Customers are tweeting about the incorrect wait times recommended by our waitlist feature. How would you investigate this?

Approach

1. Understand the wait time feature 

2. Map out the user journey

3. Brainstorm the root causes for the poor accuracy of wait time predictor with the interviewer

4. Work with the interviewer to single out the root cause

5. Implement the fix

Understand the wait time feature

Users can book a table at a restaurant, they get a wait time estimate. So that users show up to the venue on time.

Map out the user journey 

A) Users go to yelp

B) Search for the restaurant outlet

C) Book an appointment for a service and get an estimated wait time 

D) Users go the veune on time to receive a service 

Brainstorm the root causes for the poor accuracy of the wait time prediction feature 

Process-related

A) Under-staffed restaurants: Restaurants cannot keep their promise with regards to service time when they are understaffed. They end up taking longer than expected time to prepare the table.

B) Under-capacity restaurants: Restaurants that are limited in capacity does not have lot of tolerance, therefore even a minor variations in terms of some customers taking longer time to move out of the table would lead to longer than expected wait times for customer that booked appointment on-line. 

C) Restaurant tables are not earmarked for on-line booking: When restaurants use same set of tables for on-line as well as walk-in customers, surge in walk-in customers would lead to longer than predicted wait time for the on-line customers 

Technology-related

A) Auto-approval of the on-line bookings: Restaurants can serve the new customers only when they have tables available. But there are lot of factors that affect the availiability of tables to the customers and they tend to change rapidly, therefore auto-approval of the on-line booking will not be able to factor-in the real time information 

B) Point-in-time predictor: Restaurants wait time is dependent on multitude of factors such as arrival time of the customers, departure time of the customers, dwell time of the customers inside the restaurants. They are all random and are not deterministic. 

Therefore, when yelp predicts a precise wait time, more often than not, it tend to be inaccurate

Collaborate with the interviewer to single out the root cause

I would run the potentital causes by the interviewer. If, hypothetically, interviewer aligns on the point-in-time wait time predictor being the driver of the problem, I would develop a solution for the problem.

Implement the fix

Given the high volatile nature of the restaurant dynamics, I woud built-in the tolarance into the wait time predictor. I would give a range rather an a precise estimate of the wait time. 

The time interval based of wait time estimate has a higher chances of being right for the customers. It will improve the customer satisfaction.