Fundamentals of Operations Mgmt Forecasting

1
Fundamentals of Operations MgmtForecasting
Managing UncertaintyAug 23, 2012
2
Forecasting

• A statement about the future value of a variable
  of interest
• Future Sales
• Weather
• Stock Prices
• Other Short term and Long term estimates
• Several Methods
• Quantitative
• History and Patterns
• Leading Indicators / Associations (Housing Starts
  Furniture)
• Qualitative
• Judgment
• Consensus

Used for making informed Decisions and taking
Actions based on those decisions
3
Forecasting

• The quality of the Demand Forecast makes a MAJOR
  IMPACT (Positive or Negative) on
• Revenue
• Market Share
• Capital Investment
• Hiring
• Inventory
• Cost
• Profit

Cisco wrote down 2.5B in inventory in 2001
4
Forecast can be constructed along many dimensions

• Product Complexity / Granularity
• Product Line
• Family
• Model
• SKU
• Time Horizon
• Daily
• Weekly
• Monthly
• Annual
• Multi-year
• Hourly
• Unit Of Measure
• Dollars
• Units

5
A Demand Forecast serves many Purposes
WHAT is done and WHY?
Region
Product Line
Channel
Features
Product
Customer
Scheduling Factory Volumes Materials Planning
Balancing Factory Capacity Assessing Direct Cost
_at_ Mixes Analyzing Absorption implications
Revenue Planning Revenue Scenarios Allocation
Criteria Commissions Quotas
Estimating TAM and Share Pricing Targets Programs
Promotions Margins _at_ Mixes Message to Analysts
Business Need / Benefit
6
How different Functions use Forecast information
7
Constructing a Forecast
8
A forecast can be very complicated (or somewhat
simplified)
9
Features Common to all Forecasts

• Generally assumes that what drove past
  performance and behavior will drive future
  performance and behavior
• Credit Rating
• Insurance Rates
• Other
• More accurate for groups vs. individuals
• Accuracy decreases as time horizon increases

Forecasts WILL be wrong the goal is to predict
as closely as possible
10
Start with what you KNOW

• How many people will attend the next Giants game?
• Tickets already sold
• Patterns of walk up sales
• Visiting team
• Weather
• School day
• Other
• How many Sewing Machines will Singer sell this
  week?
• Orders in Backlog
• Inventory in Stores
• Production capacity
• Household Budget
• Rent
• Car Payment
• Bills
• Rest of money

11
Selecting the most useful Forecasting technique(s)

• No single technique works in every situation
• Two most important factors
• Cost
• Accuracy
• Other factors include the availability of
• Historical data
• Computers
• Time needed to gather and analyze the data
• Forecast horizon

12
Time Series Forecasts (and Behaviors)
Trend - long-term movement in data Seasonality -
short-term regular variations in data Cycle
wavelike variations of more than one years
duration Irregular variations - caused by
unusual circumstances Random variations - caused
by chance
13
Graphs help interpret Time Series data (Figure
3.1)
Irregularvariation
Trend
Cycles
90
89
88
Seasonal variations
From Stevenson, Operations Management, Ninth
Edition, McGraw Hill Irwin
14
From Stevenson, Operations Management, Ninth
Edition, McGraw Hill Irwin
15
Managing the Forecastwith Actual Results
16
Tracking Forecast Accuracy

• Level of Aggregation
• Item (Mix of individual SKUs)
• Family
• Product Line
• Channel
• Customers
• Quantity
• Time Buckets
• Final consumer sales

Absolute values and square roots eliminate the
possibility of positive and negative variances
canceling each other out key for Mix tracking
less critical for Revenue tracking
Regular tracking and monitoring with enable
Demand SENSING, as well as contribute to
increased accuracy of future forecasts
17
Consuming a Forecast

• Was it what I expected?
• Extra?
• Less?
• Early?
• Late?
• Impact on Planning and the Business

18
Historical forecast performance at ONeill
Forecasts and actual demand for surf wet-suits
from the previous season
19
Empirical distribution function of forecast
accuracy

• Start by evaluating the actual to forecast ratio
  (the A/F ratio) from past observations.

20
Causal Factors

• External
• Market conditions (e.g. paintings when the
  Painter passes away, Michael Jackson)
• New competition
• Competitors cannot supply
• Internal
• Pricing
• Promotions
• Incentives

21
Relevance of SUPPLY on Forecasts

• Historical Sales does not always equal historical
  Demand
• Stockouts
• Substitutions
• Causal Factors may distort the analysis (pricing,
  promotions, competitor performance)
• Scarcity Behavior
• Allocation
• Advance buying
• Hedging
• Hording

22
The Relevance of Time on Forecasts
23
Forecasts vary in Objective Scope depending on
Horizon
From Stevenson, Operations Management, Ninth
Edition, McGraw Hill Irwin
24
Forecast accuracy varies over time
Over
Expected SKU Errors
0
1
2
3
4

n
Time in Future (Weeks)
Under
The further into the future, the harder to
predict details with accuracy
25
Detailed Product Forecast Accuracy will vary by
Time Horizon
Current Week should approach 100
Current Month should be greater than 80
Quarter should be at least 70
26
Relationship of Lead Time, Forecast, Inventory,
and Cost
27
Hammer 3/2 timeline
Generate forecast
of demand and
submit an order
to TEC
Spring selling season
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Receive order
Leftover
from TEC at the
units are
end of the
discounted
month
28
Additional Points
29
Key Points from Newsvendor Case

• Cost of Underforecasting (Underage cost)
• Cost of Overforecasting (Overage cost)
• Using Truly Variable Cost when making decisions
  on incremental volumes
• The role of Per Cent Probability in decision
  making
• A/F Ratio (Actual Demand / Forecast)
• Bias
• Point Forecast vs. Range Forecast
• Managerial Lessons
• Understand VARIABILITY of the Demand
• Track ACTUAL Demand
• Track PREVIOUS FORECAST accuracy to future
  actuals

30
Balancing the risk and benefit of ordering a unit

• Ordering one more unit increases the chance of
  overage
• Expected loss on the Qth unit Co x F(Q)
• F(Q) Distribution function of demand
  ProbDemand lt Q)
• but the benefit/gain of ordering one more unit
  is the reduction in the chance of underage
• Expected gain on the Qth unit Cu x (1-F(Q))

• As more units are ordered, the expected benefit
  from ordering one unit decreases while the
  expected loss of ordering one more unit increases.

31
Newsvendor model performance measures

• For any order quantity we would like to evaluate
  the following performance measures
• In-stock probability
• Probability all demand is satisfied
• Stockout probability
• Probability some demand is lost
• Expected lost sales
• The expected number of units by which demand will
  exceed the order quantity
• Expected sales
• The expected number of units sold.
• Expected left over inventory
• The expected number of units left over after
  demand (but before salvaging)
• Expected profit

32
Other Points to consider

• Do not second guess the forecast
• Significant judgment and even debate contribute
  to the final forecast. But once the forecast is
  finalized it then becomes the Demand Plan of
  Record for the enterprise
• and do not say, If only we got a better
  forecast
• The forecast should be generated as a team and
  managed as a team
• Do not use the forecast to position or influence
  Supply
• The forecast is an UNCONSTRAINED, honest estimate
  of future demand. The forecast becomes one of
  several INPUTS for an integrated Supply Pla
• Consider Contribution Margin when considering
  capacity tradeoffs
• Highest margin can justify bias toward
  overbuilding
• Lowest margin can justify underbuilding
• Product Transitions are very difficult to
  forecast, but require special attention and
  monitoring
• New Product Introduction
• End Of Life

Peter Drucker The best way to predict the
future is to CONTROL it