Measuring Marketing Performance

Question 1

Estimating the effect of price promotions using transactional data

Answer 1

• In order to estimate the effect of price promotions you need to have a control group
• Historical data alone usually doesn’t provide that
• Subjects need to be treated to different prices over time

Question 2

When demand estimation goes wrong:

Gasoline example

Answer 2

Looking at historic data of gas prices, one could conclude that the prices should be raised in order to reach higher demand, as demand and prices were positively related in the past.

Look at historical data carefully: Always ask why there was a price variation in the first place.

Question 3

Why do we have price variation?

Ice cream example

Answer 3

Ice cream sales depend on price, but also on other factors, like the weather. Good weather increases demand and this in turn prices. But raising prices does not increase demand.

In order to calculate the clean price variation, all other factors need to be held constant or to be parsed out by regression.

Question 4

Good and bad price variation

Answer 4

• sales depend on price but also on other factors
• if the other factors can be observed, they can be parsed out using regression
• if they cannot be observed, there is no possibility to have a clean price variation

Question 5

Measurement of marketing performance

Answer 5

Measurement of marketing performance usually requires a control condition:

• The goal is to determine the causal effect: The difference between running the campaign or not
• Historical data seldom provides control conditions
• Other variables will be confounding the effects

Question 6

Two possibilities to get clean variation

Answer 6

Experiment:

• respects the ceribus paribus rule
• price movement is not correlated with other demand factors
• Estimation of price coefficient is not invalidated

Experiments are often costly to run

Instrumental variables:

• filters the good variation in the variable of interest
• the good variation is then independent of other demand factors

Approach requires skilled economists to be trusted

Question 7

Instrumental variables

Answer 7

1. Predict price based on good variation
2. Regress quantity on predicted price from step 1