Lyft needs more drivers: Benefits of Marginal Analysis

This activity is designed for principles courses. No prior knowledge is necessary for this activity.

There are no class size limitations.

This activity takes an estimated 15 minutes. This activity should take just one class period.

Overview

In this activity, students will analyze actual data from Lyft on the cost of recruiting new drivers and decide how to allocate additional dollars given those data. This activity demonstrates the advantages of thinking on the margin instead of using averages in a business context.

Expected Student Learning Outcomes

Students will be able to apply marginal analysis to real-world applications.

Information Given to Students

In-Class Exercise:

Around 2019, the CEO of Lyft (Logan Green) allocated money to a team of Lyft employees and told them to "bring in as many drivers as you can."*

Running advertisements through Google and Facebook were the two best options for recruiting new drivers. To help decide which company to advertise with, the team ran a data analysis summarized below:
* For the last 1000 drivers recruited, the average cost per driver was around $400 with Facebook ads and $700 with Google ads.
* For the last 25 drivers recruited, the average cost per driver was $1000 with Facebook ads and $750 with Google ads.
* Upon request, the team could gather data for the last 10,000 drivers recruited.

Imagine you are on the team and you need to decide which company to advertise with. What is the best recommendation from the list below?

A) Advertise with Facebook, since it is cheaper than Google on average for 1000 drivers recruited.

B) Advertise with Google, since it is cheaper than Facebook on average for the last 25 drivers recruited.

C) Advertise with Google, since the variation in cost per driver between the last 1000 and last 25 drivers is smaller ($700 to $750 compared with $400 to $1000), which gives more predictability in future costs.

D) Ask for average cost data on the last 10,000 drivers and choose the company with the lowest average cost, since a larger data set allows for better decision-making.

Prefatory remarks:

1) Hook/teaser: How many of you have ever used Lyft? Uber? Which is better? Any of you drive for Lyft or Uber? What made you get started? Did you ever see ads on Facebook/instagram recruiting drivers for either company?
2) Introduce concept: Marginal analysis is important in decision-making. Here's an example to highlight what is meant by "marginal thinking" compared with "average thinking" and why it can be important for business decisions.

Explanation of each answer response:

A) This option is a misconception that using the average of more data better informs decisions at the margin (in this case it does not).

B) This option gets closer to the marginal way of thinking. Since it is taking thinner slices of data it better allows us to see the trend, or rate of change (e.g. marginal analysis) to help us predict where our next dollar is best spent.

C) Stable costs can be important in certain business contexts, but in this case the goal is not stability in costs, but lowest cost per additional driver recruited (or highest value per additional dollar spent).

D) This option is a misconception that "big data" (or at least the bigger the dataset the better) can solve our problems for us.

Option (B) is the most defensible answer. This question is meant to highlight the important difference between marginal and average thinking with an actual example. Though we are looking at the average for the "last 25 drivers," it is more in line with the idea of marginal thinking compared with using the average cost of a large chunk of data in making decisions.

Some possible follow-up questions:
1) Can you think of any examples where this idea of "thinking on the margin" could be helpful instead of looking at averages?
"E.g. on average this hospital that you want to go to has 2% patient deaths over the last 20 years, lower than the average hospital deaths" (...but imagine the average was 0.1% for 19 years and 38.1% for the most recent year).

Essay question:
Your friend tells you that over the last 40 years, his university (Cartographers of the World, U) connects an average of 20% of students to jobs in their field of study. Let's assume that the average connection rate for all colleges and universities is 19%.

Given this information, does your friend have a better-than-average chance of getting a job in his field? Explain why or why not, and explain what additional data (if any) you would like to have to get a better idea of your friend's job prospects.

(Answer: the average is not the best indicator of your friend's job prospects. You would want to know the job-connection rate for each individual year for the most recent years to get a better picture of the trend).

Multiple choice question:
You are deciding between two brands of headphones which have both been available on Amazon for 3 years. Over those 3 years, Brand "A's" average review is 4.53 stars to brand "B's" 3.94 stars. In the last month, brand "A" has had many 1 and 2 star reviews while brand "B" has had many 4 and 5 star reviews. Which brand should you get?
A) "A" since it has more stars on average.
B) "B" since the most recent reviews are better than "A."
C) Indifferent since "A" is better in average stars but "B" is better in recent reviews.

(Correct answer: B)

Below is the link to a podcast providing the information for this activity:

* The scenario's information begins at minute 53:08: https://www.econtalk.org/john-list-on-scale-uber-and-the-voltage-effect/#audio-highlights
* Transcripts are also available. If you hit "Ctrl+F" in the link above and type in "53:08" in the search bar you can quickly find the example's text.
* The information came from John List, the Chief Economist at Lyft from 2018 to 2022. His Linkedin Profile: https://www.linkedin.com/in/john-list-4727b6a/