Transcript Slide 1

Marketing Optimization
Using SAS
Randy Sherrod
[email protected]
March 2008
1
Discussion Topics
 What is the impact of marketing investment on
business metrics, e.g. sales?
 How can we determine the level of marketing
investment that optimizes return?
 What data is required?
 What techniques are available?
2
Overview of Analytic Process
Data Collection: Historical Sales,
Distribution Channel, Pricing,
Marketing Investment, Competitor
Behavior, International
Macroeconomic Data
Data
Collection
Modeling
Optimization: Use the econometric
model as an input to an optimization
engine that identifies optimal sales and
marketing investment levels
Optimization
Modeling: Develop Econometric
Model(s) relating Historical Sales
with important drivers of business
Results
Results: Compares optimized
investment levels with actual,
yielding insight into opportunity to
increase sales through marketing
reallocation
Source: Cisco SMO
3
Econometric Model Quantifies Relationship Between
Bookings and Drivers – First Step to Driving Optimal
Resource Allocation
Inputs
Outputs
 Pricing
 Distribution Channel:
Sales People, Resellers,
etc.
 Marketing Investment
Econometric
Model
 Driver Elasticities
 Predicted Sales
 Macroeconomics
 Competitor Dynamics
 etc
Source: Cisco SMO
4
Elasticity Measurements that Quantify the Relationship
Between Drivers and Bookings
Definition
Relationship of Sales & Marketing
Investment
Elasticity measures the responsiveness
of Sales to changes in drivers,
calculated as:
Diminishing Returns, Elasticity<1
x
%∆ sales / %∆ driver
x
x
x
x
Relevant Cases
Sales
x
x
x
x
x
x
Elasticity<1 (inelastic):
Percentage change in sales is less than
percentage change in driver (ex. Increasing
marketing investment by 1% leads to less
than 1% increase in sales)
Elasticity>1 (elastic):
Marketing Investment
Percentage change in sales is more than
percentage change in driver
Source: Cisco SMO
5
Background Observations

determine
the optimal level
of sales
force and
 How
Whattorange
of elasticities
can we
expect?
marketing?

ForceSales
(+) Force=$400M, Marketing=$50M,
 Sales
Initial Values:
Sales=$1B.
 Total
Marketing
(+)
 Estimated Elasticities: Sales Force=0.40,

TV (+)
Marketing=0.20

Suppose
there is an
Paid Search
(+)additional $40M to allocate, how
do you split between Sales Force and Marketing to
 GDP
(+) Sales?
maximize
 $40M=10% of Sales Force0.40*0.10*$1B=$40M
 What
is the impact
increase
in Salesof GDP on marketing and
sales? What might this mean for the optimal
 $40M=80% of Marketing0.20*0.80*$1B=$160M
level increase
of investment?
in Sales
6
Modeling Possibilities
Framework
Log-linear models with SAS:
1. Proc GLM
2. Proc Reg
3. Proc Surveyreg
4. Proc Genmod
5. Proc Mixed
6. etc.
Output from these procedures quantifies
the impact of marketing on sales
Source: Cisco SMO
7
Modeling Details
Framework
Log-linear model with customer-level fixed effects:
Log Salesit=αi+β1log Competitor Advertisingt-1 + β2log Sales Forcet-1 + β3log
Marketingt-1 + β5log Cust Satisfactionit-1 + β6log GDPt-1 + Seasonality
Where:
i=customer
t=time
(-1)=lag 1 QTR
1. Imposes constant elasticity
2. Allows for many possible
response curve shapes
3. Explicitly accounts for synergies
between drivers
Source: Cisco SMO
8
Modeling Details cont.
SAS Implementation
proc surveyreg;
class customer;
model log_sales=customer log_comp_advertising_1 log_sales_1
log_marketing_1 log_cust_satisfaction_1 log_gdp_1 q4 /noint solution;
cluster time;
Creates cluster-consistent standard errors
quit;
Estimated model can then be solved for optimal levels using proc
optmodel.
Source: Cisco SMO
9
Questions?
10