5 Star Life

In this blog, I have noted my fascination with incentives. So I want to ‘step out’ a particular set of incentives.

Now this exercise is good for my mind (the brain is a muscle) but also good for my health. Last week, I walked with a client to coffee in the city and while we were starting to cross a street, a car ran the red light. This car wasn’t speeding but it wasn’t slowing down. It just breezed through the light, the intersection and our intended path. The cars that were moving through their green light stopped (luckily) and honked their horn. My client had also hung back from crossing so we weren’t in any major danger.

So why might incentives be involved here?

What if I add that the car had an Uber sticker on it?

What is the incentive system of an Uber driver? The major incentive is, likely all small business owners (as they are classified as small business owners) is to get paid. But what is the actual system of incentives for an Uber driver?

The system is the star rating. Basically, an Uber driver stays on the road and makes money when it maintains its star rating. An intended outcome is that the driver has to deliver a 5 star ride to each customer.

What are the advantages of this system?

• Customer focused
• Simple and easy to monitor
• Difficult to game (i.e. drivers cannot easily pump their rating)

So, on the face of it, the incentive system looks great for the customer and the company. But what might it do to a driver who is interacting with the range of entities that he encounters during his day?

The road is a complex set of second to second decisions, particularly when there is dense traffic and lane changes involved. A driver is constantly balancing their objectives with a range of other considerations including general safety.

Does a myopic focus on getting 5 star reviews create moral hazard within the Uber Drivers?

Let’s step it out:

1. I have to respond to alerts on my phone to ensure that I am the driver to pick up a person
2. I have to then maneuver the car as quickly as possible to that location and then be in a position for the person to conveniently enter the vehicle
3. I am given a route by the Uber application that I am to drive
4. I drive that route meeting the needs of the customer and my own needs to get a 5 star review (and hopefully get a new ride as quickly as possible).
5. I stop as close as possible to the destination for the customer to help ensure I get my 5 star review.

Does there look like glaring issues of moral hazard? My hypothesis is that there are moments of moral hazard rather than the Uber driver being constantly in conflict. These moments are Step 1, 2 and 5. These transition periods are potentially where the need to service customer (and get paid) might override the need to be safe or aware.

• Step 1 is pure distraction from the road as responding rapidly to an alert is the basis of being paid.
• Step 2 is interesting as this can involve lane changes and other directional changes to location. For instance, a risky U-turn might substantially reduce the time to the customer while playing it safe might involve a long detour (and customers are watching via the app). But drivers who have crashes on the way to the pick up are not going to get a 5 star rating.
• Step 5 might involve a reduced concern about stopping rules (i.e. stopping in a clearway) in order to drop the customer at their destination.

So, it may be that the incentives are not misaligned when averaged across the Uber driver’s day but in key moments.

What might be done?

Most people might say ‘well… driving will be automated soon’. Sure, that will mean that people won’t make these mistakes but programmers still have build the computer responses. They have to decide how the program ‘values’ different things and 5 star rating is going to be a central element to that analysis.

So it’s worth thinking about it.

For Step 1, a simple solution is to not allow drivers to respond to alerts while they are moving. It will slow the process of responding but fixes the issue. Or the app gives the ride to a driver (rather than having them respond) which probably makes the drivers employees rather than subcontractors. This would be very possible under an automated driver system where the drivers are not paid per ride (and probably not paid at all).

It should be noted the Uber driver who ran the red light in front of me looked to be interacting with their phone, potentially to respond to an alert.

Step 2 is harder in that humans can make good decisions in balancing the competing objectives of speed and safety. A rule like reducing the number of lane changes in the pre-pickup driving phase might reduce the cognitive load and enable the driver to make better decisions. Short, multi-lane change routes are problematic. The pre-pickup position is already being used by Uber to put people into more convenient (and legal) places to be picked up.

Step 5 should also involve setting destination stops that are designated based on road signage and other elements.

The question is then ‘does this undermine the core competitive advantage of Uber?’ I have thoughts on that too…