Marketing Engineering Model Marketing Actions Inputs Competitive Actions Observed Market Outputs Market Response Model Environmental Conditions Objectives.

Marketing Engineering Model

Marketing ActionsInputs

Competitive ActionsObserved Market

Outputs

MarketResponse

Model

Environmental Conditions

Objectives e.g., Profits

Product design Price

AdvertisingSelling effort

etc.

Awareness levelPreference level

Sales Level

Evaluation

(5)

Control / Adaptation

(6)

(1) (4)

(2)

(3)

Steps in Creating a Marketing Response Model

1. Develop a relationship between sales and marketing variables

• Sales = f(marketing variables)

2. Calibrate the model• Statistically or judgmentally

3. Create a profit model• Profits = unit volume x contribution margin – fixed costs

4. Optimize• What if or optimum

Linear Response Model:

• Y = a + b1X1 + b2 X2

• Examples:– Medical advertising– Conjoint analysis– Bookbinders Book Club– Price, cart, and coupon exercise

• Easy to estimate, robust, good within certain ranges• Optimum is either zero or infinity• Judgmental – sales at current level of effort and

change in sales for a one unit change in effort.

Weight Loss Response and Profit Model

• Response Model

• Profit Model

AdsAdsbaC 1.366.12ˆˆˆ

AdsAds

AdsAdCostMarC

1300$59$1.366.12

/ˆ

Linear Models - Statistical ConceptsLeast Squares

22

2

1

1

2

)ˆ(

:minimizingbySolve

ˆˆˆˆ

:EquationRegression

)((TSS)SquaresofSumTotal

)1(

)(yofVariance

iii

iiiiiiii

n

i i

n

i i

yye

yyeoreyexbay

yy

n

yy

Linear Models - Statistical ConceptsR2

2

2

2

22

222

)(1

)(

)ˆ(

)ˆ()(

yy

e

TSS

RSSTSS

TSS

ESS

yy

yyR

ESSRSSTSS

yyeyy

i

i

i

i

iii

Nonlinear Response Models: ADBUDG

b – minimum Y

a – maximum Y

c – shape, 0 < c < 1 concave; c > 1 s-shaped

d – works with c to determine specific shape

0)(

XdX

XbabY

c

c

Response function:

Expected Sales

Relative to

Base

Base

R(X )

R(X ) 1.5

R(X ) 1.0

R(X ) 0

0 1.5 Base

Effort Relative to Base

ci

ci

iXd

XbabXr

)()(

ADBUDG Model

• Examples:– Response Modeler: units of marketing effort and sales

– Conglomerate: four cities responding to sales promotion

– Spreadsheet Exercise: (Blue Mountain Coffee) sales response to advertising

– Syntex: 7 products or 9 specialties responding to number of sales calls

– John French: 4 accounts responding to call frequency

Using Solver to Estimate Response Functions

• Locate parameters and choose starting values

• Create columns for independent and dependent variables. Calculate mean of dependent variable.

• Create column of predicted dependent variables based on parameters and independent variables.

• Create column of squared errors between actual and predicted dependent variable. Sum this column.

• Use solver to search over parameters to minimize sum squared errors.

Judgmental Calibration of ADBUDG

• Data: R(Xminimum), R(Xsaturation), R(X1.0), and R(X1.5)

• Parameters:

• a = R(Xsaturation)

• b = R(Xminimum)

• d = (a-R(X1.0))/(R(X1.0)-b)

• c = ln((d*(R(X1.5)-b)/(a-R(X1.5))/ln(1.5)

Judgmental Calibration of ADBUDG

• Data: R(Xminimum), R(Xsaturation), R(X0.5), R(X1.0), and R(X1.5)

• Parameters: • a = R(Xsaturation), b = R(Xminimum)• d = (a-R(X1.0))/(R(X1.0)-b)• Solve for c using least squares over R(X0.5) and R(X1.5)

))(()()(ˆ)(

))(()()(ˆ)(

ˆ5.1

ˆ5.1

5.15.15.15.1

ˆ5.

ˆ5.

5.5.5.5.

dX

XbabXRXRXRe

dX

XbabXRXRXRe

c

c

c

c

Profit Models

• Unit Sales = f(marketing variables) Response Function• Profits = Unit Sales(margin) – fixed costs

• Example: Example on page 38

XXX

XX

XXdX

XbabX

c

c

)25$.1)($(UProfit

)9.53

()7.48.36(7.4)(U

)margin)((UProfit

)()()(U

2.2

2.2

Different Shapes of Multiplicative Model : Y= aXb

0<b<1

b<0

b>1

X=1

Y=a

Linearizable Response Models: Multiplicative Model

2121

2121

21

2121

2211

2211

2121

21

2,1 and e :backTransform

)()()()(

:equationstimate

)()()(

)()()()()(

:equationoriginal of logarithms Take

2,10

bb

ii

bbbb

ibb

XaXXXeY

iba

aLnwhereXLnXLnYLn

E

XLnbXLnbaLn

XLnXLnaLnXaXLnYLn

iXXaXY

Multiplicative Models Cont’d

• Estimate judgmentally– Sales at current level of marketing variable(s)

– Percent change in sales for a percent change in marketing variable i = exponent bi

– Yc=a Xcb

– Solve for a

Multiplicative Models Cont’d

• Examples:– Allegro: Sales = a price-b . Advc

– Nonlinear Advertising Sales Exercise– Forte Hotel Yield Management: Sales = a price-b

• Constant elasticity – exponents are elasticities• Models both increasing (adv) and decreasing

(price) functions as well as both increasing (positive feedback) and decreasing (adv and price) returns

Other Linearizable Models

• Exponential Model: Y = aebx; x > 0– Ln Y = Ln a + bX– Models increasing (b>1) or decreasing (b<1) returns .

• Semi-Log Model: Y = a + b Ln X • Reciprocal Model: Y = a + b/X = a + b (1/X)

– Models saturation

• Quadratic Model: Y = a + bX + c X2

– Supersaturation– Ideal points in MDS– Bass Model

Choose model based on:

• Theory• Fit• Pattern of error terms• Signs and T-statistics of coefficients

Response Function

CurrentSales

Response Function

Min

Max

Current Effort

Sales Response

Effort Level

Elasticity - Percent change in the dependent variable divided by the percent change in the independent variable

= (Y/Y)/(X/X) = (Y/X) (X/Y) = (dy/dx)(X/Y)

• If Y = bX then = 1 For example, if we double X (from x to 2x), Y also doubles (from bx to 2bx), so the percent change in X is always the same as the percent change in Y.

• If Y = a + bX, then Y/X = b(x)/ x = b and X/Y = X / (a + bX) and = (Y/X) (X/Y) = bX/(a+bX) <1 if a>0

Elasticities with a Multiplicative Model

Y = aXb

= (dy/dx)(X/Y)

• dy/dx = a bXb-1

= (a bXb-1) (X/aXb) = (a bXb-1 X)/aXb = b

Elasticity – A way to compare various marketing instruments

= (Y/Y)/(X/X) = (Y/X) (X/Y) = (dy/dx)(X/Y)

(Adv Existing Product) = .05 - .15 (Adv New Product) = .20 - .40• Advertising Long Term = 2X Short Term (Price) = -2.5 (Coupons) = .07

Source:Bucklin and Gupta, 1999

Elasticity in Product Classes where P&G Competes

(Adv) = .039 (Price) = -.541 (Deals) = .092 (Coupons) = .125

Source: Ailawadi, Lehmann, and Neslin 2001

Effect of Increasing Advertising

• Assume 100 units sold at $1.00/unit, 50% contribution margin, advertising elasticity of .22, and 10% A/S ratio

• No change in advertising:– Profit = (100 * $.50) - $10 = $40

• A 50% increase in advertising – sales increase by 11%– New Profit = (111 * $.50) - $15 = $40.50

Conglomerate Market Share Calculations – New York

Promotion Level

E4

Response Multiplier

P56

Non-Deal Prone Share

Deal Prone Share

Total Share

E10

0 .4 69% x .05 = 3.45

31% x .05x .4 = .62

4.07

100% 1 3.45 31% x.05 x 1 = 1.55

5.0

150% 1.64 3.45 31% x .05 x 1.7 = 2.635

6.0

Saturation 2.7 3.45 31% x .05 x 2.7 =4.185

7.635

Aggregate Response Models:Dynamics

• Dynamic response model

• Yt = a0 + a1 Xt + Yt–1

Easy to estimate.

Difficult to interpret correctly

currenteffect

carry-overeffect