Академический Документы
Профессиональный Документы
Культура Документы
Marketing Experiments
Introduction
• Experimentation is a key component of the resource
allocation portfolio that allows us to evaluate the
consequence of different marketing actions.
• In this module we will
Understand some basic experimental designs
Follow a case study of a firm that implemented experiments on TV
and the web
Correlation vs Causation
•Does skipping breakfast cause obesity?
http://www.webmd.com/diet/news/20080303/eating-
breakfast-may-beat-teen-obesity
•Alternative explanations:
Physical activity
Lack of sleep
A Basic Experiment
Choose 1,000 How to assign
Customers customers to
Experiment and
Control Group?
Test Group (500) Control Group (500)
Exposed to advertisement highlighting Exposed to old advertisement
new packaging for one month for one month
Sales Lift
Test-Control = 200 units
Sales Lift
Test-Control = 200 units
Sales Lift
Test-Control = 200 units
Sales Lift
Test-Control = 200 units
Sales Lift
Test-Control = (1,200 – 1,000) – (1,100 – 1,000)
= 100 units
Video 6: Designing an
Experiment: Etch A Sketch
Photo Credit: Etcha (Own work) [CC BY-SA 3.0 Photo Credit: By Shikhar Sharma (Own work) [CC BY-SA 3.0
(http://creativecommons.org/licenses/by-sa/3.0) or GFDL (http://creativecommons.org/licenses/by-sa/3.0)], via Wikimedia
(http://www.gnu.org/copyleft/fdl.html)], via Wikimedia Commons Commons
14000
12000
10000
Sales (units)
8000
6000
4000
2000
0
1-Feb 1-Mar 1-Apr 1-May 1-Jun 1-Jul 1-Aug 1-Sep 1-Oct 1-Nov 1-Dec 1-Jan
Video 7: Analyzing an
Experiment: Etch A Sketch
* Control Cities included (a) Charleston, South Carolina; (b) Cleveland, Ohio; (c) Indianapolis, Indiana; and (d)
Pittsburg, Pennsylvania
* Control Cities included (a) Charleston, South Carolina; (b) Cleveland, Ohio; (c) Indianapolis, Indiana; and (d)
Pittsburg, Pennsylvania
* Control Cities included (a) Charleston, South Carolina; (b) Cleveland, Ohio; (c) Indianapolis, Indiana; and (d)
Pittsburg, Pennsylvania
* Control Cities included (a) Charleston, South Carolina; (b) Cleveland, Ohio; (c) Indianapolis, Indiana; and (d)
Pittsburg, Pennsylvania
* Control Cities included (a) Charleston, South Carolina; (b) Cleveland, Ohio; (c) Indianapolis, Indiana; and (d)
Pittsburg, Pennsylvania
* Control Cities included (a) Charleston, South Carolina; (b) Cleveland, Ohio; (c) Indianapolis, Indiana; and (d)
Pittsburg, Pennsylvania
Arizona California
Color Crazy Go Go Glam Color Crazy Go Go Glam
Total/Store/Week
1.8 2.2 0.3 1.2
17 Jun – 17 Jul 2007
Lift 267%
Arizona California
Color Crazy Go Go Glam Color Crazy Go Go Glam
Total/Store/Week
1.8 2.2 0.3 1.2
17 Jun – 17 Jul 2007
Lift 267%
Arizona California
Color Crazy Go Go Glam Color Crazy Go Go Glam
Total/Store/Week
1.8 2.2 0.3 1.2
17 Jun – 17 Jul 2007
Lift 267%
Arizona California
Color Crazy Go Go Glam Color Crazy Go Go Glam
Total/Store/Week
1.8 2.2 0.3 1.2
17 Jun – 17 Jul 2007
Lift 267%
Arizona Test
Arizona Test
Arizona Test
All Retail
National Chain
Sales
Arizona Chain
Sales
Arizona Test
Selling Season
All Retail
Arizona Test
Conclusion
• Experiments assess cause and effect
• Pay attention to:
Design
Gap between test results and field implementation
Difference between test and campaign contexts
• Web experiments are cheaper and faster
Costs of experiments can be variable rather than fixed
• Experiments provide forecasts of expected ROI
This can help with determining campaign budgets