You are on page 1of 36

Experimentation

Case - In store Advertisements


Rite aid drug co. conducted an experiment to examine the effectiveness of in store audio advertisements to induce POP buys. Twenty drug stores were selected. Half were selected as test stores advts were aired The other half were control stores not aired.

Data was tracked by unit volumes and dollar sales Data was tracked for 7 days before the experiment, during the 4 week experiment and 7 days after the experiment. Products kitchen knivessmall appliances Result sales doubled where advts were aired.

POP studies sponsors Pfizer, Pepsi, P& G etc

Causal Research Design


A process where events occur in a setting at the discretion of the experimenter, and controls are used to identify the sources of variation in subjects response.

Necessary Conditions for Making Causal Inferences


Concomitant variation a cause X and an effect Y occur together or vary together Time order of occurrence of variables the causing event must occur either before or simultaneously with the effect. Absence of other possible causal factors

Concepts Used in Experiments


Independent variables: Independent variables are also known as explanatory variables or treatments. The levels of these variables are manipulated (changed) by researchers to measure their effect on the dependent variable. (in store advertising present or absent )

Dependent variables: These variables measures the effect of treatments (independent variable) on the test units. sales in case Test units: Test units are those entities on which treatments are applied. eg : stores

Experiment: An experiment is executed when the researcher manipulates one or more independent variables and measures their effect on the dependent variables while controlling the effect of the extraneous variables. Extraneous variables: These are the variables other than the independent variables which influence the response of test units to treatments.

Examples: Store size, government policies, temperature, food intake, geographical location, etc.

Experimental Design : a set of procedures specifying : - The test units - Independent variables - Dependent variables - How to control extraneous variables

Validity in Experimentation
Internal validity: Internal validity tries to examine whether the observed effect on a dependent variable is actually caused by the treatments (independent variables) in question. External validity: External validity refers to the generalization of the results of an experiment. The concern is whether the result of an experiment can be generalized beyond the experimental situations.

Environments of Conducting Experiments


Laboratory Environment - In a laboratory experiment, the researcher conducts the experiment in an artificial environment constructed exclusively for the experiment. Field Environment - The field experiment is conducted in actual market conditions. There is no attempt to change the real-life nature of the environment.

A Classification of Experimental Designs

Pre-experimental design
Pre-experimental designs do not make use of any randomization procedures to control the extraneous variables. Therefore, the internal validity of such designs is questionable. One-shot case study: X O

One-group pre-testpost-test design:


O1 X O2 Static group comparison:

Group 1 - X O1
Group 2 O2

Three types of Experimental Designs


1. Before After Without Control
Sales before exposure to Experimental Variable
time

sales after exposure to EV

Example
Glaxo wants to measure the effect of introducing music into its work areas to the output of its employees. Applied in select workshops . Workers in selected workshops ---independent variables Music -----Treatment Output of workers ------ Dependent variable

Example glaxo
Workers output workers output prior to music music with music 840 lakh units (treatment) 890 lakh units Effect of treatment = 50 lakh units per day

Quasi-experimental designs
In quasi-experimental design, the researcher can control when measurements are taken and on whom they are taken. However, this design lacks complete control of scheduling of treatment and also lacks the ability to randomize test units exposure to treatments.

Time series design: O 1 O2 O3 O 4 X O 5 O6 O7 O8

Quasi-experimental designs

Quasi-experimental designs
Multiple time series design:

Experimental Group: O1 O2 O3 O4 X O5 O6 O7 O8 Control Group: O1 O2 O3 O4 O5 O6 O7 O8

True experimental designs


In true experimental designs, researchers can randomly assign test units and treatments to an experimental group. Here, the researcher is able to eliminate the effect of extraneous variables from both the experimental and control group. Pre-testpost-test control group: Experimental Group: R O1 X O2 Control Group: R O 3 O4

Before After with Control


Experimental group Sales before exposure to time Experimental Variable in area 1 Control Group Sales before exposure to time Experimental Variable in area 1

sales after exposure to EV in area 1

sales after exposure to EV in area 1

Example
To measure the impact of clearance sale One experimental unit is selected (apparel in a retail store) sale measured (5 lakh units measured) - clearance sale held sale measured(7 lakh) One control unit is selected sale measured(3 lakh units) clearance sale not held sale measured (4.5 lakh units)

Sale before the experiment in the experimental unit : ES1 = Rs. 5 lakhs per month
Sale after the experiment in the experimental unit : ES2 = Rs. 7 lakhs per month Sale before the experiment in the control unit : CS1 = Rs. 3 lakhs per month Sale after the experiment in the control unit : CS2 = Rs. 4.5 lakhs per month

Change in sale in experimental unit over two periods = ES2 ES1 = Rs. 2 lakhs Change in sale in the control unit over two periods CS2 CS1 = Rs. 4.5 lakhs Rs. 3 lakhs = Rs. 1.5 lakhs Treatment Effect : ( ES2 ES1) ( CS2 CS1) = Rs. 2 lakhs - Rs. 1.5 lakhs = Rs. 0.5 lakhs

True experimental designs


Post-test only control group design: Experimental Group: R X O1 Control Group: R O2

Solomon four-group design:


Experimental Group 1 : Control Group 1: Experimental Group 2: Control Group 2: R O1 X O2 R O3 O4 R X O5 R O6

After only with controls


EG CG = effect of treatment Example : Impact of free samples on sales Experimental Group : sample of soap mailed in a territory ; Rs 5 off on one cake of soap coupon mailed. Sale = 400 coupons

Control Group : Rs 5 off on one cake of soap coupon mailed to equal no.. Sale = 250 coupons

Effect of treatment = 150 coupons

Applications
Product Design Package Design Pricing Policies Promotion Policies (product) Distribution Policies

Formal Experimental Designs


Randomised Block Design Latin Square design Factorial Design

Randomized Block Design


There is only one major external variable such as sales, store size, income etc. Blocking by one variable

Which advt is the most effective ? Commercial A - No humor Commercial B - Some humor Commercial C - high levels of humor
Block No.
1 2 3

Store Comm A Patronage


HEAVY MEDIUM LOW A A A

Comm B
B B B

Comm C
C C C

NONE

Latin Square Design


Statistical control of two variables . Each external or blocking variable is divided into an equal number of blocks or levels. The independent variable is also divided into the same no. of variables

Equal no. of rows, columns and treatment levels


Interest in the store
Store patronage High Medium Low and None High B C A Medium A B C Low C A B

Factorial Design
A factorial design is used to measure the effects of two or more independent variables at various levels. So, a 3 * 3 = 9 cells. Thus, 9 different commercials would have to be aired each having a specific level of store information and humor.

Amount of store information Low Medium High

Amount of humor

No humor A D G

Medium humor High humor B E H C F I