有客户之间建立忠诚度方面具有优势。与客户的重要交互发生在许多领域，如下所示。

• Design assistance: helping customers in the design of new products or improvement of existing ones
  设计协助:帮助客户设计新产品或改进现有产品。

• Determination of customer needs: assessing the customer's business and creating or expanding product offerings to help the customer achieve better business results
  确定客户需求:评估客户的业务，创建或扩展产品，以帮助客户取得更好的业务成果

• Information and communications: collecting and analyzing customer data to support targeted marketing campaigns and practices, such as Amazon.com Inc.'s suggestions of books to buy based on a customer's past purchases

  收集和分析客户数据，以支持有针对性的营销活动和实践，如亚马逊公司(Amazon.com Inc.)根据客户过去的购买情况提出购买书籍的建议。

Customer order management is a major CRM activity and also plays a major role from a supply chain operations efficiency and customer service perspective in enabling the following:

客户订单管理是一项重要的CRM活动，它也从供应链运营效率和客户服务的角度发挥了重要的作用。

• Fast and accurate order entry and tracking
  快速准确的订单录入和跟踪

• Ability to provide and meet promised delivery dates and quantities
  提供并满足承诺的交货日期和数量的能力

• Handling customer inquiries and service complaints, returns, and repair requests
  处理客户查询及服务投诉、退货及维修要求

• Accurate and timely shipping documentation, invoicing, and recoding of sales history
  准确和及时的发货单，发票和销售历史记录

Along with marketing management, CRM plays an important role in demand management.

与市场营销管理一起，CRM在需求管理中起着重要作用。

Demand Planning 需求计划

Demand planning is the recognition of demand, which comes in two forms:
需求计划是对需求的确认，它有两种形式:

• demand forecasts 需求预测
• management of actual customer orders from internal and external customers
  管理来自内部和外部客户的实际客户订单

Forecasts and customer orders drive operations, and manufacturers have to get these right in order to understand what, how much, and when to produce products. Different types of orders are discussed under sources of demand in the next section on characteristics of demand.

预测和客户订单驱动操作，制造商必须获得这些权利，以便了解生产什么、生产多少、何时生产产品。下一节将讨论不同类型的订单在需求来源下的需求特征。

It is important to understand the characteristics of demand in order to understand the demand forecasting process. These two topics will be addressed in the next two sections.
要理解需求预测过程，了解需求特征是很重要的。这两个主题将在下两节中讨论。

Characteristics of Demand 需求的特征

In this section we will discuss the following characteristics of demand that are key influences on demand forecasting:
在本节中，我们将讨论需求的以下特征，它们是影响需求预测的关键因素:

• independent versus dependent demand 独立需求与依赖需求
• sources of demand 需求来源
• demand patterns 需求模式

Independent Versus Dependent Demand 独立需求与依赖需求

There are two types of demand: independent and dependent demand. Independent demand is the basis for demand planning in terms of forecasts and customer orders.

需求有两种类型:独立需求和依赖需求。独立需求是预测和客户订单需求计划的基础。

• Independent demand is demand for a product that is not related to the demand for any other product. An automobile manufacturer forecasts demand for motor vehicles; demand for the motor vehicles is independent.

独立的需求是对与任何其他产品的需求无关的产品的需求。汽车制造商预测汽车需求;对汽车的需求是独立的。

• Dependent demand is demand for a product that is related to the demand for another product. Dependent demand is calculated.In the automobile manufacturer example, it is not necessary to forecast the demand for tires, steering wheels, and other components. Demand for these components is dependent on the number of vehicles forecasted and will be derived during a planning process called material requirements planning (MRP), which is further discussed in Session 4. Tires and other components bought independently of automobiles, however, would be considered independent demand.

依赖需求是指对另一种产品的需求。计算相关需求。在汽车制造商的例子中，不需要预测对轮胎、方向盘和其他部件的需求。对这些部件的需求取决于预测的车辆数量，并将在称为材料需求计划(MRP)的过程中得到，这将在第4课中进一步讨论。不过，独立于汽车购买的轮胎和其他零部件将被视为独立需求。

This is an important concept in demand planning. Forecasting deals with independent demand.

这是需求规划中的一个重要概念。预测涉及独立需求。

Sources of Demand 需求来源

There are five common sources of independent demand used in forecasting and managed during order processing.

在订单处理过程中，有五种常见的独立需求来源用于预测和管理。

• Forecast -- Predictions of future demand based on quantitative or qualitative methods, or a combination of the two
  预测——基于定量或定性方法或两者结合的预测未来需求

• Customer orders -- These orders are from external customers for particular products or numbers of products, often referred to as actual demand as opposed to a forecasted demand, including service parts and spare parts. (Note that the same part planned by MRP as a component of an end item would represent dependent demand.)
  客户订单——这些订单来自外部客户，针对特定产品或产品数量，通常指的是实际需求，而不是预期的需求，包括服务部件和备件。(请注意，MRP计划的与成品的组成部分相同的部分将代表依赖需求。)

• Replenishment orders form distribution centers -- Based on customer orders placed on distribution centers and on distribution center forecasts
  配送中心补货订单——基于配送中心的客户订单和配送中心预测

• Interplant transfers -- Orders for manufactured parts or components from other divisions in the company
  工厂间转移——公司其他部门制造的零部件的订单

• Other sources of demand -- Examples include orders for products needed for marketing and product demonstrations
  其他需求来源——例如营销和产品演示所需的产品订单

Note that these are characterized as sources of independent demand. Independent demand is the basis for forecasting.

注意，这些是独立需求的来源。独立需求是预测的基础。

Demand Patterns and Stability 需求模式和稳定性

The plotting of demand data in a time series reveals that demand can take on different patterns. Forecasters analyze and attempt to understand these patterns in terms of shape and stability.

在时间序列中绘制需求数据表明需求可以呈现不同的模式。预测者分析并试图从形状和稳定性的角度理解这些模式。

Patterns

There are four basic demand patterns, or components:
有四种基本的需求模式或组成部分:

• Trend: Visual 2-15 shows three examples of trends -- increasing, decreasing, and level. In the increasing and decreasing trends, demand changes at a steady rate from period to period. This particular example shows a linear rate of increase and decrease, but the year-to-year rate of change could be exponential. The trend pattern also can be level.

趋势:图2-15展示了趋势的三个例子——增加、减少和水平。在增加和减少的趋势中，需求在一个时期一个时期一个稳定的速度变化。这个特别的例子显示了线性的增长和下降速度，但是年与年的变化速度可能是指数级的。趋势模式也可以是水平的。

demand patterns trend

• Seasonal: Visual 2-16 shows the regular fluctuation of quarterly demand over the course of three years -- the result of seasonal influences such as weather, school calendars, and social customs. Seasonal demand patterns also can occur in shorter time intervals, such as monthly, weekly, daily, or hourly -- as in sales of consumer products -- and can show signs of trend.

季节性:图2-16显示了三年内季度需求的规律波动——季节影响的结果，如天气、校历和社会习俗。季节性需求模式也可以在更短的时间间隔内出现，如月度、每周、每日或每小时——如消费品销售——并可以显示趋势迹象。

demand-patterns-seasonal

• Random: Visual 2-17 shows demand fluctuation due to random occurrences such as inclement weather, special events, and the vagaries of human behavior. Over time, random demand will vary near the average and the variations will cancel each other out. If variations are small, the random variation is considered stable. When they are large, it is considered unstable. Random variation can exist within a trend or seasonal demand pattern but does not alter the general pattern.

随机:图2-17显示需求波动，如恶劣天气、特殊事件和人类行为的变幻莫测。随着时间的推移，随机需求会在接近平均值的地方发生变化，这些变化会相互抵消。如果变化很小，则认为随机变化是稳定的。当它们很大时，就被认为是不稳定的。随机变化可以存在于趋势或季节需求模式中，但不会改变一般模式。

demand-patterns-random

• Cyclical: Cyclical demand is characterized by wave-like fluctuations that take place over long time spans, such as several years, and are tried to external influences such as the business cycle. Forecasting of cycles is the domain of economic forecasters, though businesses recognize their impact on demand and sales.

周期性:周期性需求的特征是波动，这种波动会在很长一段时间内发生，比如几年，并且会受到外部影响，比如商业周期。虽然企业认识到周期对需求和销售的影响，但对周期的预测是经济预测者的领域。

Stable Versus Dynamic Demand 稳定需求对比动态需求

stable-versus-dynamic-demand

Demand patterns may change over time.
需求模式可能随时间而改变。

• Patterns that retain their general shape over time are called stable and are easier to forecast.
  随时间保持其一般模型的模式称为稳定模式，更容易预测。

• Dynamic patterns are less predictable, though technically in the long run they may have a random pattern. Dynamic changes make the prediction and forecasting of demand difficult and prone to error.
  动态模式的可预测性较低，但从技术上来说，它们可能具有随机模式。动态变化使需求预测和预测变得困难，容易出错。

Visual 2-18 presents a simple model that highlights the implications of the difference between stable and dynamic demand.
图2-18提供了一个简单的模型，强调了稳定需求和动态需求之间的差异。

• The model assumes that average demand is the same over time for both stable and dynamic demand.
  该模型假设稳定和动态需求的平均需求随时间而变化。

• Though average demand is the same in the stable and dynamic demand scenarios shown, the wide and irregular fluctuations of dynamic demand make it difficult to develop and accurate forecast.
  虽然在所示的稳定和动态需求情景中，平均需求是相同的，但动态需求的大而不规则的波动使其难以发展和准确预测。

• Variations in the stable demand scenario are relatively predictable and remain consistently within predetermined or expected upper and lower limits, allowing for random variations. It is easier to develop a forecast based on average demand and develop an inventory policy -- for example, safety stock - for this scenario.
  稳定需求场景中的变化是相对可预测的，并始终保持在预定的或预期的上限和下限范围内，允许随机变化。在这种情况下，根据平均需求制定预测和制定库存政策(例如安全库存)更容易。

Forecasting 预测

In this section, we will address the following important concepts relating to the purposes and principles of forecasting:
在本节中，我们将讨论与预测的目的和原则有关的下列重要概念:

• purposes and uses of the forecast 预测的目的和用途
• principles of forecasting 预测原理
• principle of data collection and preparation 数据收集和准备原则

Purposes and Uses of the Forecast 预测的目的和用途

In Session 1, we learned about the multilevel planning methodology used to plan priorities and capacity: manufacturing planning and control (MPC). Demand forecasts support planning at the following three levels:
在第1课中，我们学习了用于计划优先事项和能力的多级计划方法:制造计划和控制(MPC)。需求预测支持以下三个层面的规划:

• business planning 商业计划
• sales and operations planning (S&OP) 销售及运营计划
• master scheduling 主调度

Forecasts become more detailed as the planning moves from the business planning to the master scheduling level, whose output is the master production schedule (MPS).

随着计划从业务计划转移到主调度级别，预测变得更加详细，主调度级别的输出是主生产计划(MPS)。

Principles of Forecasting 预测的原则

In Introduction to Materials Management, the authors point out that there are four major, simple, and common-sense principles of forecasting. They include the following:
在《物料管理导论》中，作者指出预测有四种主要的、简单的、常识性的原则。他们包括以下:

Principle 2 is especially an important one. It is counterintuitive to think that forecasts can predict the exact demand for and item during a specific period in the future, but their forecasts can be highly useful if they also include an estimate of the probable range of error.

原则2尤其重要。认为预测可以预测未来某一特定时期内的确切需求和项目，这是违反直觉的，但如果预测还包括对可能的误差范围的估计，那么预测将非常有用。

Principle 4 holds an important lesson for supply chain management. Because forecasts are much more accurate in the short term, companies that shorten their production lead times can react to short-term rather than long-term demand forecast data. As you will learn in this course, lead-time reduction is a major operational concern for manufacturing enterprises.

原则4是供应链管理的重要一课。由于短期预测要准确得多，缩短生产交货期的公司可以对短期而非长期需求预测数据做出反应。正如您将在本课程中了解到的，缩短交货期是制造企业的一个主要操作问题。

Data Collection and Preparation Principles 数据收集和准备原则

Most forecasts are based on historical data, and the collection and preparation of the data for use in forecasting is of utmost importance. It requires judgment and sound analytical skills.

大多数预测是基于历史数据，而收集和准备用于预测的数据是极其重要的。它需要判断力和良好的分析能力。

There are three major principles of data collection and preparation:

收集和准备数据有三个主要原则:

Principle 1: To match the items planned at the S&OP level, the forecast should be at the product family or group level. At the MPS level, the forecast needs to be at the end item or item option level for all items within product families -- the actual end items.

原则1:为了与S&OP级计划的项目相匹配，预测应该在产品系列或集团级。在MPS级别，预测需要位于产品系列中的所有项目的最终项目(实际的最终项目)或项目选项级别。

Principle 3 : Customer A orders two same-size large lots a year and customer B orders same-size small lots once a month. It would be reasonable to forecast customer B's demand on a monthly average basis, but to do the same for customer A actually will mask the umpires of customer A's demand -- resulting in a very inaccurate aggregate monthly forecast.

原则3:客户A每年订购2个相同大小的大批次，客户B每月订购大小相同的小批次。按月平均预测客户B的需求是合理的，但对客户A做同样的预测实际上会掩盖客户A的需求，从而导致非常不准确的月度总体预测。

data-collectoin-example

Forecasting Techniques 预测技术

Forecasting techniques, also referred to as methods, are in two categories. The second category, quantitative techniques, consists of two sub-categories:

预测技术也称为方法，可分为两类。第二类，定量技术，包括两个子类:

• qualitative 定性
• quantitative 定量
  • extrinsic 外在
  • intrinsic 内在

Qualitative Techniques 定性技术

Qualitative techniques are based on intuition and informed opinion; they therefore tend to be subjective.

定性技术基于直觉和知情的意见;因此，它们往往是主观的。

They are a necessary supplement to quantitative techniques for business planning. For example, qualitative techniques assist in forecasting general business trends and the demand for new products through market research, historical analogy, test marketing, and the Delphi method of having a panel of experts.

它们是商业计划定量技术的必要补充。例如，定性技术通过市场研究、历史类比、测试营销和德尔菲法专家小组的方法来帮助预测一般的商业趋势和对新产品的需求。

These techniques are used for medium- to long-range planning, where business conditions, opportunities, and trends are likely to change and where quantitative techniques are not fully adequate.

这些技术用于中长期规划，业务条件、机会和趋势可能会改变，而定量技术还不够充分。

Quantitative Techniques: Extrinsic 定量技术:外在

Extrinsic techniques are based on the idea of correlation and cause and effect. They rely on external indicators to make projections or forecasts of demand. For example, housing starts influence the sale of building materials. Housing-starts data published by the home building industry therefore as used as a leading indicator of demand for a wide range of other products associated with home building such as appliances, carpeting, and other home furnishings.

外在技术是建立在相关、因果观念的基础上的。它们依靠外部指标来预测或预测需求。例如，房屋开工会影响建材的销售。房屋开工数据由房屋建筑行业发布，因此被用作与房屋相关的其他产品需求的领先指标，如家电、地毯和其他家具。

There are two commonly used types of leading indicators used in making extrinsic forecasts:

在进行外部预测时，常用的先行指标有两种:

• Economic : some examples include housing starts, construction contract awards, contract wards by the military, and orders for large military and commercial aircraft.

经济方面:一些例子包括房屋开工、建筑合同授予、军方合同病房以及大型军用和商用飞机订单。

• Demographic: Examples include changes in the population such as the birth rate, age distribution, income and education levels, and ethnicity.

人口统计:例如出生率、年龄分布、收入和教育水平以及种族等人口的变化。

Extrinsic forecasts tend to look beyond the short term and are used mainly to forecast total company demand and demand for smilies of products rather than individual end items.

外部预测往往着眼于短期之外，主要用于预测公司对特定产品的总需求和需求，而不是单个最终项目。

Quantitative Techniques: Intrinsic 定量技术:内在

Intrinsic forecasting techniques use time-sequenced historical data, or time series data, for an item as the source for projecting further demand. This technique is based on number of assumptions.

内在预测技术使用时间序列的历史数据，或时间序列数据，为项目作为预测未来需求的来源。这种技术基于许多假设。

• What happened in the past helps you understand what will happen in the future.
  过去发生的事情可以帮助你了解将来会发生什么。

• Whether or not time series data are available in company information systems -- the usually are.
  公司信息系统中是否有时间序列数据——通常是。

• What patterns of demand are revealed in time series data; for example, trend, seasonal, random, and cyclical should be major factors in forecasting future demand.
  时间序列数据显示了哪些需求模式;例如，趋势、季节、随机和周期性应该是预测未来需求的主要因素。

Examples of two basic intrinsic forecasting techniques are discussed below: moving averages and exponential smoothing.

下面讨论两种基本内在预测技术的例子:移动平均法和指数平滑法。

Moving-Average Forecast 移动平均预测

Introduction

Moving-average forecasts are based on the following principles:
移动平均预测基于以下原则:

• Best used when demand is stable, there is little trend or seasonality, and demand variations are random.
  当需求稳定时最好使用，因为几乎没有趋势或季节性，或需求变化是随机的。

• When past demand fluctuates because of random variation, as it often does, it is better to forecast based on average demand than second-guess the effect of random fluctuation.
  当过去的需求因随机变化而波动时，通常情况下，基于平均需求进行预测比事后猜测随机波动的影响要好。

Sample Calculations 样本计算

12-month-demand-history

Let's assume that we are at the end of December and need to forecast demand for the next month, January. Demand of the last 12 months has shown random variation. Average demand was 79, and the sum of positive and negative variances from the average was zero. Forecasters face a few choices, including

假设我们在12月底，需要预测下个月的需求，1月份。过去12个月的需求呈现随机变化。平均需求为79，正方差和负方差之和为0。预测者面临着一些选择，包括

• using the demand for the previous January: 92
  使用上一个1月份的需求:92

• using the demand for the previous month, December: 84
  使用上月需求计算，十二月:84

• using the average demand for preceding months -- for example, the January to December average is 79.
  以前几个月的平均需求为例，1月至12月的平均需求为79。

The average demand calculation for previous periods applied to the forecast period is in part the basic logic of moving-average forecasting. In this case, we calculated the average demand for January through December to arrive at 79.

应用于预测期的以前各期的平均需求计算部分是移动平均预测的基本逻辑。在这种情况下，我们计算了1到12月的平均需求到79。

These are three different results for a demand pattern that shows random variation and has no obvious upward or downward trend. Being random, however, demand just happens to be higher in both January and November. The advantage of using an average is that it neutralizes that impact of random variation on the forecast and results in a safer forecast.

是三种需求模式的不同结果，这些需求模式呈现随机变化，没有明显的上升或下降趋势。然而，由于是随机的，1月和11月的需求恰好都比较高。使用平均值的优点是它可以抵消随机变化对预测的影响，从而产生更安全的预测。

moving-average-forecast-sample

Moving-Average Forecast Logic 移动平均预测逻辑

The moving-average forecast uses the average of demand for the past three to six months, for example, to forecast demand for the next month. We'll use three months in this example.

移动平均预测用过去三到六个月的平均需求来预测下个月的需求。在这个例子中，我们将使用3个月。

At the end of month 3, the average demand for months 1 through 3 would be the forecast for month 4.

在第3个月末，第1个月到第3个月的平均需求将是第4个月的预测。

At the end of month 6, the average demand for months 4 through 6 would be the forecast for month 7.

在第6个月末，第4到第6个月的平均需求将是第7个月的预测。

At the end of month 7, the average demand for months 5 through 7 would become the forecast for month 8, and so forth.

在第7个月末，第5个月到第7个月的平均需求将成为第8个月的预测，以此类推。

Moving-average forecasting issues 移动平均预测问题

The data shown below are for two moving-average forecast scenarios based on the same demand data: a three-month moving-average forecast and a six-month moving-average forecast.

下面显示的数据是基于相同需求数据的两种移动平均预测场景:3个月移动平均预测和6个月移动平均预测。

Note that an upward demand trend begins in month 6. The forecasts in both scenarios are slow to react, although the three-month moving-average forecast reacts a lot quicker in months 7, 8, and 9 to the sudden increases in demand.

请注意，需求上升趋势从第6个月开始。这两种情况下的预测反应迟缓，尽管三个月移动平均预测在7月、8月和9月对需求突然增加的反应快得多。

moving-average-forecasts-example

The lessons learned are as follows:
经验教训如下：

• The moving-average forecast will lag the development of a rising or falling trend.
  移动平均预测将滞后于上升或下降趋势的发展。

• The further back the moving-average forecast reaches for data to include in average demand, the greater the lag. For example, the six-month moving-average forecast for month 6 included three more months of earlier demand than the three-month moving-average forecast.
  移动平均预测值越接近平均需求，延迟越大。例如，第6个月的6个月移动平均预测包括比3个月移动平均预测提前3个月的需求。

• On the other hand, the three-month moving-average forecast may have overreacted if the demand surge is an aberration and demand suddenly falls to the level before period 6.
  另一方面，如果需求激增是一种反常现象，而需求突然降至第6期前的水平，3个月移动平均预测可能反应过度。

• The moving-average forecast works best when demand is stable and continuous. In this type of situation, it will filter out random variation.
  当需求稳定且连续时，移动平均预测效果最佳。在这种情况下，它会过滤掉随机变化。

Exponential Smoothing 指数平滑法

Logic 逻辑

The logic of the exponential smoothing forecast is straightforward and as follows:
指数平滑预测的逻辑很简单，如下:

• Take the old forecast and the demand for the latest, most current period.
  以旧的预测和最新的需求为例。

• Assign a weighting factor or smoothing constant (α, alpha) to the latest period's demand to give it an appropriate weighting relative to the forecast.
  分配一个权重因子或平滑常数(α,阿尔法)给最新阶段的需求，给它一个适当的相对于预期的权重。

• Calculate the weighted average of the old forecast and the latest demand.
  计算旧预测的加权平均值和最新的需求。

Calculation Summary 计算概要

The exponential smoothing forecast technique is similar to the moving-average forecast, but requires less data management and is easier to calculate.

指数平滑预测技术类似于移动平均预测，但需要较少的数据管理，更容易计算。

• At the end of a current period (for example, month 1), start the exponential smoothing process with a moving-average forecast for the next period -- month 2 -- as described in a previous section. You have now established an old forecast.
  在当前周期结束时(例如，第1个月)，使用前一节中描述的下一周期(第2个月)的移动平均预测启动指数平滑过程。你现在建立了一个旧的预测。

• At the end of month 2, use the actual demand and the old forecast for month 2 to calculate an exponential smoothing forecasts for month 3.
  在第2个月底，利用第2个月的实际需求和旧的预测，计算第3个月的指数平滑预测。

• To the actual demand for month 2, assign a weighting factor or smoothing constant α, which means that the weighting for the old forecast component is 1 - α.
  第2个月的实际需求,分配一个权重因子或平滑常数α,这意味着旧的预测组件权重为1 - α。

• Calculate the forecast for month 3 based on a weighted average calculation as follows:
  根据加权平均计算，计算第3个月的预测:

  New forecast = (α)(latest demand)+(1-α)(previous forecast)
  新的预测=(α)(最新需求)+(1-α)(以前的预测)

Smoothing Constant 平滑常数

Selection of an appropriate smoothing constant (α) requires analysis of demand patterns and judgment. A low alpha value such as .2 will give much more weight to the old forecast (1 - .2 = .8), which might not be appropriate if there is an upward or downward demand trend. In either case, a higher α would be appropriate.
选择合适的平滑常数(α)需要分析需求模式和判断。像.2这样的低alpha值会使旧的预测(1 -。2 = .8)更有分量，如果有向上或向下的需求趋势，这可能是不合适的。无论哪种情况，较高的α都是合适的。

By using spreadsheets or simulation software, it is possible to run simulations with different alpha values to see which one best fits the historical demand pattern for particular products.

通过使用电子表格或模拟软件，可以使用不同的α来运行模拟，看看哪一个最适合特定产品的历史需求模式。

Seasonal Forecast 季节性预测

For products with a seasonal demand pattern, the forecasting process is as follows. Note that these steps assume seasonal periods are quarters of the year.

对于具有季节性需求模式的产品，预测过程如下。注意，这些步骤假设季节周期是一年的四分之一。

• Step 1: Calculate a seasonal index of demand for each period, which might be for each quarter of a year, to account for seasonality in historical demand.
  第1步:计算每个时期的季节性需求指数，可能是每年的每个季度，以反映历史需求的季节性。

• Step 2: Develop deseasonalized demand by developing total forecast for a year and divide it by the number of periods, or quarters. This takes the seasonality out. The result is average demand across all seasons.
  第2步:通过制定一年的总预测除以周期数或季度数，来制定消除季节变动后的需求。这就去掉了季节性。结果是所有季节的平均需求量。

• Step 3: Develop a seasonal forecast for each quarter by multiplying the deseasonalized demand by each quarter's seasonal index.
  第3步：通过消除季节变动后的的需求，的需求乘以每个季度的季节性指数，来制定每个季度的季节性预测。

Seasonal Index/Seasonality (Step 1)

季节指数/季节性(第1步)

The first step in the forecast is to determine seasonality by calculating the seasonal index for each period, or quarter, using the equation shown. The indices indicate the degree of variation from the average through each quarter within an annual cycle.

预测的第一步是通过计算每个时期或季度的季节指数来确定季节性。这些指数表明在一个年度周期内，从平均水平到每个季度的变化程度。

Seasonal index = (period average demand) / (average demand for all periods)

季节指数=(期平均需求)/(各期平均需求)

• Period average demand can be calculated using the data for years 1 to 3 in the demand table. For example, for quarter 1 the period average demand equals 128.
  周期平均需求可使用需求表中1至3年的数据计算。例如，第一季度平均需求为128。

• The average demand for all quarters is equal to the three-year average of total demand (400) divided by the number of quarters (4), which equals 100.
  所有季度的平均需求，等于三年平均总需求(400)除以季节数(4)，即100。

• Based on the equation for the seasonal index, the index for quarter 1 is 1.28. The seasonal index for quarter 2,3, and 4 are calculated in the same way.

  根据季节指数方程，第一季度指数为1.28。季2、季3和季4的季节指数也是用同样的方法计算的。

    Seasonal index (Q1) = 128 / 100 = 1.28
  

demand history

Deseasonalized Forecast (Step 2)

The second step consists of two parts:
第二步包括两部分:

• Make the forecast for the next year.
  预测下一年。

• Deseasonalize the forecast -- distribute it evenly across the four quarters.
  对预测进行调整——在四个季度内平均分配。

Let's assume that, based on product improvements and an increased marketing plan, the company forecasts demand for year 4 will be 420.
让我们假设，基于产品改进和增加的营销计划，公司预测第四年的需求将是420。

How will this forecast be distributed across the four quarters of year 4? The annual forecast, divided by four, results in an average demand of 105 units per quarter. We call this deseasonalizing the demand.
这一预测将如何在第四年的四个季度中分布? 年度预测除以4，结果是每季度平均需求105台。我们称之为需求的不稳定。

Deseasonalized demand (average demand / period) = (annual forecast) / (number of periods) = 420 / 4 = 105 units

稳定需求(平均需求/周期)=(年度预测)/(周期数)= 420 / 4 = 105个单位

Seasonal Forecast (Step 3)

季度预测（第3步）

The third and final step is to convert deseasonalized, or average, demand for the four periods into a seasonal or seasonality-adjusted forecast. The process is simple : For each quarter, multiply the deseasonalized forecast by the seasonal index established in Step 1.

第三步，也是最后一步，是将四个时期的消除季节变动后的或平均的需求，转化为季节性或季节性调整后的预测。过程很简单:对于每一个季度，用第一步建立的季节指数乘以消除季节变动后的的预测。

Expected quarter demand               =     (seasonal index) * (deseasonalized forecast demand) 
预期季度需求                            =     (季节指数)*(消除季节变动后的预测需求)

Expected first quarter demand         =    1.28 * 105    = 134 units
预计第一季度需求

Expected second quarter demand        =   1.02 * 105     = 107 units
预计第二季度需求

Expected third quarter demand         =    .75 * 105     = 79 units  
预计第三季度需求

Expected fourth quarter demand        =    .95 * 105     = 100 units
预计第四季度需求
Total forecast demand                 =                 420 units
总预测需求

Tracking the Forecast 跟踪预测

The purpose of tracking the forecast is to compare it within actual demand to measure and improve its accuracy. In the process, we can learn the following:

跟踪预测的目的是在实际需求范围内进行比较，以衡量和提高预测的准确性。在这个过程中，我们可以学到以下内容:

• why demand differs from the forecast -- has a demand changed or has a trend developed?
  为何需求与预测不同?需求是否有变化或有发展趋势?

• the need to plan around the error, such as adjusting demand history to account for exceptional one-time reasons for error
  需要围绕错误进行计划，例如调整需求历史，以应对异常的一次性错误原因

• how to improve the forecast
  如何改善预测

This section introduces and provides awareness for some of the basic concepts of tracking the forecast. These basic concepts include the following:
本节介绍并提供跟踪预测的一些基本概念。这些基本概念包括:

• forecast error -- both bias and random 预测误差——偏差和随机
• measuring forecast error 测量预测误差

Bias 偏差

bias VS random

Refer to the data in the Visual 2-50. Bias exists when the cumulative variation of actual demand from the cumulative forecast is not zero. Cumulative demand is 90 units higher than forecast.

参考图2-50中的数据。当累积预测的实际需求的累积变化不为零时，存在偏差。累计需求比预期高90个单位。

This means that the average monthly demand was 115 units (690 units / 6 months) instead of 100.

这意味着月平均需求量为115台(690台/ 6个月)，而不是100台。

There are some actions to take and alternatives to consider:

我们需要采取一些行动，并考虑以下备选方案:

• Investigate and take into account the cause of the error, such as sales promotions, large on-time orders, a customer shutdown, machine breakdowns, exceptional weather events, and new customers.
  调查并考虑错误的原因，如促销、大订单、客户关机、机器故障、异常天气事件和新客户。

• If necessary, adjust the demand history.
  如有必要，调整需求历史。

• Change the monthly average forecast.
  更改月度平均预测。

Measuring Forecast Error 测量预测误差

The degree of deviation from average demand presents a challenge to inventory management. How much safety stock, for example, should be carried to deal with random variations in demand and to achieve desired customer service levels? This is why it is important to measure and evaluate forecast error.

偏离平均需求的程度对库存管理提出了挑战。例如，为了处理需求的随机变化并达到预期的客户服务水平，需要携带多少安全库存?这就是衡量和评估预测误差为什么重要的原因。

Mean Absolute Deviation (MAD) 平均绝对偏差

Here is a quick, high-level summary of a simple and proven statistical approach to measuring and evaluating forecast error. Remember, forecast error, or deviation, is the actual period demand minus the forecast (A-F). If actual demand exceeds forecast, the error will have a positive value and vice versa. In calculating MAD, however, the MAD is used to determine the dispersion of the forecast error around average demand. Let's illustrate by calculating the value of the MAD using the data and formula below.

这里有一个快速的，高级的总结，一个简单的和经过验证的统计方法来测量和评估预测误差。记住，预测误差或偏差，是实际周期需求减去预测(A-F)。如果实际需求超过预测，误差将为正值，反之亦然。然而，在计算MAD时，MAD用于确定预测误差在平均需求附近的离散程度。让我们用下面的数据和公式来计算MAD的值。

MAD

normal-distribution

Now let's look at the statistically determined dispersion of forecast error using a normal distribution curve for the data shown above. (See Visual 2-53)
现在，让我们用正态分布曲线来研究统计上确定的预测误差的离散性。(图2-53)

• The center is the central tendency or average, which in this case is the forecast.
  中心是中心趋势或平均值，在这种情况下是预测。

• Forecast error, which are not shown in the visual, are randomly dispersed on both sides of the central tendency.
  预测误差不显示在图上，随机分布在中心趋势的两边。

• Statistically, in a normal distribution, 60 percent of the forecast error will fall within ±1 MAD of the average, 90% will fall within ±2 MAD, and 98% will fall within ±3 MAD.
  正态分布,预测误差的60%将会平均落在±1 MAD,90%落在±2 MAD,98%会在±3 MAD。

• If the actual demand is tightly bunched around the center, then the dispersion is narrow and forecast accuracy is relatively high -- and vice versa for demand that is widely dispersed.
  如果实际需求集中在中心附近，则离散度较窄，预测精度较高，反之则需求分散较广。

• As shown in the Visual 2-53 in the example we just calculated, 60% of the errors, or period demand quantities, were within ±22 units, 90% were within ±44 units, and 98% were within ±66 units of the average.

如图2-53我们刚计算的,60%的错误,或期间需求数量,在±22个单位,90%在±44个单位,98%是在±66单位的平均水平。

Measurements such as MAD are important because they indicate the relative cost of different levels of customer service. For example, by using safety factor tables, planners can convert the MAD of 22 units into safety stock requirements for desired service level percentages.

诸如MAD之类的度量很重要，因为它们指示了不同级别的客户服务的相对成本。例如，通过使用安全系数表，规划者可以将22个单元的MAD转换为所需服务级别百分比的安全库存需求。

Standard Deviation 标准差

MAD and standard deviation are alternate methods to measure forecast error. Standard deviation is covered in more detail in other CPIM modules, especially as it relates to determining safety stock levels.

MAD和标准差是衡量预测误差的替代方法。标准偏差在其他CPIM模块中有更详细的介绍，特别是在确定安全库存水平时。

Uses of Forecast Measurement 预测测量的使用

You should come away from this discussion within an appreciation of how measuring and understanding forecast error can help a company deal with demand uncertainty in the following areas:

你应该从这个讨论中认识到如何测量和理解预测误差可以帮助公司处理以下领域的需求不确定性:

• identifying changes and trends in demand
  识别需求的变化和趋势

• identifying and adjusting for forecast error that results from random events
  识别和调整随机事件导致的预测误差

• adjusting the period forecast as close to the true forecast average demand as possible to eliminate or minimize bias in the forecast
  根据实际预测的平均需求调整期间预测，以消除或减少预测的偏差

• making decisions on safety stock and service levels based on the degree of random variation, or forecast error
  根据随机变化的程度或预测误差来决定安全库存和服务水平

Supply Chain Management Implications 供应链管理的影响

Finally, from a supply chain management viewpoint, one way to deal with forecast error is to reduce reliance on forecasts, especially long-term forecasts. This requires continuous process improvement to reduce manufacturing lead times through

最后，从供应链管理的观点来看，处理预测错误的一种方法是减少对预测的依赖，特别是长期预测。这需要持续的过程改进以减少生产的前置时间。

• increasing collaboration with customers and suppliers in the area of demand management, especially in sharing demand information and collaboration planning, forecasting, and replenishment
  加强与客户和供应商在需求管理领域的合作，特别是在分享需求信息和协同计划、预测和补货方面

• increasing manufacturing flexibility internally and operations integration externally with customers and suppliers.
  增加内部生产的灵活性，并与客户和供应商进行外部整合。

These steps will enable manufacturers to react quickly to shorter-term forecasts as they become more flexible and efficient.

这些措施将使制造商在更加灵活和高效的情况下，能够迅速对短期预测做出反应。