Intro to Supply Chain Chapter 02: Forecasting and Demand Planning

Demand Planning

combines statistical forecasting techniques and judgment to construct demand estimate for products and services.

Independent Demand

demand for an item unrelated to the demand for other items, such as finished products or spare parts…demand is forecasted for these items.

Dependent Demand

demand for an item directly related to other items or finished products, such as a component or highly used materials…demand is calculated for these items.

Forecasting

business function that estimates future demand for products so they can be purchased or manufactured in appropriate quantities in advance of need.

• uses historical data for future trends

• uses math to predict future demand

Short-term forecasting

3 or less months, used for tactical decisions (specific, short-term)

Medium-term forecasting

Used to develop a strategy over the next 6-18 months.

Long-term forecasting

Moe than 2 years of predictions, used to detect general trends and identify major turning points.

Good Forecasting

• More effective planning

• Reducing inventory costs and stockouts

• Better customer service

Bad Forecasting

• Root of all problems

• Surplus or Shortage

• Can annoy customers

Factors Influencing Demand

• Market Changes

• Seasons

• Competition

• Pricing

• Changing consumer preferences

etc

Forecasting Error Theory

the farther out into the future you forecast, the more wrong you will be.

Qualitative Forecasting

forecasting based on opinion and intuition, preferred for newer items, used when data is unavailable, best for long-range forecasting, depending on forecaster’s expertise

• Personal Insight

• Jury of Executive Opinion

• Delphi Method

• Historical Analogy

• Customer Surveys

Quantitative forecasting

forecasting using math models and historical data, preferred for older/mature items

• Time Series

  • Naive

  • Simple Moving Average

  • Weighted Moving Average

  • Exponential Smoothing

  • Linear Trend

• Cause N Effect

  • Simple Regression

  • Multiple Regression

Personal Insight (Qualitative Forecasting)

based on the insight of the most experienced/senior person

Adv:

• Fastest and cheapest

• can provide good forecast

Dis:

• relies on one person’s judgement

• unreliable

Jury of Executive Opinion (Qualitative Forecasting)

A panel formed by management that conducts a series of forecasting meetings until they reach a consensus agreement.
Adv:

• multiple experts/experiences

• no time needed to collect data by survey

dis:

• may have bias

• some quieter panelists may be overshadowed by more outspoken panelist (bias)

Delphi Method (Qualitative Forecasting)

Same as Jury of Executive Opinion EXCEPT each expert is collected separately to avoid bias/influences

adv:

• decisions are not caused by groupthink

• useful for new products

• best for long-term forecasting

dis:

• may be some bias

• company must spend time collecting data

• external experts —> less confidentiality

• can be time consuming

Historical Analogy (Qualitative Forecasting)

A judgmental forecasting technique based on identifying a sales history comparable to a present situation, such as the sales history of a similar product.

Adv

• potential to provide lots of info that can be used for new products

• relatively inexpensive

Dis:

• there may be no similar product to historically compare

• No 2 products, even similar are NOT identical, so may be inaccurate even though it’s similar.

Customer Survey (Qualitative Forecasting)

Customers are directly approached and asked for their opinion about a particular product (phone, email, online)

adv:

• it is a direct method of assessing information from the primary source

• simple to administer

• no bias if questions are made well

dis:

• can have bad questions —> unreliable answers

• customers may just ignore survey

• time consuming and costly to survey large population

Time Series (Quantitative Forecasting)

based on the assumption that the future is an extension of the past. History is used to predict future… forecasts for future demand rely on understanding past demand.

Cause N Effect (Quantitative Forecasting)

Assumes that one or more factors predicts future demand

Trend Variations

• Identification of trends is a common starting point for a trend

• S Curve —> initial slow growth, usually new product (slow, then fast, then reduces a bit, then plateaus)

• Asymptotic —> demand reaches a plateau over-time

• Exponential —> Slow start but keeps going up

Random Variations

• Instability in data caused by random occurrences

  • unpredictable weather, labor strikes, war, natural disasters

Seasonal Variations

• Patters of variations within the year, high demand at some time, low at others

  • holidays, seasons

Cyclical Variations

Wave-like patterns that last longer than one year and extend over multiple years

• not easily predictable

  • business cycles, china’s growth, GDP

Naive Forecast

last period’s actual demands are used as this period’s forecast without adjustment.

• simple —> worked last time, works this time

adv

• stable for mature products

• easy to determine

dis

• only for mature products

• variations in demand creates inventory issues

Simple Moving Average

mathematical result that is calculated by averaging the demand of past data.

adv

• short-term forecasting, no seasonality is present

• addresses random variations

dis

• fails to identify seasonal trends

• create shortages when demand increases, and surplus when demand decreases.

Weighted Moving Average

A technique that puts more weight on recent data and less on past data through a weighting factor.

adv

• more accurate than simple weighing

• allows unequal weighing

dis:

• still not a perfect way to forecast

• more inconvenient and costly than exponential smoothing.

Exponential Smoothing

Exponential smoothing weights past observations with exponentially decreasing weights to forecast values…smoothing factor always btwn 0 and 1 and is based on company experts.

adv;

• more responsive to trends than previous methods

• Accepted Because:

  • surprisingly accurate, relatively easy

  • user can understand easily

  • small storage

dis

• still lag behind trends, esp upward trends.

Linear Trend Forecast

Simplistic Forecasting technique that imposes a line of best fit to series historical data

adv:

• provide accurate forecast even through random variations

dis:

• seasonal and cynical variations are softened, making it more beneficial for annual forecast rather than monthly forecast

Selecting Weights for Weighted Moving Average

• Experience and/or trial and error

• recent past is more indicative than past past.

• data is seasonal, weights should reflect appropriately

  • forecasting swimsuit sales for august…should have higher weight on July than December

Simple Linear Regression

models the relationship between a single independent variable and a dependent variable (demand) by fitting a linear equation to the observed data.

Multiple Linear Regression

models the relationship between a 2 or more independent variable and a dependent variable (demand) by fitting a linear equation to the observed data.

Fundamentals of Forecasting

• Forecasting is often wrong, but the skill is to detect and adapt quickly

• the more specific the forecast is, the less accurate it is

• it is easier to forecast smaller time periods than larger time periods.

• simple forecasting trumps methodological forecasting

• one correct forecast does not mean your forecasting method is perfect. (could’ve been chance)

• If you don’t use your data regularly, trust it less.

• All trends eventually end

• Hard to eliminate bias, so know that most forecasts have bias.

• Tech is not the answer to better forecasting.

  • it is a tool to make you more accurate..

Impacts of social media

• Evaluate brand health (understanding of customer’s feelings on product/service)

• Improve demand prediction (can understand if product is trendy or not)

• Address a crisis (social media can show negative thoughts of public)

• Research competition (social media often compares similar products)

Forecast Error

measured in units/percents

Error Measurement

Critical in tracking forecasting accuracy, monitoring exceptions, benchmarking forecasting process.

Forecast Error Value (FEV) Formula

FEV = Actual Demand - Forecasted Demand

FEV% = ((A - F) / A) x 100

Mean Absolute Deviation (MAD)

measures the size of the forecast error in units.

• Σ|A - F|) / n

• n = number of time periods

Mean Absolute Percent Error (MAPE)

measures size of error in %

• Σ ((| A - F |)/ A))/ n

Bullwhip Effect

small shifts in consumer demand cause increasingly larger fluctuations in orders and inventory as you move up the supply chain

Alleviate the Bullwhip Effect

• Collaboration (sharing info and data)

• Sync Supply Chain (plan better between independent suppliers)

• Reducing Inventory (using JIT or quick adaptions to effectively have perfect inventory, no surplus or shortage)

Collaborative Planning, Forecasting & Replenishment (CPFR)

business practice that combines the intelligence of multiple trading partners who share their plans, forecasts, and delivery schedules to ensure smoothness in supply.

• Better customer service

• lower inventory costs

• improved quality

Supply Chain Agility

ability to quickly and efficiently respond to changes in demand or supply without sacrificing

• reconfiguring people, processes, and goods to adapt to new market conditions

• Helps avoid knee-jerk reactions to vulnerable markets