Market Mix Modeling Overview for Data Analytics

Market mix modeling, ad stock, power, lag, and channel effectiveness

1. © 2002 Marketing Management Analytics – www.mma.com
Marketing Mix Models

2. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:2
What is Marketing Mix
Modeling?
An analytical approach which quantifies
the sales effect of marketing activity and
the financial return on that investment.
The output is used to simulate the effects
of alternative marketing plans and forecast
sales into the future.
All work, for all clients, is
completely custom.

3. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:3
Why do marketers do
Marketing Mix Analysis?
To get a true empirical relationship between
marketing activity and sales.
To conduct Return on Investment analysis.
Benchmarking (are we the same / better /
worse than we were last year?).
To find out how one business unit's marketing
interacts (if at all) with another’s (media “halo”).
To learn upside potential and downside risk in
changing marketing spend.

4. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:4
Issues that can be addressed...
How much should we spend?
What is the recommended level of
spending?
How much is enough?
How should we spend it?
Which brands should receive support and
how should our budget be allocated?
When should we spend it?
Before or after a price increase?
Immediately before a competitive launch?
Where should it be spent?
National vs. Local
Cites, Rural
1122
3344

5. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:5
Marketing Mix Elements
Step 1
determine the contribution
to incremental volume from
each marketing mix
element
Marketing Spending
Step 2
overlay marketing
spending for each mix
element to evaluate ROI
Custom Software
Step 3
utilize MMA’s custom software for
optimization and simulation of
marketing spending and advertising
return
Marketing Mix Modeling

6. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:6
Mix Modeling Timeline
Data
Collection
& Validation
Model Specification
& Validation
Analysis
& Review
12-13 weeks
Project Steps & Timing
4-5 Weeks
Kick-Off Final
Presentation
4 Weeks 3 Weeks1
4 Weeks 3-5 Weeks 1

7. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:7
Data Collection
Media
Nielsen
CMR
Agency
Integrated
Database
Sales
(Internal,
Syndicated)
Promotion
Act Media
Catalina
Internal
Shipments
Web
Promotion
Financial
1
2
Data Analysis
System
The capability to integrate several
disparate data sources
The ability to identify what
variables should be included in
your model.

8. © 2002 Marketing Management Analytics – www.mma.com
Marketing Mix Modeling
(How is it Done?)

9. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:9
Modeling: What to measure?
Start with a Dependent Variable :
• Sales, or
• Awareness, or
• New Customers or
• Brand Interest
• Sales by Segment (i.e. Heavy vs. Light)

10. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:10
Modeling: What are the Drivers
Gather information on all variables that possibly influence your dependent variable
Independent Variables:
•TV
•Print
•Radio
•Outdoor
•Internet
•Promotions
•PR
•Coupons
•Sampling
•Direct Mail
•Competitive
•Economic
•Environment/Weather
•Industry Trends
•Etc.

11. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:11
The models relate changes in sales to the
changes in marketing support and other
causal factors present during each week.
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77
0
10,000
20,000
30,000
40,000
50,000
60,000
70,000
High
Activity
Low
Activity
Sales
High
Activity
Week 1 13 22 33 41 49 58 67 70 78
TV GRPs - 214 142 - 50 - 89 284 65 -
Radio GRPS - 100 - 60 - 40 - 40 - -
In-Store Promotions 10 75 - 30 5 - - 20 - 25
Support

12. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:12
Model Validation
Face Validity
Holdout Tests
Model
Stats
Other
- S igns
- Vo lum e Trends
- Trade Contribution
H oldout Errors
&
Patterns
R-Square, Durbin
W atson, A vg % E rror,
F-S tats, T-S tats, etc.
O ther validations
against prior learnings
or known relationships

13. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:13
Models need to be able to predict volume in the
near future. Model holdout validations can help
evaluate a model’s predictiveness.
The model should “explain” as much of the
weekly variance in sales as possible. The R-
Squared statistic quantifies a models fit.
Model residuals should exhibit no patterns and
significant weekly misses should be understood.
Average errors and Durbin Watson statistics can help
with these issues.
Analyzed Market Aggregate
5/2/938/1/9310/31/931/30/945/1/947/31/9410/30/941/29/954/30/95
-100,000
0
100,000
200,000
300,000
400,000
500,000
600,000
700,000
Sales
Fit Period
Average
Error: 2.3%
Validation
Period
Average
Error: 2.8%
104 Weeks Ending 4/95
Model EstimateActual SalesResiduals
Basic Model Stats & Holdout
Validation

14. © 2002 Marketing Management Analytics – www.mma.com
Marketing Mix Modeling
What Does It Answer?

15. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:15
Incremental volume will be broken out
into the various marketing elements.
23.6%
50.0%
7.5%
7.5%
3.8%
7.5%
Prior
Year
Current Year
Volume Contribution - Example
Direct TV Base Promotions Print Radio Corp TV
Total Volume 110.0 MMTotal Volume 102.0 MM
7.8%
3.9%
4.9%
9.7%
54.2%
19.5%

16. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:16
Comparing ROI efficiencies across vehicles
shows that all TV and Trade spending is
very efficient.
$3.3
$2.6
$3.6
$3.8
$4.2
$3.0
TV Adv Trade SMF Print Radio Total

R
e
t
u
r
n

P
e
r

$

M
a
r
k
e
t
i
n
g
Marketing Efficiencies - ExampleMarketing Efficiencies - Example
(Incremental $ Sales Per Marketing $)(Incremental $ Sales Per Marketing $)

17. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:17
Changes in yearly support, resulting contribution
to sales and efficiency will be reported.
1256
872
1.9%
2.2%
$3.41
$2.69
Direct TV Advertising Performance - Summary
(Example)
Support
TRPs
% Contribution
% of Total Brand Volume
Efficiency
Cost per Incr. Vol
Prior YearCurrent Year

18. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:18
Evaluation of Copy Effectiveness
TV Advertising and Cost EfficiencyTV Advertising and Cost Efficiency
Copy A vs Copy BCopy A vs Copy B
Volume Effectiveness
Volume Per MM Impressions
Cost Efficiency*
Cost Per Incremental Unit
$0.42
$0.51
Prior Year - Copy A Current Year - Copy B

Copy A
Total Copy A = 7,535
7,458
7,987
8,586
All Other New Flavor Copy B

19. Copyright © 2002 Marketing Management Analytics ‌

www.mma.com
Page:19
ROI efficiencies can also be compared across categories
by brand, department and marketing investment.
1.24
0.84
0.66
1.10
0.85
1.24
1.20
0.93
1.98
1.34
0.68
0.57
0.48
0.40
1.17
0.90
2.10
2.34
0.74
0.70
0.61
1.34
0.54
0.68
1.50
0.45
0.56
1.17
0.80
0.65
1.30
1.20
1.14
1.34
0.94
1.30Brand 1
Brand 2
Brand 3
Brand 4
Brand 5
Brand 6
Brand 7
Brand 8
Brand 9
1.22
0.84
0.76
0.60
0.72
2.54
0.61
0.65
3.02
Profit per $1 Invested
Total Mix TV Media Print In-storeCons. Promo.