Lecture 4

Question 1

Which channel gets the credit with Multi-Touch attribution

Answer 1

1. Rule-based solutions
2. Model-based: Markov Chain
3. Experiments and approximating experiments

Question 2

Rule-based

Answer 2

1. last-touch attribution
2. time-decay attribution
3. first-click attribution
4. linear attribution

Question 3

Problems with last-touch attribution

Answer 3

mo credit to other channels “assisted” in the conversion

just because channel is the “last” does not mean that it deserves all the credit

Question 4

Model-based: Markov-based approach

Answer 4

memoryless property: probability of transition depends only on current state

Question 5

calculating markov model

Answer 5

1. remove the channel from the graph
2. measure conversions without channel

Question 6

What shows us the removal effect

Answer 6

gives us a way to measure the contribution of any channel in producing conversion

Question 7

calculating the removal effect

Answer 7

1 - (reduced conversions/sum all path conversions)

Question 8

Incrementality

Answer 8

a way of measurement of an effect that would have occurred

Question 9

Attribution

Answer 9

is the set of rules that determine how much credit is assigned to each channel within different touchpoints

Question 10

What is the main difference between attribution and incrementality

Answer 10

Incrementality measurement uses a statistical approach rather than trying to attribute on a singular user level.

Question 11

counterfactual

Answer 11

describes a causal situation in the form: “If X had not occurred, Y would not have occurred”

Question 12

how to calculate ATE on a dataset

Answer 12

take the difference from treated and non treated and compute the average on treated

Question 13

selection bias

Answer 13

[y,d=1]-[y,d=0]