Forecasting in Business Planning

Inputs
Market Conditions Competitor Action Consumer Tastes Products Life Cycle Season Customers plans

Economic Outlook
Business Cycle Status Leading Indicators-Stock Prices, Bond Yields, Material Prices, Business Failures, money Supply, Unemployment

Forecasting Method(s) Or Model(s)

Management Team

Outputs Estimated Demands for each Product in each Time Period Other Outputs

Other Factors
Legal, Political, Sociological, Cultural

Processor

Forecast Errors Feedback

Sales Forecast Forecast and Demand for Each Product In Each Time Period

Production Capacity Available Resources Risk Aversion Experience Personal Values and Motives Social and Cultural Values Other Factors

Forecasting Methods

Forecasting

Qualitative Or Judgmental

Quantitative Or Statistical

Projective

Causal

Forecasting Basics
Types
Qualitative --- based on experience, judgment, knowledge; Quantitative --- based on data, statistics;

Methods
time series models (e.g. moving averages and exponential smoothing); causal models (e.g. regression)

Forecasting Approaches(1)
JUDGEMENTAL APPROACHES: The essence of the judgmental approach is to address the forecasting issue by assuming that someone else knows and can tell you the right answer. They could be experts or opinion leaders. EXPERIMENTAL APPROACHES: When an item is "new" and when there is no other information upon which to base a forecast, is to conduct a demand experiment on a small group of customers and extrapolated to the wider population. Test marketing is an example of this approach. RELATIONAL/CAUSAL APPROACHES: There is a reason why people buy our product. If we can understand what that reason (or set of reasons) is, we can use that understanding to develop a demand forecast. They seek to establish product demand relationships to relevant factors and/or variables e.g. hot weather to cold drinks consumption. TIME SERIES APPROACHES: A time series is a collection of observations of welldefined data items obtained through repeated measurements over time.

Forecasting Approaches(2)
In general, judgment and experimental approaches tend be more qualitative While relationship/causal and time series approaches tend be more quantitative

Still, these qualitative methods are also scientifically done with results that are expressed in indicative numbers and broad trends
Time series/causal methods are completely based on statistical methods and principles

Qualitative Approach
Qualitative Approach
Usually based on judgments about causal factors that underlie the demand of particular products or services Do not require a demand history for the product or service, therefore are useful for new products/services Approaches vary in sophistication from scientifically conducted surveys to intuitive hunches about future events. The approach/method that is appropriate depends on a products life cycle stage

Qualitative Methods
Educated guess Executive committee consensus Delphi method Survey of sales force Survey of customers Historical analogy Market research

Forecasting Methods
-judgmental approach(a)
Surveys - this involves a bottom up method where each individual/respondent contributes to the overall result; this could be for product demand or sales forecasting ; also for opinion surveys amongst employees, citizen groups or voter groups for election polls Sales Force Composites- where the similar bottom up approach is used for building up sales forecasts on any criteria like region-wise or product wise sales territory groupings from sales force personnel Consensus of Executive Opinion -normally used in strategy formulation by sought opinions from key organizational stakeholders- managers, suppliers, customers, bankers and shareholders Historical analogy- used for forecasting new product demand as similar to the previously introduced new product benefiting from its immediacy that same demand influencing factors will apply

Forecasting Methods
-judgmental approach(b)
Consensus thro Delphi method especially for new product developments and technology trends forecasting It is the most formal judgmental method and has a well defined process and overcomes most of the problems of earlier consensus by executive opinion This involves sending out questionnaires to a panel of experts regarding a forecast subject. Their replies are analyzed, summarized, processed and redistributed to the panel for revisions in light of others arguments and viewpoints. By going thro such an iterative process say 3-4 times, the final panel forecast is considered as fairly accurate and authentic Yet, difficulties do exist in planning, administering and integrating member views into a meaningful whole Course Booklet has a separate chapter on the Delphi method( page 107 onwards)

Forecasting Methods
-judgmental approach(c) Method
Short term accuracy
POOR POOR TO FAIR VERY GOOD POOR FAIR TO GOOD

Medium term accuracy

POOR POOR TO FAIR GOOD FAIR TO GOOD FAIR TO GOOD

Long term accuracy

VERY POOR POOR FAIR FAIR TO GOOD FAIR TO GOOD

Cost

Personal insights Panel consensus Market survey Historical analogy Delphi method

VERY LOW LOW VERY HIGH MEDIUM HIGH

Forecasting Methods
- experimental approach
Customer surveys- thro extensive formal market research using personal or mail interviews, and newly thro internet modes; also build demand models for a new product by an aggregated approach Consumer panels- particularly used in initial stages of product development and design to match product attributes to customer expectations Test marketing- often used after product development but before national launches by starting in a selected target market/geography to understand any problems or issues to fine-tune marketing plans and avoid costly mistakes before going in a big way Customer buying data bases- based on selected and accepted individuals/families on their buying behavior , patterns and expenditures captured using electronic means direct from retailer sales data; gives extensive clues on buying factors, customer attitudes, brand loyalty and brand switching and response to promotional offers

Forecasting Methods
- relationship/causal approach(1)
Its basic premise is that relationships exist between various independent demand variables( like population, income, disposable incomes, age, sex etc to consumer needs/wants/expectations( dependent variables) Before linking these, we need to find the nature and extent of these causes/relationships in mathematical terms as regression( linear/multiple)equations Once done, they can be used to forecast the dependent variable for available independent variables Various types of causal methods follow in next slide

Forecasting Methods
- relationship/causal approach(2)
Econometric models like discrete choice and multiple regression models
used in large-scale or macro-level economic forecasting Input-output models used to estimate the flow of goods between markets

and industries, again in macro-economic situations

Simulation models used to establish raw materials and components demand based on MRP schedules , driven by keyed-in product sales

forecasts; to reflect market realities and imitate customer choices

Life-cycle models which recognize product demand changes during its various stages(i.e. introduction/growth/maturity/decline) particularly in

short life cycle sectors like fashion and technology

Forecasting Methods
- time series approach(1)
Fundamentally, uses historical demand/sales data to determine future demand Basic assumptions are that :
Past data/information is available This data/information can be quantified Past patterns will continue into the future and projections made( though in reality may not always be the case !)

They involve statistical methods of understanding and explaining patterns in time series data( like constant series e.g. annual rainfall; trends e.g. growing expenditure with incomes; seasonal series e.g. umbrella demand during rainy season; and any random/unexplained noise where actual value= underlying pattern+ random noise)

Forecasting Methods
-time series approach(2)

Static elements:
Trend Seasonal Cyclical Random
Moving average Simple exponential smoothing Exponential smoothing (with trend) Exponential smoothing (with trend and seasonality)

Adaptive elements:

7-15

Time Series
-static elements
Trend component- persistent overall downward or upward pattern; due to population, technology or long term movement Seasonal component- regular up and down fluctuations due to weather and/or seasons whose pattern repeats every year Cyclical component- repeated up and down movements; due to economic or business cycles lasting beyond one year but say every 5-6 years Random component- erratic, unsystematic, residual fluctuations due to random events or occurrences like one time drought or flood events

Time Series: Moving average

The moving average model uses the last t periods in order to predict demand in period t+1. There can be two types of moving average models: simple moving average and weighted moving average The moving average model assumption is that the most accurate prediction of future demand is a simple (linear) combination of past demand.

Time series: simple moving average

In the simple moving average models the forecast value is At + At-1 + + At-n Ft+1 = n

is the current period.

Ft+1 is the forecast for next period

n A is the forecasting horizon (how far back we look), is the actual sales figure from each period.

Example: forecasting sales at Kroger

Kroger sells (among other stuff) bottled spring water

Month Jan Feb Mar

Bottles 1,325 1,353 1,305

Apr
May Jun Jul

1,275
1,210 1,195 ?

What will the sales be for July?

What if we use a 3-month simple moving average?

FJul =

AJun + AMay + AApr 3

= 1,227

What if we use a 5-month simple moving average?

FJul =

AJun + AMay + AApr + AMar + AFeb

5

= 1,268

1400 1350 1300 1250 1200 1150 1100 1050 1000 0 1 2 3 4 5 6 7 8

5-month MA forecast 3-month MA forecast

What do we observe?
5-month average smoothes data more; 3-month average more responsive

Stability versus responsiveness in moving averages

Time series: weighted moving average

We may want to give more importance to some of the data

Ft+1 =

wt At + wt-1 At-1 + + wt-n At-n

wt + wt-1 + + wt-n = 1

is the current period.

Ft+1 is the forecast for next period n A w is the forecasting horizon (how far back we look), is the actual sales figure from each period. is the importance (weight) we give to each period

Why do we need the WMA models?

Because of the ability to give more importance to what happened recently, without losing the impact of the past.
Demand for Mercedes E-class
Actual demand (past sales) Prediction when using 6-month SMA Prediction when using 6-months WMA

Jan Feb Mar Apr May Jun Jul Aug

Time

For a 6-month SMA, attributing equal weights to all past data we miss the downward trend

Example: Kroger sales of bottled water

Month
Jan Feb Mar Apr

Bottles
1,325 1,353 1,305 1,275

What will be the sales for July?

May
Jun Jul

1,210
1,195 ?

6-month simple moving average

FJul =

AJun + AMay + AApr + AMar + AFeb + AJan 6

= 1,277

In other words, because we used equal weights, a slight downward trend that actually exists is not observed

What if we use a weighted moving average?

Make the weights for the last three months more than the first three months

6-month SMA July Forecast 1,277

WMA 40% / 60% 1,267

WMA 30% / 70% 1,257

WMA 20% / 80% 1,247

The higher the importance we give to recent data, the more we pick up the declining trend in our forecast.

How do we choose weights?

1. Depending on the importance that we feel past data has
2. Depending on known seasonality (weights of past data can also be zero).

WMA is better than SMA because of the ability to vary the weights!

Time Series: Exponential Smoothing (ES)

Main idea: The prediction of the future depends mostly on the most recent observation, and on the error for the latest forecast.

Smoothin g constant alpha

Denotes the importance of the past error

Why use exponential smoothing?

1. Uses less storage space for data 2. Extremely accurate 3. Easy to understand 4. Little calculation complexity

5. There are simple accuracy tests

Exponential smoothing: the method

Assume that we are currently in period t. We calculated the forecast for the last period (Ft-1) and we know the actual demand last period (At-1)

Ft Ft1 ( At1 Ft1 )

The smoothing constant expresses how much our forecast will react to observed differences If is low: there is little reaction to differences. If is high: there is a lot of reaction to differences.

Example: bottled water at Kroger

Month Jan Feb Mar Apr May Jun Actual 1,325 1,353 1,305 1,275 1,210 ? Forecasted 1,370 1,361 1,359 1,349 1,334 1,309
= 0.2

Example: bottled water at Kroger

Month Jan Feb Mar Apr May Jun Actual 1,325 1,353 1,305 1,275 1,210 ? Forecasted 1,370 1,334 1,349 1,314 1,283 1,225
= 0.8

Impact of the smoothing constant

1380 1360 1340 1320 1300 1280 1260 1240 1220 1200 0 1 2 3 4 5 6 7 Actual a = 0.2 a = 0.8

Trend..
What do you think will happen to a moving average or exponential smoothing model when there is a trend in the data?

Impact of trend

Sales
Actual Data

Forecast

Regular exponential smoothing will always lag behind the trend. Can we include trend analysis in exponential smoothing?

Month

Exponential smoothing with trend

FIT: Forecast including trend

FITt Ft Tt

: Trend smoothing constant

Ft FITt 1 (At 1 FITt 1 ) Tt Tt 1 (Ft FITt 1 )

The idea is that the two effects are decoupled, (F is the forecast without trend and T is the trend component)

Example: bottled water at Kroger

At Jan Feb Mar Apr May Jun 1325 1353 1305 1275 1210 Ft 1380 1334 1344 1311 1278 1218 Tt -10 -28 -9 -21 -27 -43 FITt 1370 1306 1334 1290 1251 1175 = 0.8 = 0.5

Exponential Smoothing with Trend

1400 1350 Actual 1300 1250 1200 1150 0 1 2 3 4 5 6 7 a = 0.2 a = 0.8 a = 0.8, d = 0.5

Linear regression in forecasting

Linear regression is based on

1. Fitting a straight line to data

2. Explaining the change in one variable through changes in other variables.

dependent variable = a + b (independent variable)

By using linear regression, we are trying to explore which independent variables affect the dependent variable

Example: do people drink more when its cold?

Alcohol Sales Which line best fits the data?

Average Monthly Temperature

The best line is the one that minimizes the error

The predicted line is

Y a bX
So, the error is

i yi - Yi
Where: is the error y is the observed value Y is the predicted value

Least Squares Method of Linear Regression

The goal of LSM is to minimize the sum of squared errors

Min

i2

What does that mean?

Alcohol Sales

So LSM tries to minimize the distance between the line and the points! Average Monthly Temperature

Least Squares Method of Linear Regression

Then the line is defined by

Y a bX
a y bx

xy nxy b x nx
2 2

How can we compare across forecasting models?

We need a metric that provides estimation of accuracy

Errors can be:

Forecast Error

1. biased (consistent) 2. random

Forecast error = Difference between actual and forecasted value (also known as residual)

Measuring Accuracy: MFE

MFE = Mean Forecast Error (Bias)

It is the average error in the observations

MFE

A F
i 1 t

1. A more positive or negative MFE implies worse performance; the forecast is biased.

Measuring Accuracy: MAD

MAD = Mean Absolute Deviation

It is the average absolute error in the observations

MAD

A F
i1 t

1. Higher MAD implies worse performance. 2. If errors are normally distributed, then =1.25MAD

MFE & MAD: A Dartboard Analogy

Low MFE & MAD: The forecast errors are small & unbiased

An Analogy (contd)

Low MFE but high MAD: On average, the arrows hit the bullseye (so much for averages!)

MFE & MAD: An Analogy

High MFE & MAD: The forecasts are inaccurate & biased

Key Point
Forecast must be measured for accuracy! The most common means of doing so is by measuring the either the mean absolute deviation or the standard deviation of the forecast error

Measuring Accuracy: Tracking signal

The tracking signal is a measure of how often our estimations have been above or below the actual value. It is used to decide when to re-evaluate using a model.

RSFE (At Ft )
i1

RSFE TS MAD

Positive tracking signal: most of the time actual values are above our forecasted values Negative tracking signal: most of the time actual values are below our forecasted values
If TS > 4 or < -4, investigate!

Example: bottled water at Kroger

Month Jan Feb Mar Actual 1,325 1,353 1,305 Forecast 1,370 1,361 1,359

Month
Jan Feb Mar Apr May Jun

Actual
1,325 1,353 1,305 1,275 1,210 1,195

Forecast
1370 1306 1334 1290 1251 1175

Apr
May Jun

1,275
1,210 1,195

1,349
1,334 1,309

Exponential Smoothing ( = 0.2)

Forecasting with trend ( = 0.8) ( = 0.5)

Question: Which one is better?

Bottled water at Kroger: compare MAD and TS

MAD Exponential Smoothing Forecast Including Trend TS

70

- 6.0

33

- 2.0

We observe that FIT performs a lot better than ES Conclusion: Probably there is trend in the data which Exponential smoothing cannot capture

Which Forecasting Method Should You Use

Gather the historical data of what you want to forecast Divide data into initiation set and evaluation set Use the first set to develop the models Use the second set to evaluate Compare the MADs and MFEs of each model